52 pages including cover

Knowledge Digest for IT Community

ISSN 0970-647X

Volume No. 41 | Issue No. 8 | November 2017 50/-

Mac ine
www.csi-india.org

Le rning

COvER STORY RE SEARCH FRONT

Machine Learning:
Messiah of 21st Century 6
TECHNiCAl TRENDS
From Machine Learning to Deep
Learning: Trends and Challenges 10
Fuzzy Based Machine Learning: A Promising Approach 21
A RTiClE
Using statistical approach to Machine
Translation for Indian languages 25
SECURiTY CORNER
Empirical Survey on Advances of Network Forensics
in the Emerging Networks 31
Meetings with Goa Chief Minister & Goa Chamber of
commerce and Managing Committee meeting of CSI
Goa chapter
CSI President & Hon. Treasurer along with
MC Members of CSI Goa Chapter met Hon’ble
Chief Minister of Goa, Shri Manohar Parikkar
and apprised him of CSI and its activities.
President & Hon Treasurer of CSI along with
MC Members of CSI Goa Chapter also met
officials of Goa chamber of Commerce

National Seminar on “Role of ICT for Reaching to

Unreached”
A one day seminar was organised at C. M. Science College of L. N.
Mithila University, Darbhanga, Bihar with the technical support of CSI
Patna Chapter on the theme “Role of ICT for Reaching to Unreached”
on 13 October 2017. The Seminar was inaugurated by Prof. (Dr.) S.
K. Singh, the Hon’ble Vice Chancellor of L. N. Mithila University,
Darbhanga, Bihar and Prof. (Dr.) A. K. Nayak, National Secretary of
Computer Society of India & Director Indian Institute of Business
Management, Patna delivered the key note Address. Prof Shams Raza,
Past Chairman CSI, Patna Chapter; Dr. Arvind Kumar, Principal; Prof.
(Dr.) S. N. Singh, Head, Dept. of Physics and Dr. Ajay Thakur of Physics
department also spoke.
 CSI CommunICatIonS Volume No. 41 • Issue No. 8 • NoVemBeR 2017

Contents
Chief Editor
S S AgRAwAl
KIIT Group, Gurgaon
Editor Cover Story
PRAShANT R. NAIR
Amrita University, Coimbatore Machine Learning: Messiah of 21st Century 6
Tawseef Ayoub Shaikh and Rashid Ali
Published by
A. K. NAYAK Technical Trends
Hony. Secretary
From Machine Learning to Deep Learning: Trends and Challenges 10
For Computer Society of India Apoorva Mishra and Anupam Shukla

Editorial Board:
Machine Learning and its application in Technology and Media industries 13
Arun B Samaddar, NIT, Sikkim Krishna Murthy
Bhabani Shankar Prasad Mishra,
KIIT University, Bhubanewar
Debajyoti Mukhopadhyay, MIT, Pune
Research Front
J. Yogapriya, Kongunadu Engg. College, Trichy Challenges of Theoretical Computer Science and Computing Systems 16
M Sasikumar, CDAC, Mumbai,
R. Ponnusamy
R Subburaj, SRM University, Chennai Fuzzy Based Machine Learning: A Promising Approach 21
Sujamol S., Sreeja Ashok and U Krishna Kumar
R K Samanta, SiliguriInst. of Tech., West Bengal
R N Behera, NIC, Bhubaneswar
Sudhakar A M, University of Mysore
Articles
Sunil Pandey, ITS, Ghaziabad Using statistical approach to Machine Translation for Indian languages 25
M. Sasikumar and Prakash B. Pimpale
Shailesh K Srivastava, NIC, Patna
Machine learning at the Edge of Internet of Things 28
Vishal Mehrotra, TCS
Mohammad Khalid Pandit, Roohie Naaz Mir and Mohammad Ahsan Chisthi

Security Corner
Design, Print and Dispatch by
Empirical Survey on Advances of Network Forensics in the Emerging Networks 31
gP OFFSET PvT. lTD. A. R. Jayakrishnan and V. Vasanthi

P l US
Meetings with Goa Chief Minister and Goa Chamber of commerce 2nd Cover
Please note: CSI Members win International recognition in SEARCC 12
CSI Communications is published by Computer
Society of India, a non-profit organization. Obituary - Dr. A. Shamugasundaram 12
Views and opinions expressed in the CSI
Communications are those of individual CSI Executive Committee (ExecCom) meeting at Nashik on 7 October 2017 20
authors, contributors and advertisers and they
may differ from policies and official statements Past President visit to Shambhunath Inst. of Engineering & Technology 24
of CSI. These should not be construed as legal
or professional advice. The CSI, the publisher,
“GeekExpo” - Project Exhibition 27
the editors and the contributors are not Report of ‘3rd IT Excellence Awards 2017’ by CSI Ghaziabad Chapter 37
responsible for any decisions taken by readers
on the basis of these views and opinions. CSI Young IT Professional (YITP) Awards 2017 – A Report 38
Although every care is being taken to ensure
genuineness of the writings in this publication,
Pilgrims in the Digital World – Book Review 41
CSI Communications does not attest to the CSI Reports 42
originality of the respective authors’ content.
© 2012 CSI. All rights reserved. Student Branches News 44
Instructors are permitted to photocopy isolated
articles for non-commercial classroom use
52nd Annual Convention of Computer Society of India 3rd Cover
without fee. For any other copying, reprint or Call for Papers-International Conference on Social Transformation - Digital Way Back Page
republication, permission must be obtained
in writing from the Society. Copying for other Printed and Published by Prof. A. K. Nayak on behalf of Computer Society of India, Printed at G.P. Offset Pvt. Ltd.
than personal use or internal reference, or of
articles or columns not owned by the Society
269 / A2, Shah & Nahar Industrial Estate, Dhanraj Mill Compound, Lower Parel (W), Mumbai 400 013 and published from
without explicit permission of the Society or Computer Society of India, Samruddhi Venture Park, Unit-3, 4th Floor, Marol Industrial Area, Andheri (East), Mumbai 400 093.
the copyright owner is strictly prohibited. Tel. : 022-2926 1700 • Fax : 022-2830 2133 • Email : hq@csi-india.org

3
CS I C o mm u n I Ca t I o n S | n o VE m B ER 2 0 1 7
Editorial
Dear Fellow CSI Members,
The theme for the Computer Society of India (CSI) Communications (The Knowledge Digest
for IT Community) November 2017 issue is Machine Learning
“The human condition is plagued with a labyrinth of shortcomings, frailties and limitations that
hinder man from reaching his fullest potential. Therefore, it only makes sense that we find
ourselves at the next phase in human evolution where restricted man merges with the infinite
possibilities of hyper-evolving technologies. This techno-human transmutation will prove to be
‘the’ quantum leap in human progression. The harmonization of technologically extending oneself,
Prof. (Dr.) S. S. Agrawal consciousness, artificial intelligence and machine learning will reverse the failures of genetic Prof. Prashant R. Nair
Chief Editor Editor
predisposition and limitation.”
James Scott, Senior Fellow, Institute for Critical Infrastructure Technology

After a series of thematic issues focusing on ICT in applications such as education, governance, agriculture and health, CSI
Communications is focusing on machine learning in this issue. Machine learning is a field of computer science that gives
computers the ability to learn without being explicitly programmed. Machine learning is employed in a range of computing tasks
where designing and programming explicit algorithms with good performance is difficult or infeasible; example applications
include email filtering, detection of network intruders and computer vision.
The Cover story in this issue is Machine Learning: Messiah of 21st Century by Tawseef Ayoub Shaikh and Rashid Ali. In this article,
the authors have traced the evolution of machine learning and also classified the same.
The technical trends showcased are “Machine Learning and its application in technology and media industries” by Krishna Murthy
and “From Machine Learning to Deep Learning: Trends and Challenges” by Apoorva Mishra & Anupam Shukla
In Research front, we have two contributions, One is “Fuzzy Based Machine Learning: A Promising Approach Towards Better
Decision Making” by Sujamol.S, Sreeja Ashok & U Krishna Kumar. The other one is “Challenges of Theoretical Computer Science
and Computing Systems” by R.Ponnusamy.
Other articles in this issue on machine learning provide us information on applications, of machine learning in healthcare, I ndian
languages, IoT and bioinformatics.
The articles are as follows:
¬ “Using statistical approach to Machine Translation for Indian languages” by M. Sasikumar & Prakash B. Pimpale
¬ “Machine learning at the Edge of Internet of Things” by Mohammad Khalid Pandit, Roohie Naaz Mir & Mohammad Ahsan
Chisthi
In Security front, A R Jayakrishnan & V.Vasanthi have performed an “Empirical Survey on Advances of Network Forensics in the
Emerging Networks”
This issue also contains reports of CSI Young IT Professional (YITP) Awards 2017 National Round Reports, Book Review and CSI
2017 convention notifications, CSI activity reports from chapters & student branches and calendar of events
We are thankful to entire ExecCom for their continuous support in bringing this issue successfully.
We wish to express our sincere gratitude to the CSI publications committee, editorial board, authors and reviewers for their
contributions and support to this issue.
We look forward to receive constructive feedback and suggestions from our esteemed members and readers at
csic@csi-india.org.

With kind regards,

With kind regards,

Prof. (Dr.) S. S. Agrawal Prof. Prashant R. Nair

Chief Editor Editor

www.csi-india.org
4
C S I C o mm u n I C a t I o n S | n o V E m B ER 2 0 1 7
 Computer Society of India
http://www.csi-india.org
Mr. Sanjay Mohapatra Shri.gautam Mahapatra Prof. A. K. Nayak
President Vice President Hon. Secretary
President’s Message
From : President, Computer Society of India
Date : 01 November, 2017
Email : president@csi-india.org / Cell : (91) 9861010656
Season’s Greetings!
Meeting with goa Chief Minister
President & Hon Treasurer of CSI along with MC Members
of CSI Goa Chapter met Hon’ble Chief Minister of Goa, Shri
Manohar Parikkar and apprised him of CSI and its activities.
Hon’ble CM advised CSI to work on e-waste management
system and assured to provide all support at Goa for promotion
of IT / ITES.
ExecCom & CSI YITP at Nashik on 8th October 2017
CSI ExeCom meeting took place at Nashik on 8th October
2017. CSI ExeCom members along with Chair, Publications
Committee, Chief Editor of CSIC, CSI NSC, NC Chair &
Members participated in ExecCom and many issues for the
improvement of CSI were discussed and deliberated.
CSI Young IT Professional (YITP) Final round also took place at
Nashik on 8th Nov 2017 and almost 14 top teams from regional
rounds from different regions participated. Detailed report of
YITP is published in this issue of CSI Communications.
Meeting with goa Chamber of commerce
President & Hon Treasurer of CSI along with MC Members of
CSI Goa Chapter met officials of Goa chamber of Commerce.
It was decided that CSI Goa Chapter and Goa Chamber of
Commerce will organize events jointly for promotion of IT at
Goa and will also have inter-exchange membership activities.
visit to CSI goa Chapter
President & Hon. Treasurer visited CSI Goa chapter on 5th
October 2017. It was found that, CSI Goa Chapter premises
is having some legal (title) issues with property. CSI office
space is in the 3rd Floor of the building, which is in very bad
shape. Urgent renovation is needed failing which building
may collapse at any time. This issue was discussed with MC
of Goa Chapter and it is resolved to move ahead upon earlier
ExeCom’s decision during 2011/2012 to sell the property and
buy a new one. Current MC of Goa Chapter requested CSI
President for a loan of Rs.30 lakhs for the purchase of new
premises to augment the proceeds of the sale of the existing
building. President updated MC of Goa Chapter to take this
issue to CSI ExecCom for final approval.
International Participation
CSI is a member of At Large structure (ALS) of ICANN,
since 2012. ICANN is the community of individual Internet
users (known as the At-Large Community) of the Internet
Corporation for Assigned Names and Numbers (ICANN). At
Abu Dhabi, ICANN General Assembly of APRALO is scheduled
for 28 Oct - 3 Nov 2017. Detailed report will be published in
5
C S I C o mm u n I C a t I o n S | n o V E m B ER 2 0 1 7
TM
Mr. Manas Ranjan Pattnaik Dr Anirban Basu
Hon. Treasurer Immd. Past President
next issue of CSIC.
Meeting with
CSI Publications
Company
A discussion was
held on 14th October
2017 with Director s
of CSI Public ations
Company. CSI
President request ed
the directors of
Publications Comp any & Past Presidents of CSI, Sri S
Mahalingam & Prof. S V Raghavan that all issues of the
journal, CSI Transactions on ICT in pdf format will uploaded
in CSI portal for the benefit of its members. CSI will also
place an order to CSI Publications Company for 500 copies
of CSI Transactions on ICT for dispatching to all Institutional
members of CSI. This will help the brand building of CSI. It was
also decided to explore ways for a sustainable model of CSI
Transactions on ICT. Balance sheet of Publications Company
was discussed and it was proposed to conduct one Editors
Consortium. CSI will invite all Editors of different publications
of CSI for this event.
visit to CSI Educational Directorate
President & Hon Treasurer visited CSI Educational Directorate
(ED) on 14th October 2017 and it was found that, CSI ED building
condition is deteriorating and needs urgent repairs. Boundary
condition of building has collapsed and roof condition is also
not good. During last cyclone, parapet boundary had fallen
and as advised by ExeCom, it was decided do necessary
repairs. A budget of Rs.20 to Rs.25 lakhs has been earmarked
for repair, maintenance and painting of the CSI ED building.
Hon Treasurer will be overseeing this. President, CSI also
persuaded with CSI Senior Members at Chennai for renewal
of Land at Taramani.
Meeting with CSI Chennai Chapter Chairman & hon
secretary on 14th October 2017
Meeting with Chairman & Hon Treasurer of CSI Chennai
Chapter was held at CSI ED and discussions were carried on
for expanding the Chennai Chapter activities.
With kind regards
Sanjay Mohapatra
President, CSI
 COvER STORY
Machine Learning: Messiah of 21st Century
Tawseef Ayoub Shaikh and Rashid Ali
Department of Computer Engineering, Aligarh Muslim University, Uttar Pradesh, India
tawseef37@gmail.com, rashidaliamu@rediffmail.com
1. Introduction
“People worry that computers
will get too smart and take over
the world, but the real problem
is that they’re too stupid and
they’ve already taken over the
world.” -Pedro Domingos
Since the time humans landed their
feet on the earth, they have been always
trying to improve their life in one or the
other way around. Counting from the
earlier discoveries of the fire, clothing,
food, houses etc., the human civilization
was always in effort to through light on
new innovative ideas which will make
their lives comfortable and sustainable.
In this phase of continues development
humanity has seen a revolution in
industrial sector in the 18th century
when they discovered the mechanical
engine which later on took many shapes
and eased humanity in capturing the
light year distances in few units of
time. This phase was the dramatic shift
in which the muscular energy of the
living beings were transformed into the
machines which are doing many times
the amount of work the humans could
do in very less amount of time. Be it the
automobile engines, aero planes, ships
and other types of machines commonly
used in day to day life. From the time
man has landed in to the ocean of
Information Technology sector, a new
revolution has stepped in the human
history and a rigorous effort is been
applied to impart the Intelligence inside
the hearts of the same machines. This
continues evolving effort has given rise
to what is commonly known as Artificial
Intelligence (AI). Machine Learning
(ML) being at the core of Artificial
Intelligence (AI) is the thrust and sexiest
research area in the 21st century. Since
the human brain can’t remember the
information for large periods of time
and is computationally hard for them
to perform any work on a large amount
of data. Computers (Machines) on
C S I C o mm u n I C a t I o n S | n o V E m B ER 2 0 1 7
the other hand have got the inherent
ability to remember the information for
Machine Learning
Suppervised Unsuppervised Reinforcement
Machine Learning Machine Learning
Fig 1: Machine learning (Ml) types
billions and billions of years and could
also perform any hard computational
task in milliseconds. This made them
a perfect candidate to make them
learn so that they can automatically
perform the tasks which human brain
was lacking, thus outputting the most
debatable topic of machine Learning
(ML), nowadays.
2. Machine learning
Machine Learning (ML), a term
coined by Arthur Samuel coined in 1959
at IBMisabranchof Artificial Intelligence
in which machine is trained so as to
impart it the ability to automatically
learn and improve from the experience
without being explicitly programmed.
Its focus lies in the development of the
intelligent programs that can access a
particular amount of the data (training
data) and is then used to automatically
work on the new data (test data)
which it has never met with. Machine
Learning is a broad multi-disciplinary
filed which has its roots in Statistics,
Algebra, Data Mining, Data Analytics
etc., which makes it difficult to come up
with a unique definition of it. According
to Tom M. Mitchell’s [1] definition of
Machine Learning (ML) which defines
it as:, “ A computer program is said to
learn from experience E with respect to
some class of tasks T and performance
measure P if its performance at tasks
6
CSI HIStory SerIeS
in T, as measured by P, improves with
experience E”. Learning is used when a
Learning
Active Learning Active Learning
pattern exists in the given data which we
can’t pin down manually. It is because
of the Machine Learning that the
present century has been witness to the
landmark discoveries like Autonomous
Helicopter, Handwriting Recognition,
Natural Language Processing (NLP),
Computer Vision, Speech Recognition,
Recommendation Systems, Decision
Support Systems (DSS), email filtering,
detection of network intruders or
malicious insiders working towards
a data breach, optical character
recognition etc. Machine Learning is
at the root of the tree and has lots of
branches and sub branches. Broadly
speaking, this fantastic branch of
innovation is divided into following
categories:
1. Supervised Machine Learning
2. Unsupervised Machine Learning
3. Reinforcement Learning
4. Active Learning
5. Online Learning
1) Supervised Machine Learning:
This is the learning [2] under
a supervision that is why it is
also known as learning from a
Teacher. Here the input data with
its corresponding desired output is
provided to the computer program
for the purpose of making it learn.
Dataset {(x1, y1), (x2, y2)….(xn, yn)} is
inputted to the program and the
www.csi-india.org
 COvER STORY
goal is to build a model in order
to find a target function y=f(x) so
as to map input values to their
corresponding output values.
Supervised Machine Learning is
further divided into two categories:
Supervised Machine learning
Classification Regression
Fig 2: Supervised Machine learning (Ml)
types
1(a) Classification : In Classification
the output predicted value ‘y’ is
categorical/discrete in nature and
the right answers are neatly given.
E.g. given the input breast cancer
data of the patients containing
various features like tumor size,
tissue lesion color, the machine
can learn and predict on the new
datasets whether the patient is
suffering from Breast cancer or
not. It is used in:
Email: Spam /Not Spam?
Online Transactions: Fraudulent
(Yes / No)?
Tumor: Malignant / Benign?
Classification in general terms is
like YES, NO (0, 1 or PREENCE,
ABSENSENEEor TRUE, FALSE) type
of learning. Logistic Regression,
Support Vector Machines (SVM),
Decision Trees, Random Forests
are the techniques which lie under
the realm of Classification. The
output predicted value can also
be in multiple classes giving rise
to what is known as Multiclass
classification. e.g.
Email foldering/tagging: Work,
Friends, Family, Hobby
Medical diagrams: Not ill, Cold, Flu
Weather: Sunny, Cloudy, Rain,
Snow
1(b) Regression:Inittheoutputpredicted
value ‘y’ is continuous/real valued
in nature. e.g. in predicting the
price of the house given its input
parameters like location, number
of rooms, furnished, area etc.,
the corresponding house price is
predicted. Given ‘m’ as number
of training examples, ‘x’ as input
variable and ‘y’ as output/target
C S I C o mm u n I C a t I o n S | n o V E m B ER 2 0 1 7
Binary Classification Multi Class Classification
X2 X2
X1 X1
Fig 3: Binary Classification vs. Multi level Classification
variable, the goal is to find a cost
function f(x->y), which will minimize Logistic Regression
the difference between the actual
value and the predicted value. Decision Trees
m= Number of training examples
Support Vector
X’s= “input” variable / features Machine (SVM)
Y’s= “output” variables / features
Random Forests
ii) Unsupervised Machine Learning
(Clustering): This is the learning
[3] where no external supervision is Artificial Neural Network
required. The computer program is
Supervised learning
built in a way that it automatically Analytical Learning
finds the useful patterns in the
data based upon a similarity matrix Bayesian Statistics
and groups them together into
Clusters. This is the most practical Boosting
form of the machine learning since
in the real life most of the data Kernel Function
comes in unlabeled form. It is
used in Google News, Genomics, Ensemble Classifiers
Social Network Analysis, Market
Segmentation, Astronomical Data Conditional Random Field
Analysis, and Cock Tail Party
Problem. It is further divided into Minimum Complexity
its child branches as: Machines
ii(a) Clustering: Finds a relation among
the similar data and groups them etc.
together in a cluster. The similarity
measure normally used is Euclidian
distance. Other similarity measures
Fig 4: Supervised Ml Approaches
like Mahalanobis distance, Cosine Rank based, Correlation based,
Similarity, Hamming Distance, Manhatten, Jaccard can also be
Unsupervised Ml
Clustering hierarchical Clustering gaussian Mixture Models
Fig 5: Unsupervised Machine learning (Ml) types
7
 COvER STORY

used. space. used to update our best predictor for

future data at each step, as opposed
ii(b) Hierarchical Clustering: Here the
data is being grouped in small to batch learning techniques
State Value Function
clusters based upon a similarity which generate the best predictor
by learning on the entire training
measure and the corresponding Brute Force data set.
sub clusters are then again
clustered based upon some similar Value Function
parameter until some convergence Batch learning

Supervised learning
condition is satisfied.
Monte Carlo Methods Online learning: recursive
ii(c) Gaussian Mixture Model: Each least squares
mixture component represents a Direct Policy Search
unique cluster specified by:{πk, μk,
Stochastic gradient descent

Online Ml
σk}, where ‘πk’ is the data inputted,
Temporal difference
‘μk’ is its mean and ‘σk’ is the
Methods
corresponding variance. Incremental stochastic
gradient descent
Singular Value
K-Means
Decomposition (SVD) Progressive learning
Hierarchical Clustering
etc. Online convex optimisation
Gaussian Mixture Models Fig. 7: Reinforcement Ml Approaches
(GMM) Follow the leader (FTL)

iv) Active Learning: Active learning etc.

Expectation Maximization
[5] is a special case of semi-
(EM) supervised machine learning in
which a learning algorithm uses Fig. 9 : Online Ml Approaches
Neural Network the mechanism of query the user
Supervised learning

(or some other information source)

Anomaly Detection to interact and obtain the desired 3. Machine learning Tools
outputs at new data points. In Lot of tools with new novel machine
Hebbian Learning statistics literature it is sometimes learning algorithms implemented in
also called optimal experimental different languages in time to time
design. phase. These tools are being used as
Principal Component
a backbone for performing any data
Analysis (PCA)
Uncertainty sampling analytics task by using the algorithms
implemented in it and they also provide
Independent Component a healthy platform to develop new data
Analysis (ICA) Expected error reduction
analytics algorithms for tackling the
new challenges of this data rich age.
Supervised learning

Singular Value Balance exploration and Below is a list of the machine learning
Decomposition (SVD) exploitation tools:
3.1) Some open Source tools/
etc. Query by committee
Software’s
Weka, CNTK, KNIME, RapidMiner,
Expected mode change Deeplearning4j, R, Mahout, H2O, GNU
Fig. 6 : Unsupervised Ml Approaches
Octave, MOA (Massive Online Analysis),
Exponentiated Gradient Tanagra, Orange, Python, Shogun,
iii) Reinforcement Learning: In it a Exploration for Active TensorFlow, Torch etc.
computer program [4] is given Learning
3.2) Some Proprietary software
access to a dynamic environment
MATLAB, Neural Designer,
in for performing a certain goal Fig. 8 : Active Ml Approaches
Mathematica, Google Prediction API,
(such as playing a game against
Oracle Data Mining, NeuroSolutions,
an opponent or driving a vehicle). v) Online Learning: Online machine Amazon Machine Learning,
Feedback in terms of rewards and learning [6] is a method of machine Microsoft Azure Machine Learning,
punishments is provided to the learning in which sequential order STATISTICA Data Miner, Splunk, SAS
program as it navigates its problem of data availability is followed and is Enterprise Miner etc.
www.csi-india.org
8
C S I C o mm u n I C a t I o n S | n o V E m B ER 2 0 1 7
 COvER STORY

4. Machine learning Applications In short Machine Learning is going 2004.

Machine Learning has now laid its
footprints on every sector of human life.
Hardly anything is there which has not
come under the domain of ML. Since
data lies at the heart of ML. And with the
rapid increase in the number of smart
phones across the globe, the interaction
between humans and computer and
machines and machine interaction
is also producing data at a very fast
exponential growing pace. Following
are some of the directions where ML is
showing its positive promise:
Medical diagnosis, Bioinformatics,
Cheminformatics, Sequence mining,
Recommender systems, Computer
vision, Search engines, Natural
language processing, Natural
language understanding, Robot
locomotion, Brain–machine interfaces,
Classifying DNA sequences, General
game playing, Information retrieval etc.
5. Conclusions
Machine Learning has already
stepped in the competition of the human
beings. The intelligent machine like
the IBM Watson, modern sophisticated
Robots are all set to ease the human
life by performing lot of complex tasks.
In a sense we have reached in an age
where it brings a picture that a silent
competition is going on between human
intelligence and artificial intelligence.
to dramatically change the map of
the human life particularly in the field
of healthcare where a lot of effort
and research is going on to develop
intelligent CAD (Computer Assisted
diagnosis) which are supposed to
diagnosis a particular disease well
advance in time that can lead to its early
cure and hence will save lot of live by
providing high healthcare quality. In
short, Machine Learning is the Messiah
of this present data Analytics age.
Refrences
[2] T. Mitchell, “Machine Learning”,
McGraw Hill. p. 2, 1997, ISBN 0-07-
042807-7.
[3] S. B. Kotsiantis, “Supervised Machine
Learning: A Review of Classification
Techniques”, Informatica Vol-31, pp:
249-268, University of Peloponnese,
Greece, 2007.
[4] Z. Ghahramani, “Unsupervised
Learning”, Advanced Lectures on
Machine Learning LNAI 3176, Springer,
pp:1-32, Gatsby Computational
Neuroscience Unit University College
London, UK, 2004.
[5] LP. Kaelbling and AW. Moore,
“Reinforcement Learning A Survey”,
Journal of Artificial Intelligence
Research, Vol-4, pp: 237-285, 1996.
[6] Y. Baram, Ran E. Yaniv and K. Luz,
“Online Choice of Active Learning
Algorithms”, Journal of Machine
Learning Research, Vol-5, pp: 255–291,
[7] S. Hoi, J. Wang and P. Zhao, “LIBOL: A
Library for Online Learning Algorithms”,
Journal of Machine Learning Research,
Vol- 15, pp: 495-499, 2014.
Some online free Tutorials for
Machine Learning:
[1] Machine Learning by Andrew Ng,
Stanford University (www.coursera.org/
learn/machine-learning)
[2] Intro to Machine Learning by Sebastian
Thrun and Katie Malone (https://
in . u d a c it y . c o m / c o u r s e / i n t r o - to -
machine-learning--ud120)
[3] Learning from Data (Introductory
Machine Learning) by Yaser S. Abu-
Mos t afa ( w w w . e d x . o r g / c o u r se/
learning-data-introductory-machine-
caltechx-cs1156x-0#!)
Some best books for Machine
Learning:
[1] Learning From Data by Yaser S. Abu-
Mostafa, Malik Magdon-Ismail and
Hsuan-Tien Lin (AMLbook.com)
[2] Machine Learning: A Probabilistic
Perspective by Kevin R Murphy
[3] The Elements of Statistical Learning by
Jerome H. Friedman, Robert Tibshirani
and Trevor Hastie (https://web.stanford.
edu/~hastie/Papers/ESLII.pdf)
[4] Patterns Classification by Richard O.
Duda and Peter E. Hart
[5] Information Theory, Inference and
Learning Algorithms by David J.C.
MacKay (http://www.cambridge.
org/0521642981)
n

About the Authors

Tawseef Ayoub Shaikh, research scholar in department of computer engineering, Zakir Hussain College of Engineering &
Technology (ZHCET), Aligarh Muslim University Aligarh India. His interests are AI, Machine Learning, Data Analytics and is
presently working in Biomedical data analytics field.

Dr. Rashid Ali, Associate Professor at department of computer engineering, Zakir Hussain College of Engineering &
Technology (ZHCET), Aligarh Muslim University Aligarh India. Dr. Rashid Ali is a senior IEEE member in addition of having life
time membership of Computer Society of India. His interests are AI, Web Mining, and Soft Computing etc.

9
C S I C o mm u n I C a t I o n S | n o V E m B ER 2 0 1 7
 TEChNICAl TRENDS

From Machine Learning to Deep Learning:

Trends and Challenges
Apoorva Mishra* and Anupam Shukla
Soft Computing and Expert System Laboratory, ABV -Indian Institute of Information Technology & Management, Gwalior, Madhya Pradesh, India – 474015
apoorvamish1989@gmail.com; dranupamiiitm@gmail.com

Overview of Machine learning

In the present world, machine
learning is omnipresent. You might
have used it several times even without
knowing that you are using it. Machine
learning (ML) is a branch of artificial
intelligence (AI) which endeavors
to impart computers (or any other
electronic machine) the ability to learn
without being explicitly programmed
[2]. It deals with the study and design
of algorithms that can learn from and
make predictions on data. It is a data
analysis approach that automates
analytical model building. With the
improvement in hardware technology,
the use of machine learning algorithms
is increasing day-by-day. Many sectors
like healthcare, finance, pattern
recognition, traffic management and
route planning, robotics and Unmanned
Aerial Vehicles (UAVs), decision support Fig. 2 : Classification of Machine learning Techniques
systems for business intelligence,
etc. have been revolutionized by Classification of Machine learning  Neural Networks
the application of Machine learning Techniques  Naive Bayes Classifier
algorithms. Deep learning [5-6] is one Machine learning algorithms can  Decision Trees
of the recent and fastest growing areas be broadly classified into the following  Linear Regression
of Machine learning and is a buzz word four categories as depicted by figure-2. • Unsupervised learning
in industries as well as in academia. 1. Supervised learning In unsupervised learning, no
The emergence of deep learning has 2. Semi-supervised Learning element of supervision is involved. The
also led to the renaissance of neural 3. Unsupervised Learning computer is trained with unlabeled
networks. The relationship among deep 4. Reinforcement Learning data [4]. Unsupervised learning plays
learning, machine learning and artificial • Supervised learning a vital role in those cases where the
intelligence is represented by the Venn Supervisedlearningisverycommon human expert doesn’t know what to
diagram as shown in Fig. 1. in classification problems, where the look for in the data. The common types
aim is to classify the given input data or of supervised learning approaches are:
AI
image into predefined output classes. Association Rules, K-means clustering
Machine Learning For example: Handwritten Digit and etc-.
character recognition, classification • Semi-supervised Learning
Deep learning
of animals or objects depending Semi-supervised Learning lies
upon the input image, medical image between the two approaches mentioned
classification for diagnosis of various above. In this approach a combination of
diseases, etc-. In supervised learning, labeled and unlabelled data is used for
there is some supervision while the training.
algorithm is being trained. Here, labeled
data is used for training. Different types • Reinforcement Learning
Fig. 1 : Representation of Deep learning,
of supervised learning approaches are Reinforcement Learning involves
Machine learning and Artificial
Intelligence as follows: the mechanism of reward and

www.csi-india.org
10
C S I C o mm u n I C a t I o n S | n o V E m B ER 2 0 1 7
 TEChNICAl TRENDS
punishment for the process of learning.
In this type of learning, the objective is
to maximize the reward and minimize
the punishment.
Emergence of Deep learning
Deep learning has emerged as the
hottest area of Machine learning [8-11],
which is based on a group of algorithms
in which high level abstractions of
the data are modeled by using deep
graph having various processing
layers and multiple non-linear and
linear transformations. Deep learning
promises to replace hand-engineered
features with feature extraction
techniques that are hierarchical,
unsupervised or semi-supervised [1].
In a deep neural network, there
are multiple hidden units between input
and output layers. With the advent of
big-data and powerful computational
resources, the use of deep learning
has become more widespread as
the performance of deep learning
algorithms increases with the amount
of data that is available for training,
and they also require tremendous
computational power [7].
The popular deep learning based
algorithms are as follows:
• Convolutional Neural Network
(CNN)
CNN is a variant of the feed-forward
network that involves various layers
and deeply connected architecture [3].
CNNs are trained with a modified form
of the back-propagation algorithm. They
are capable of recognizing patterns with
a lot of variability and hence are widely
used for this purpose.
• Recurrent Neural Network (RNN)
RNN is a special type of Artificial
Neural Network (ANN). It has directed
cycles between connections. The
dynamic temporal behavior is exhibited
by the internal state of the network.
Unlike ANN, arbitrary sequences can
be processed by using the internal
state of RNN. There is a time-varying
real-valued activation in every unit
and modifiable real-valued weight in
each connection. The Long short term
memory (LSTM) is an RNN based
deep learning system that overcomes
the vanishing gradient problem of
traditional RNN. The recurrent gates
called forget gates are augmented in
C S I C o mm u n I C a t I o n S | n o V E m B ER 2 0 1 7
LSTM. It is very useful for time-series
data.
Applications of Deep learning
Some of the broad application
areas of the deep learning techniques
are mentioned below.
 Medical image classification.
 Pattern Recognition
 Robot trajectory generation
 Automatic Game Playing.
 Patient monitoring in ICUs and
their mortality prediction.
 Automatic photograph description
 Feature extraction
 Intrusion detection
 Prediction in Finance
 Autonomous driving of vehicles
 Geological data analysis
 Logistics
 Speech processing
 Image Caption Generation.
 Handwritten digits and character
recognition.
 Natural language processing.
Challenges in Deep learning
 Requires tremendous amount of
data for proper training.
 Lack of generic intelligence, most
of the deep learning models work
for the specific purpose only.
 Requires too many computational
resources (like GPUs)
 Running time is very high varying
from hours to days.
 Lack of flexibility.
Conclusion
Machine learning algorithms have
been playing a pivotal role for decades
in almost every field like Finance,
Weather forecasting, Medical science,
Engineering optimization problems,
Software cost prediction, Pattern
matching etc-. With the improvement
in hardware technology and availability
of huge amount of training data, deep
learning techniques have emerged as
the most important machine learning
algorithms. Some of the popular deep
9
learning techniques are CNN, RNN,
LSTM, etc-. Although deep learning
techniques give far better results than
the shallow traditional machine learning
algorithms, there are some issues
associated with them like: the need for
huge amount of data for training, need
for expensive hardware resources and
the requirement for very high running
time. Despite these limitations, deep
learning algorithms are very useful and
are being widely used for solving various
problems related to different fields.
References
[1] Karpathy and L. Fei-Fei, Deep visual-
semantic alignments for generating
image descriptions,” in Proceedings
of the IEEE Conference on Computer
Vision and Pattern Recognition, pp.
3128-3137, 2015.
[2] Samuel, Arthur L. “Some studies in
machine learning using the game of
checkers. II—recent progress.” IBM
Journal of research and development
11.6 (1967): 601-617.
[3] Krizhevsky, I. Sutskever, and G. E.
Hinton, “Imagenet classification with
deep convolutional neural networks,”
in Advances in neural information
processing systems, pp. 1097-1105,
2012.
[4] H. Lee, P. Pham, Y. Largman, and A. Y.
Ng, “Unsupervised feature learning for
audio classification using convolutional
deep belief networks,” in Advances in
neural information processing systems,
pp. 1096-1104, 2009.
[5] Y. Zheng, Q. Liu, E. Chen, Y. Ge, and J. L.
Zhao, “Time series classification using
multi-channels deep convolutional
neural networks,” in International
Conference on Web-Age Information
Management, pp. 298-310, Springer,
2014.
[6] Yu, Dong, and Li Deng. “Deep learning
and its applications to signal and
information processing [exploratory
dsp].” IEEE Signal Processing Magazine
28.1 (2011): 145-154.
[7] Sharma A, Shukla A, Tiwari R, Mishra
A. Mortality Prediction of ICU patients
using Machine Leaning: A survey.
InProceedings of the International
Conference on Compute and Data
Analysis 2017 May 19 (pp. 49-53). ACM.
10.1145/3093241.3093267
[8] Poria, Soujanya, Erik Cambria,
and Alexander Gelbukh. “Aspect
extraction for opinion mining with a
deep convolutional neural network.”
Knowledge-Based Systems, 108 (2016):
42-49.
[9] Polson, Nicholas G., and Vadim O.
 TEChNICAl TRENDS

Sokolov. “Deep learning for short-term [10] Affonso, Carlos, et al. “Deep learning [11] Evermann, Joerg, Jana-Rebecca Rehse,
traffic flow prediction.” Transportation for biological image classification.” and Peter Fettke. “Predicting process
Research Part C: Emerging Expert Systems with Applications 85 behaviour using deep learning.”
Technologies 79 (2017): 1-17. (2017): 114-122. Decision Support Systems (2017).

About the Authors

Apoorva Mishra (Life Membership No. I1504911) is a Ph.D. scholar at ABV-Indian Institute of Information
Technology and Management (ABV-IIITM), Gwalior, India. He has filed one Indian patent and has published 20
research papers in various peer-reviewed international/national journals and conferences, including those of
the world-renowned publishers like IEEE, Springer, Elsevier, ACM etc-. He has received B.E (honors) degree in
Computer Science & Engineering from Chhattisgarh Swami Vivekanand Technical University (CSVTU), Bhilai
in 2011, and has worked as ‘Assistant System Engineer’ at Tata Consultancy Services (TCS) till May 2012. He
has received the National level Scholarship from the MHRD, Government of India, for pursuing M.Tech. He has
also received Gold Medal from CSVTU, Bhilai, for scoring highest marks in M.Tech. He is a member of various
renowned profesisonal societies like ACM, IAENG, and ISTE (life member).

Prof. Anupam Shukla (Life Membership No. 01099955) is a Professor in the Department of Information and
Communication Technology (ICT) at ABV-Indian Institute of Information Technology and Management (ABV-
IIITM), Gwalior. He has 29 years of administrative, research, and teaching experience. He has been awarded
“Distinguished Professor” by Computer Society of India, Mumbai, 2017,“Dewang Mehta National Education
Award” for best Professor, 2016, ‘Prof. Rais Ahmed Memorial Award, Acoustical Society of India’ in 2015 and
has received many other awards and honors. He is the author of three patents, three books, published by
international publishers such as CRC Press, USA & Springer-Verlag, Berlin, 176 peer-reviewed publications,
editor of three books published by IGI Global Press, USA, and mentor of 18 doctorates and 116 postgraduate
theses. He is also an editorial board member and a reviewer of various international journals of repute.

CSI Members win

International recognition in SEARCC
At the recently concluded SEARCC Annual Conference, the following CSI
Members received recognition for their work in three categories for whom
they were nominated by CSI.
The prize distribution was held on September 13, 2017 at Colombo and all
three travelled to Colombo (with financial support of SEARCC) to be part of
the prize distribution ceremony.
• SEARCC Educator of the year: Dr Siddharth Bhattacharya (Winner)
• SEARCC Professional of the Year: K Sabarish (Finalist)
• SEARCC Researcher of the Year: Dr Anupam Shukla (Finalist)

PSG College of Technology, MSc Engg. from College of

Dr. A. Shamugasundaram
Obituary Founder Chairman, CSI
Coimbatore chapter
Dr. A. Shamugasundaram was the founder
Chairman of CSI Coimbatore chapter and
played very active role in the chapter. He was
instrumental in making BIG 86 a huge success.
He encouraged students to participate in
student activities.
Dr. A. Shanmugasundaram was born on 15th Feb. 1933 at
Thanjavur. Did his Bachelors in Electrical Engineering at
12
C S I C o mm u n I C a t I o n S | n o V E m B ER 2 0 1 7
Engineering Guindy and stood first in both the programmes.
He was the first UK Commonwealth scholarship awardee
from India and did his Ph.D at Royal College of Science and
Technology, University of Glasgow UK.
He was awarded the MCKENZIE memorial prize and STOKES
memorial prize by the University of Madras. In 1958, he joined
as a Lecturer in the Electrical Engineering department of
PSG College of Technology and elevated as Principal during
June 1986.
He has guided several Ph.D Scholars engaged in industrial
R&D resulting in development of many products.
www.csi-india.org
 TEChNICAl TRENDS
Machine Learning and its application in
Technology and Media industries
Krishna Murthy
Deloitte Consulting
Machine Learning - essentially
a form of artificial intelligence (AI)
whereby computers learn without
being explicitly programmed with new
information – is clear. At exceptional
speed, for example, complex algorithms
can identify subtle but important data
patterns that humans could never
have spotted. In ‘learning’ from that
information, the ‘machine’ can predict
patterns ahead, and then act to process
that knowledge to maximise future
business.
Machine learning is a modern
and highly sophisticated technological
application of a long – established
notion – study the past to predict the
future.
Machine learning and Technology
Industry:
Machine learning has become a
huge trend in the technology industry in
the last few years, as a more accessible
side of artificial intelligence, with
computers learning to complete tasks
without being directly programmed to
do so.
In 2016 there were many machine
learning pilots and trials in the industry,
with acquisitions of machine learning
start-ups by tech giants. Deep learning,
a branch of machine learning that
produces efficient algorithms to model
high level data, is also becoming
mainstream.
2017 will be the first year we’ll
start seeing real-world impacts from
machine learning technologies, with
companies moving out of the acquisition
stage, and into the implementation
phase.
Self-driving cars
Perhaps the most popular
application of machine learning in the
consumer field of vision is in driverless
cars technology. Many of these cars are
currently in the testing phase, but the
idea of self-driving cars on public roads
C S I C o mm u n I C a t I o n S | n o V E m B ER 2 0 1 7
is in its infancy.
As self-driving cars take to the
road, it will prove essential that they are
able to respond to the situations around
them in real time. This means that all
the information gained through the
sensors must be processed in the car,
rather than being submitted to a server
or the cloud for analysis, which could
cost crucial time.
As a result, machine learning
will be central to the car’s digital
infrastructure, enabling it to learn
from the conditions it observes. One
particularly interesting use of this data
will be in mapping – while self-driving
cars can have maps programmed on
to them, they will need to be able to
update these maps automatically in
response to the real-world conditions,
and the vehicle will have to learn a new
navigational network on its own.
Manufacturing
Like self-driving cars, the
manufacturing sector collects a huge
amount of data from sensors attached
to every aspect of the production line
during the growth of the IoT.
However, that information simply
isn’t being fully utilised. As multiple
parameters data is collected from
complex systems, analytics can be a
daunting task. The biggest application
of machine learning in manufacturing
will be in anomaly detection.
While earlier some analytics was
done in this space, 2017 will see more
of supervised learning and building
models to predict the failures.
Machine learning will also be used
to drive collaborative robots proof of
concepts in factories that can learn
by observing the production line and
data streams, and be able to smartly
optimise the production process to
lower production costs and speed
production cycles without the time and
financial costs of a human having to
9
analyse the data.
Finance
The financial industry is renowned
for the vast amounts of data it holds
– from transaction data to customer
data, and everything in between. This
volume is unlikely to decrease in
the future, and the finance sector is
increasingly looking to make the most
of the data that it holds. To date this has
been largely analysed using statistical
analysis tools, however, the challenge is
sorting through such a wealth of data in
a timely fashion.
In 2016, there was a lot of talk in
the industry about the potential for
deploying machine learning and several
pilot schemes, and we predict that
2017 will be the first year we see this
beginning to take off as mainstream in
finance.
Financial institutions will
increasingly lean on machine learning
to devise new business opportunities,
deliver customer services and even
detect banking fraud as it is taking
place.
Portfolio management companies
today use traditional methods like
analysing margin profiles, free cash
flow, return ratios, growth profile,
pricing power that are available through
reports.
Machine learning in the meantime
can help to scout social media,
websites and other sources, analysing
unstructured data to get some additional
inputs and make better decisions.
Retail
The science of online
recommendations has become
increasingly complex, however, in 2017,
this will become even more nuanced
as more data streams such as social
media are incorporated to provide
better recommendations.
If you were to browse for cat food
online, at the moment, you may be
 TEChNICAl TRENDS

able to see recommendations for cat- processing and automation, as bots process.
related products. In the coming year, increasingly take on tasks for IT and In contrast, deep learning allows
e-commerce sites will be able to use business processes. us to feed data into an artificial neural
more buying trends and customer data Machine learning and Media: network—a computer system modeled
to provide nuanced recommendations Machine learning is being applied after the human brain—and use
that will accurately reflect products you in the media in a number of ways. One algorithms to help the computer learn
might like to buy. area that has seen huge growth is autonomously.
While online retail is already around the personalisation of content – For example, in 2012, a team at
experiencing the early stages of the analysis of data to make conclusions Google fed images from 10 million
machine learning deployment, one and predictions. YouTube videos into an artificial neural
of the most exciting things is the Imagine news articles that network. The system identified distinct
application of this technology in the write themselves. Envision bots that objects within the images, such as cats
physical store environment. identify fake news before it’s widely (apparently there are tons of cat videos
Retailers will be able to analyse disseminated. on YouTube).
customers as they walk in, and we These things will be possible Why was this such a big deal?
will start to see this analysis used because of deep learning, in which Because no one told the computer what
to help the customer find the right computers learn human-behavior a cat actually looks like. It learned on its
products and appropriate offers. With without being explicitly programmed. own.
the incorporation of video analytics, Prior to deep learning, computers The opportunities to leverage this
retailers will be able to analyse which had to be programmed to exhibit human technology within the media industry
products people are looking at, and even behavior. will be enormous. As publishers are
where they are looking on the product – For example, if you wanted a squeezed for ad dollars and face the
whether that’s the price, the features or computer to learn how to write a news duopoly that is Facebook and Google,
the picture on the box. article, you’d have to figure out the AI/Machine Learning will help them
By considering this data against behaviors of a journalist, and manually stay competitive in an era of dominant
customer purchases, retailers will be input those behaviors into the computer. social platforms.
able to make the best recommendations This was a painfully time-consuming To that end, here are a few
for other products that the customer
may want to consider.
Developments in technology
In order to support industries in
transformations, the big developmental
shift we will see in machine learning is
one of hardware upgrades rather than
software.
In order to support the speed of
insights that machine learning can
offer, machine learning processing is
increasingly moving from the cloud to
edge computing where time-sensitive
information can be processed as close
as possible to its origin.
Alongside this, we are seeing
increasing investment in multi-core
technologies and parallel computing
that can boost this speed further. 2017
is the year that hardware begins to
catch up with software in the wider
deployment of machine learning
technology e.g. enhancements in
GPU technologies coupled with hybrid
systems like CPU+FPGA.
Machine learning is beginning to
affect us both at work and at home,
and the possibilities for industries are
truly transformative. Enterprises will
soon find machine learning at the heart
of decision making in core business
www.csi-india.org
14
C S I C o mm u n I C a t I o n S | n o V E m B ER 2 0 1 7
 TEChNICAl TRENDS

applications of AI/Machine Learning
that media companies should consider.
Automation in the newsroom
This is all about making work more
efficient.A computer system will be able
to identify the contents of an image and
write the caption.
In fact, a recent study by AI
researchers shows that humans prefer
computer-generated captions 25% of
the time they’re displayed.
The technology isn’t perfect, but
as its accuracy continues to improve,
survey results will increasingly favor
computer-generated outputs.
Similarly, AI can’t write nuanced,
eloquent news articles yet, but it’s getting
close. For example, the Associated Press
(AP) is already leveraging deep learning
to automate news articles that have
structured data, such as corporate
earnings reports and baseball game
recaps.
Ultimately, automation will enable
media companies to shift resources to
higher value initiatives, and that will be
good for both consumers and editorial
teams.
Computer-Generated Earnings
Report, Associated Press
Publishers that detect emotion
We’ve known for some time that
technology can be cognitively intelligent.
However, Affectiva—an AI company
born out of MIT’s Media Lab—is helping
it become emotionally intelligent, too.
Powered by emotion recognition
software, a streaming service like
Netflix could analyze the look on your
face to understand your feelings and
serve up content to fit your mood.
Sad? Here’s a ‘feel good’ movie.
Energized? Check out this action flick.
Silly? Maybe you’ll enjoy this comedy.
Recommendations across content
types
The opportunity to enhance
recommendations extends beyond
emotionally intelligent software. If
you’ve been around the industry, you’ve
heard time spent being referred to as the
new “currency” of advertising.
Ad buyers want to know how much
time people spend on your content,
what percentage of video ads they
complete—that sort of thing.
One way to maximize time spent
would be to suggest content across
multiple formats. For example, Fortune
Magazine could recommend its Most
Powerful Women podcast series to users
who read articles about women CEOs
on Fortune.com.
AI could discover similarities
between text, video, and audio assets,
and guide users through content that
meets their interests across different
formats.
It’s something to consider as the
industry shifts to “video first,” and
experiences continued buzz in audio
and podcasting.
Smarter content management
AI will improve our ability to
recommend content, but what about
our ability to distribute it?
With numerous platforms to find
audience at scale—including Facebook,
YouTube, Twitter, and Snapchat—it can
be difficult to determine the optimal
distribution strategy.
This reality, coupled with the fact
that many publishers rely on these
platforms for over 40% of their traffic,
creates a problem in legitimate need of
a solution.
Content management systems like
WordPress VIP and Drupal are fantastic,
but what if they could publish content
across multiple platforms, and predict
when and where (i.e. what platform) to
publish in order to maximize referral
traffic?
What if they could predict virality?
AI will make these things possible.
Cultivating a smarter workforce
AI can also be leveraged to build a
more knowledgeable workforce.
Communications teams are
commonly bogged down by the manual
process of discovering and sharing
news with their employees. The ability
to share the right information at
the right time, thus enabling better
business decisions, can be a difficult
undertaking.
It doesn’t have to be this way,
though. AI will be able to automate
this process and predict what type of
news and information will be most
valuable to your employees at any given
time—whether you’re a team of 50 or
50,000.
We believe that information is only
as valuable as the ability to access,
understand, and share it at scale, and
we’re excited about using AI to make
that easier for media companies, and
ultimately, the world at-large.

About the Author

Mr. Krishna Murthy Pradhan is a high caliber IT Management Stalwart with 19+ years of experience including 4
years of international exposure in USA, UK, Argentina and Mexico. Demonstrated success in the areas of Solution
Delivery, Program Management and Project Management, giving importance to client relationship, deliverable,
timelines, budgetary & quality parameters for smooth execution of complex enterprise level assignments
and client engagements. Currently working as Solution Sustainment Manager at Deloitte Consulting India,
Hyderabad Office and managing Custom Development/Delivery/Sustainment of Big Data & Analytics based
assignments from Offshore.

Like CSI on facebook at :

https://www.facebook.com/CSIHQ

15
C S I C o mm u n I C a t I o n S | noVEmBER 2017
 R E S E A R C h FRONT
Challenges of Theoretical Computer Science
and Computing Systems
R. Ponnusamy
Professor, Dept. of CS & Engg., CVR College of Engineering, Hyderabad - 501510, India. E-mail: r_ponnusamy@hotmail.com
The existing challenges of theoretical computer science and computing systems were documented in
the order of importance. The issues were critically discussed and gaps in the research were portrayed.
The need for further research is emphasized. These problems are classified under polynomial, cognitive
science, bio-informatics, bio-medicine, application development and other new technologies.
Keywords: Theoretical computer science, computing systems, polynomial, cognitive science, bio-informatics, bio-medicine.
1. Introduction:
Modern sciences play vital roles
in changing the social life style. The
concepts of computers have taken active
momentum after the Second World War.
It has seen many revolutionary changes,
products and systems that impacted the
business and life for several decades
and even today continuing. But, several
future visions of computing systems are
not yet completely fulfilled. Even today
there are several issues persist [1-23]
in the computing world. On finding the
appropriate solutions to these problems
the society will enjoy automated,
trusted, enhanced, quick services
through computing world.
Computer science is the study
of the theory, experimentation, and
engineering that form the basis for
the design and use of computers.
Alternatively, it is a science that
investigates automatic algorithmic
procedure to do the required
computation or task. A computer
researcher working in the field isolates
the problems into theoretical and
practical frameworks. All over the
world several researchers documented
their challenges and issues. Even
though the discipline grown has well,
these issues are still unsolved. In order
to increase the understanding of the
prevailing issues and challenges this
review is undertaken. This may help the
researchers to formulate the suitable
research projects for the current issues.
This article enlists some of the core
challenges that exist in the field today.
C S I C o mm u n I C a t I o n S | n o V E m B ER 2 0 1 7
2. Challenges and Issues:
There are several major closed-
form challenges in Computer Science,
the theoretical computer science issues
are such as Computational complexity
theory, Polynomial (P) versus Non-
deterministic Polynomial (NP),
Cryptography, Integer factorization,
Clustered Planar, Multiplication
Algorithm, Fourier Transform, Tree
Rotation and Integer Multiplication and
Matrix Multiplication Algorithm.
The computing system issues are
Big-data Applications and Analysis,
Artificial Intelligence and Robotics,
Systems Biology Computational
Biophysics, Computational
Cognitive Science, Computational
Neurobiology, Computer Assisted
Automated Education, Automated
Information Processing Centers,
Natural Language Processing,
Human-Computer Interaction,
Multimedia, Computer Artificial
Limbs and Components, Automated
Brain Intension Identification, Brain
Damage Recovery System, Semantic
Web, Advanced Technologies for
Hardware, High-end Communication
Devices, Advanced Networking
System, Space Communication and
Networking, Automated Programming
Environments, Internet Security
System, Online Electronic Voting,
Automated Software Verification and
Theorem Proving.
If the computing world finds the
suitable solutions for these problems
then the society will enjoy trusted better
16
computing services.
3. Computational complexity
theory and Applications
In computational complexity
theory is a critical one which identifies
whether the given algorithm is capable
of solving the problem with the existing
resources and also determines it
is capable of accepting the given
input or not. These resources are
time, memory, communication and
processing etc. Firmly related fields in
computer engineering are investigation
of algorithms. All the more correctly,
computational complexity theory tries
to order issues that can or can’t be
comprehended with properly limited
assets. There are different issues are
there in the field of theoretical computer
science. These issues are detailed in
the following section one by one.
Secondly there are limitations in
developing novel solutions in various
application development. In the
absence of these solutions the current
system restricts the user to access
the services. In fact, much innovative
research on hardware would be able
to take us to the new era of computing
world. Also, the solutions will envisage
the low cost computing in the world.
The developing countries are not able to
afford the solutions and reachability will
be much more compared to the present
era of computing.
3.1 P versus NP problem
In general the computing problems
can be classified in to two categories,
www.csi-india.org
 R E S EA R Ch FRONT
that is either it may be solved by
algorithms in a Polynomial (P) time
or Non-deterministic Polynomial (NP)
time [13, 7]. It is only a question of
whether it is solvable in a deterministic
or non-deterministic polynomial time
and resources. As of now the non-
deterministic polynomial (NP) problems
are classified as either complete or hard
problems. In case of complete problem
the solutions are possible and in the
case of hard problems the solutions are
possible only by converting the problem
in to a complete problem by means
of approximation or constrains. The
computing world should find suitable
solutions for this.
3.2 Cryptography
A public key encryption method
is used as a method of assuring
the confidentiality, authenticity and
non-reproducibility of electronic
communications [7]. It needs a private
key to decrypt the messages. Anybody
knows the private key can decrypt
messages sent over the public network
and this is a big challenge in the security
world and it needs alternate solutions
to overcome this problem.
3.3 Integer factorization
In number theory, whole number
factorization is the disintegration of
a composite number into a result of
littler whole numbers. In the event that
these whole numbers are additionally
limited to numbers, the procedure is
called prime factorization [15]. When
the numbers are large, no efficient,
non-quantum integer factorization
algorithm is known. So far it has not
been demonstrated that no proficient
algorithms exists; finding an efficient
algorithm for this problem is a big
challenge, till this time.
3.4 Clustered Planner
The clustering can be depicted
combinatorially by a gathering of
subsets of the vertices with the end goal
that for every two subsets, either both
are disjoint or one is contained in the
other. It is not required that the grouping
is maximal [14] nor that each vertex has
a place with a bunch and it is obscure
whether it is conceivable to develop
grouped planar drawings in polynomial
time. There are many cases which have
polynomial time algorithms. Finding a
C S I C o mm u n I C a t I o n S | n o V E m B ER 2 0 1 7
best polynomial time algorithm is a big
challenge till date.
3.5 Fourier transform
A Fast Fourier Transform
calculation registers the Discrete
Fourier change of a grouping, or its
reverse. Fourier Transform changes
over a signal domain from its area to
a frequency domain in the recurrence
space [17]. A FFT quickly registers such
changes by factorizing the DFT network
into a result of inadequate elements.
Subsequently, it figures out how to
diminish the many-sided quality of
registering the DFT, which in the event
that one essentially applies the meaning
of DFT. Fast Fourier transforms are used
for many applications in engineering
and science. Finding less complex and
efficient method is a big challenge till
today.
3.6 Tree rotation
In discrete science, tree turn
is an operation on a parallel tree.
Progressions are made in the structure
without meddling with the request of
the components [18]. A tree turn moves
one hub up in the tree and one hub down
and it is utilized to change the state of
the tree, and specifically to reduce its
tallness by moving smaller sub-trees
down and bigger sub-trees up, bringing
about enhanced execution of many tree
operations. There exists an irregularity
in various depictions with regards to the
meaning of the heading of turns; some
say that the course of revolution mirrors
the bearing that a hub is moving upon
pivot while others say that the bearing
of revolution reflects which sub-tree
is turning. It is another big issue for
theoretical computer scientist.
3.7 Multiplication algorithm
A multiplication algorithm is an
algorithm to multiply two numbers [16].
Depending on the size of the numbers,
different algorithms are in use; efficient
multiplication algorithms have existed
since the advent of the decimal system.
But, still much better algorithms are
needed to reduce the number of basic
operations.
3.8 Matrix multiplication algorithm
Many different algorithms do exist
for matrix multiplication [20]. They are
performed in different hardware setups
like parallel, distributed and standalone
23
setups. But it is still unknown which
does in optimal time.
3.9 Big-data applications and
Analysis
The system should able to
quickly analyze and process the semi
structured and unstructured data using
machine learning[7] and patterning is
yet another challenge in the present era
of computing.
3.10 Artificial intelligence and robotics
It is to make a computer to become
intelligent one to perform efficient
tasks, to make effective decisions like
humans. Self-driving cars and swarms
of search-and-rescue robots are a
good illustration [7]. The system needs
some intelligence (common sense) to
do the works and things automatically.
Designing such system is a challenging
one at present.
3.11 Systems Biology
In understanding modeling
interactions of multiple systems in a
living organism [7], including immune
systems and cancer development, to
develop the suitable treatment therapy
is another challenge.
3.12 Computational Biophysics
The modeling and understanding
mechanical, electrical, and molecular-
level interactions inside an organism
[7] is another big challenge for the
computer scientists.
3.13 Computational Cognitive Science
There is a very large gap between
what is known about the understanding
of brain structure and the functional
capabilities of a living brain; closing
this gap is one of the big challenges in
modern science and engineering [21].
DNA analysis and genetics have also
become computer-based in the last
twenty years. Development of artificial
brain neural system in understanding
the nature is the big challenge for
computer scientist.
3.14 Computational neurobiology
It is another area of research to
understand how organisms process
incoming information and react to it,
control their bodies, store information,
and think [7]. The existing connectionist
neural models are not fully complying
with the requirements. It is another
modern research challenge for
 R E S EA R Ch FRONT
computer scientists.
3.15 Computer-assisted automated
education
It essential to design a system
to teach a lesson spontaneously [11].
Cheat-proof automated support for
exams and testing, essay grading,
generation of multiple-choice
questions. Support for learning specific
skills, such as programming by giving
immediate feedback and correction of
mistakes and giving suitable example
narration are the challenging one in the
present era of computing.
3.16 Automated Information
Processing Centers
Automated collection, storing,
indexing and retrieval are major issues
[7] in the present days. Development of
such systems will enable the designers
to handle huge set of unstructured data
generated by the humans in different
forms.
3.17 Natural-Language Processing
The system should understand
the human-generated documents and
should be able to communicate with
the humans in a normal manner [4].
This needs syntax, semantic, pragmatic
level of conversation and dialogs. This
will also solve many other computer
interaction issues. Design of solutions
for such problems is a greater.
3.18 Human-computer interaction
The current generation systems are
not able to understand tacit knowledge
available with the user. Moreover the
user interaction is restricted with the
keyboard and few selective devices.
But, the future generation system
should understand the human mind
states such as emotion, intent[5], social
interactions and spontaneously give the
suitable answer on behalf of the user
as well as do the tasks. Design of such
system is a challenging one.
3.19 Multimedia
In the design of video or audio
or multi-media data the system is
not able to understand any semantic
structure associated (scene analysis,
comprehendingmovement, synthesizing
realistic multimedia) with it [7]. Most of
the time the designer is bothered about
the quality of display data, rather than
semantic structure. At present people
are doing it with the explicit annotation.
C S I C o mm u n I C a t I o n S | n o V E m B ER 2 0 1 7
Any multimedia-data analysis without
considering semantics will be a futile
exercise. It is another challenge in the
future generation multimedia.
3.20 Computerized Artificial Limbs and
Components
Of late even extra ambitious
envisioning programs are underway
for systems to control Artificial Limbs.
Engineers envision automatic implants
capable of receiving the signals from
hundreds of the brain’s nerve cells
and then wirelessly transmitting that
statistics to an interface tool that might
decode the mind’s intentions [1]. The
interface could then ship indicators to
an Artificial Limb, or maybe at once to
nerves and muscular tissues, giving
guidelines for imposing the preferred
actions. This could help the physically
handicapped people to have better and
easy movements.
3.21 Automated Brain intension
Identification
Some brain studies have explored,
with some achievement, implants that
could literally study the thoughts of
immobilized sufferers and signal an
external computer [1], giving human
beings who are not able to talk a way
to communicate with the outside world.
This can have a substantial opportunity
to have exclusive applications.
3.22 Brain Damage Recovery System
Brain is the probably most
complicated system which has even
greater impact on human health.
The use of new AI insights can be
used for repairing broken brains [1].
Damage from injury or disease to the
hippocampus, a brain shape essential
for gaining knowledge of and memory
can disrupt the proper electric signaling
among nerve cells; this is needed for
forming and recalling recollections with
understanding of the right signaling
styles in wholesome brains, engineers
have begun to layout computer chips
that mimic the brain’s very own verbal
exchange abilities. Such chips may be
useful in cases where healthy brain
tissues is starved for information due
to the barrier imposed via damaged
tissue in principle; indicators from
the healthy tissue can be recorded
by an implantable chip, which might
then generate new signals to pass the
18
damage. Such an electronic charge
signaling route should help restore
ordinary reminiscence capabilities to an
impaired brain.
3.23 Semantic Web
The semantic net [8] is an extension
of the contemporary net in which facts
are given properly-described, which
means, enabling computer systems and
those to work in cooperation. Gap exists
between automated processing and
accessing of information in the present
web setup. The present challenges are
(i) the availability of content, (ii) ontology
availability, development and evolution,
(iii) scalability, (iv) Multi-lingualism,
(v) visualization to reduce information
overload, and (vi) Stability of Semantic
Web languages. If these things are
solved then the users will see a different
kind of Internet tomorrow.
3.24 Advanced Technologies for
Hardware
Development of optical and
quantum computing or optical
computing [2, 3] hardware technology
will enable us to develop the three-
dimensional circuits and chips. Even
though the physicists are involved in
that, the computer scientist has also
little things to do with this development.
It will have a high-impact in increasing
the processing power.
3.25 High-end Communication Devices
Modeling and the usage of new
styles of digital switches (Memristors,
devices using carbon nano-tubes,
etc), quantum communication [22,
7] and improvement of high-stop
sensing devices to understand diverse
communications occurring around the
device is the predominant challenge for
electronic scientist.
3.26 Advanced Networking
Ubiquitous networking by giving the
low cost service [7] to access multiple
services is another important issue in
the communication world.
3.27 Space Communication and
Networking
Days are not for behind for human
to travel to the other planets. It is also
essential to develop a new network
system for space communication [7].
Development of such a network is
another challenge.
www.csi-india.org
 R E S EA R Ch FRONT
3.28 Automated programming
Environments
Complete Automated program
generation [12] is only a partial success
in the age of programming. The
next generation programming must
support the automated programming,
compilation, and debugging and system
development. Also, automated program
language conversion (converting the
code from one language to another) is
another problem. It is the big deal for
computer scientist.
3.29 Internet Security System
The security system development
is another challenge in the Internet
[23]. Even though security is provided at
different levels the creation of trusted
secured environment in the public
network is still a major issue.
3.30 Online Electronic Voting
All over the world big democracies
are moving towards electronic voting
system. Development of trusted state-
of-art online electronic voting system
[10] with foolproof security is a big
challenge.
3.31 Automated Software Verification
and theorem proving
Software verification in the dynamic
environment [19] is another challenge
for the software developers. The logical
entailment and formal reasoning is the
another challenge in the computing
system. This system development
should enable us to solve the theorem
proving and diagnostics system.
3.32 Automatic Personalized Medicine
System
One goal of biomedical engineering
these days is fulfilling the promise of
personalized medicine [1]. Medical
doctors have long identified that people
vary in their susceptibility to ailment
and their reaction to treatments, but
clinical technologies have generally
been presented as “one length fits
all.” Current cataloging of the human
genetic endowment, and deeper know-
how of the body’s complement of
proteins and their biochemical inter-
actions, provide the prospect of figuring
out the specific factors that determine
illness and well being in any man or
woman. A critical manner of exploiting
such data would be the improvement of
techniques that permit docs to forecast
C S I C o mm u n I C a t I o n S | n o V E m B ER 2 0 1 7
the benefits and side consequences of
capability treatments or treatments.
“reverse-engineering” the mind, to
decide how it plays its magic, should
provide the twin benefits of supporting
treat sicknesses at the same time as
imparting clues for brand new methods
to computerized Artificial Intelligence.
Advanced computer intelligence, in flip,
should allow computerized analysis and
prescriptions for remedy. Automatic
catalogs of fitness data should beautify
the clinical device’s ability to tune the
unfolding of sickness and analyze the
comparative effectiveness of various
procedures to prevention and therapy.
The engineering demanding situations
for enabling drug discovery replicate
the ones needed to permit customized
medicine: improvement of greater
effective tools and techniques for
speedy evaluation and prognosis in
order that a variety of medicine can
be fast screened and right remedies
may be promptly carried out. Modern
capsules are frequently prescribed
incorrectly or unnecessarily, promoting
the development of resistance without
actual clinical benefits. Quicker, Extra
precise diagnoses may additionally
lead to greater targeted and effective
therapies. Antibiotics that attack a huge
range of bacteria have usually been
sought, becausedoctorscouldn’tusually
make certain of the ideal bacterium
inflicting an infection. Units which can
determine the real wrongdoer right
away may want to cause using greater
narrowly centered tablets, reducing the
risk of selling resistance. Developing
organism-specific antibiotics should
come to be one of the century’s most
essential engineering challenges.
4. Conclusion
In general the engineers play vital
roles in creating the new products and
systems in diverse fileds. The computer
scientist is not exception for that. All
over the world the computer people
create novel software and systems every
day. Hence, this article aims to review
the major problems and an issue for the
computer engineering scientist about
the different difficulties exists today. It
surveys a few hot issues and difficulties
that need to understand in the field. A
large portion of the issues anticipated
that would locate the fundamental novel
23
solutions. The appropriate researchers
should take-up the research on these
issues for future computing system.
References:
[1] Grand Challenges For Engineering,
National Academy of Engineering of
the National Academics, http://www.
engineeringchallenges.org
[2] Richard P. Feynman, Simulating
Physics with Computers, International
Journal of Theoretical Physics, VoL 21,
Nos. 6/7, 1982.
[3] Nolte, D.D., Mind at Light Speed: A New
Kind of Intelligence, ISBN 978-0-7432-
0501-6, 2001.
[4] Bates, M, Models of natural language
understanding, Proceedings of the
National Academy of Sciences of the
United States of America. 92 (22): 9977–
9982, 1995.
[5] https://www.interaction-design.org/
literature/book/the-encyclopedia-of-
human-computer-interaction-2nd-ed/
[6] Computing Research Association http://
cra.org/]
[7] 13 Of 2015’s Hottest Topics In Computer
Science Research - Forbes,2015 https://
www.forbes.com
[8] Jorge Cardoso, “The Semantic Web
Vision: Where Are We?”, IEEE Intelligent
Systems, vol. 22, no. , pp. 84-88,
September/October 2007, doi:10.1109/
MIS.2007.97
[9] http://www.indianspacestation.com/
research/general/575-challenges-in-
deep-space-communication
[10] https://www.scientificamerican.com/
article/pogue-the-challenges-of-
digital-voting/
[11] Wang Tiantian, et. al., AutoLEP: An
Automated Learning and Examination
System for Programming and its
Application in Programming Course,
First International Workshop on
Education Technology and Computer
Science, 2009. ETCS ‘09,7-8 March
2009, ISBN: 978-0-7695-3557-9,DOI:
10.1109/ETCS.2009.18.
[12] Alan W. Biermann, Automatic
programming : A tutorial on formal
methodologies, Journal of Symbolic
Computation, Volume 1, Issue 2, June
1985, Pages 119-142.
[13] Gerhard J. Woeginger, Open problems
around exact algorithms, Discrete
Applied Mathematics 156 (2008) 397 –
405.
[14] Cortese, Pier Francesco; Di Battista,
Giuseppe, “Clustered planarity”, Proc.
21st Symp. Computational Geometry
(SCG’05), New York: ACM, pp. 32–34,
doi:10.1145/1064092.1064093, 2005.
[15] Richard Crandall and Carl Pomerance.
Prime Numbers: A Computational
Perspective. Springer. ISBN 0-387-
 R E S EA R Ch FRONT

94777-9,2001. geometry, Journal of the American SIAM News. 38 (9), 2005.

[16] Fürer, M. Faster Integer Multiplication
in Proceedings of the thirty-ninth
annual ACM symposium on Theory
of computing, June 11–13, 2007, San
Diego, California, USA.
[17] Christos H. Papadimitriou, Optimality
of the Fast Fourier transform, Journal
of the ACM (JACM), Volume 26 Issue 1,
Jan. 1979, Pages 95-102.
[18] Sleator, Daniel D.; Tarjan, Robert
E.; Thurston, William P. ,Rotation
distance, triangulations, and hyperbolic
Mathematical Society, 1 (3): 647–681,
1988, doi:10.2307/1990951.
[19] M. Popovic et al.,A formal software
verification concept based on
automated theorem proving and reverse
engineering, Proceedings. Ninth
Annual IEEE International Conference
and Workshop on the Engineering of
Computer-Based Systems, 2002,8-11
April 2002, ISBN: 0-7695-1549-5.
[20] Robinson, Sara, Toward an Optimal
Algorithm for Matrix Multiplication,
[21] Harnish, R. M. (2002). Minds, brains,
computers: an historical introduction
to the foundations of cognitive science.
Malden, MA: Blackwell Publishers.
[22] Nasibulin, Albert G.; et al., A novel
hybrid carbon material, Nature
Nanotechnology,2007 2 (3): 156–161.
[23] h t t p s : / / w w w . c e d m a g a z i n e . c o m /
article/2000/11/internet-security-
challenges
n

About the Author

Dr. R. Ponnusamy received the M.Tech. in Computer Science & Engineering from Pondicherry University, Pondicherry, India
in 2000. He has a Ph.D. in Computer Science & Engineering from the Dept. of Computer Science & Engineering, College
of Engineering, Anna University, Chennai in 2008. He has also received a D.C.Sc. (Engg) (Post-Doctoral Degree), from the
Universita Popolare Degli Studi De Milano, Italy in 2015. He is Professor of Computer Science and Engineering and has more
than 20 Years of collegiate service in Tamilnadu, India. He is currently Professor in the Department of Computer Science and
engineering at CVRCE, Hyderabad. His areas of interest are Artificial Intelligence, Human-Computer Interaction, Information
System and Higher Education. He has presented/published more than 85 papers in various international conference and
journals. He has been awarded Young Scientist Fellowship Award by the Tamilnadu State Council for Science & Technology and
visited Super Computer Education Research Centre of Indian Institute of Science, Bangalore, India during the year 2005-2006.

CSI Executive
Committee (ExecCom)
meeting at Nashik on
7th October, 2017

Kind Attention: Prospective Contributors of CSI Communications

Please note that Cover Theme for December 2017 issue is Cyber Physical Systems. Articles may be submitted in the
categories such as: Cover Story, Research Front, Technical Trends, Security Corner and Article. Please send your contributions
by 20th November, 2017.
The articles should be authored in as original text. Plagiarism is strictly prohibited.
Please note that CSI Communications is a magazine for members at large and not a research journal for publishing full-
fledged research papers. Therefore, we expect articles written at the level of general audience of varied member categories.
Equations and mathematical expressions within articles are not recommended and, if absolutely necessary, should be
minimum. Include a brief biography of four to six lines, indicating CSI Membership no., for each author with high resolution
author photograph.
Please send your article in MS-Word format to to Editor, Prof. Prashant R. Nair in the email ids csic@csi-india.org with cc to
prashant@amrita.edu
(Issued on the behalf of Editorial Board CSI Communications)

Dr. S S Agrawal
Chief Editor

www.csi-india.org
20
C S I C o mm u n I C a t I o n S | n o V E m B ER 2 0 1 7
 R E S EA R Ch FRONT
Fuzzy Based Machine Learning:
A Promising Approach
Sujamol S., Sreeja Ashok and U Krishna Kumar
Department of Computer Science & I.T, School of Arts & Sciences, Amrita University, Kochi - 682 024
why Fuzzy Based learning is
Important
The real life application gives less
importance to boolean variables. They
mainly concentrate on linguistic variable
that have values as words or sentences.
For example, the area medical field,
come across many levels of uncertainty
as there are different manifestation of a
single disease depending upon a patient
and occurrence of a single symptoms
in multiple disorders. All the entities
related with a disease are described
using linguistic words and will have
values in the range 0-1. The traditional
mathematics lies on precisely defined
points, functions, sets and probability
measures. Here an element x either
belongs to a particular set X or it
doesn’t. They are purely dependent on
binary logic. A precise statement loses
its meaning and a meaningful statement
loses its precision when complexity is
increased. Even though human brain
resembles a binary computer based
on exact information containing zeros
and ones, human heart is compared
with a chemical computer that uses
fuzzy logic for analyzing information
that can’t be easily defines by zeros
and ones. The task of decision making
has met with many difficulties when
dealing with real world problems. The
traditional crisp logic is based on binary
reasoning where everything is classified
as Yes (True) or No (False). Fuzzy is
described as mathematics of words
and was developed by Lofti.A.Zadeh in
1965. The fuzzy logic provides a work
space for computation with words and
offers a hand in managing uncertainty
during the designing of expert systems.
It has now become an unavoidable part
of machine learning as it can handle
imprecise and uncertain situation.
Fuzzy logic can take value between
0 and 1 and these values represent
degrees of uncertainty and models
human cognitive decision making. 0’s
C S I C o mm u n I C a t I o n S | n o V E m B ER 2 0 1 7
Table 1: Comparison between Traditional logic and Fuzzy logic
Traditional Logic Fuzzy Logic
(x) = 1 if x  A μA:x→[0,1]
0 if x  A
Allows only two binary values : 0 Permitts a range of value between 0 and 1
and 1 where 0 represent no membership and 1
Where 0 stands for True and 1 for represent complete membership.
False
Based on models from differential Uses Knowledge generated through rules
equations
Requires complete set of data Concentrates on incomplete and fuzzy data
Linear Non Linear
Table 2: Brief history of Fuzzy logic - Important Milestones
Year Invention
1965 Fuzzy Sets(Lofti Zadeh,Seminar)
1966 Fuzzy Logic(P.Marinos,Bell Labs)
1972 Fuzzy Measure(M.Sugeno,TIT)
1974 Fuzzy Logic Control(E.H.Mamdani)
1980 Control of Cementkiln(F.L.Smidt,Denmatk)
1987 Sendai Subway Train Experiment(Hitachi)
1988 Stock Trading Expert System(Yamaichi)
1989 LIFE(Lab for International Fuzzy Engineering)
2003 First class on fuzzy logic is held at Clarkson University
2005 Studies in gene expression using Fuzzy Logic
2006- Successful implementation of fuzzy models in control theory, disease
2016 prediction,Faurd detection etc
stands for false values and 1’s for true logic.
values. Fuzzy set is a form of knowledge Basic Elements of a Fuzzy System
representation suitable for notions that Fuzzy system (Fig: 1) consist of
cannot be defined precisely, but which following 5 elements
depend upon their contexts. They are an
extension of crisp sets .A crisp set is a Rule Base
traditional way where each element is
allowed with full membership or partial Fuzzification Defuzzification
membership. It is denoted as (x) =1
when x belongs to set A and (x) =0, if
Knowledge Base
x doesn’t belong to A where (x) is the
membership function of x in A. Table
1 represents the difference between Crisp Input Crisp output
traditional and fuzzy logic system and
Table 2 details the brief history of fuzzy Fig. 1: Fuzzy System
23
 R E S E A R C h FRONT
 Linguistic Variables
 Membership functions for
Fuzzification
 Fuzzy Rules
 Fuzzy set operations
 Defuzzification
linguistic variables: Variables
whose values are words or sentences.
When a system is to be modeled, the
system parameters are decided and
they are set as linguistic variables.
Now these variables are assigned with
corresponding membership values.
These variables handle partial truth
between completely true and completely
false.
Fuzzification: The process that
converts a crisp input data into fuzzy
value by attaching a membership
grade to each element. This is done
with the help of suitable membership
function. The output of a membership
function is a value between 0 and
1.The following table(Table 3) shows
different membership functions and
corresponding equations where a ,b and
m are the lower limit, upper limit and
modal value respectively.
Fuzzy Rule Base: It is a simple
IF-THEN rule with a condition and
a conclusion that is constructed to
control the output variable. Fuzzy rules
are evaluated and the results of the
individual rules are combined using
fuzzy set operations.
Defuzzification: The final result,
which is a fuzzy value, is converted
to crisp value using defuzzification
method. Fig: 2 below shows working
of an air conditioner which is based on
fuzzy logic and artificial intelligence.
Fuzzy Logic System (FLS) adjusts the
temperature of the room according
to the current temperature of the
room and the target value.The room
temperature is compared with the
target temperature by the fuzzy engine
and a suitable command to heat or cool
the room is produced.
FlS
Air
Conditioner
Fig 2: Fuzzy based Air Conditioner
C S I C o mm u n I C a t I o n S | n o V E m B ER 2 0 1 7
Table 3: Membership Functions
Triangular Membership Function 0 if x <= a
A(x) = (x-a)/(m-a) if x € [a ,m]
(b-x)/(b-m) if x € [m ,b]
0 if x >= b
Singleton Membership Function 0 if x ≠ m
S(x) =
1 if x = m
Trapezoidal Membership Function T(x) = 0 if (x <= a) or (x >= d)
(x-a)/(b-a) if x € (a,b)
1 if x € (b,c)
(d-x)/(d-c) if x € (b,d)
Gaussian Membership Function G(x) = exp[-(x-m)2 /2σ2]
Fuzzy Based Machine learning element to belong to a particular cluster
Machine learning algorithms varies in the interval 0 to 1.
primarily aims at extracting knowledge Hard Clustering Soft Clustering
from data and they employ traditional
methods of clustering, classification
and associations for this purpose.
Fuzzy set are applied in conjunction
with these methods to produce more
flexible results. Fuzzy based knowledge
representation and reasoning is widely
used in designing machine learning
algorithms and models. They gave Cluster 1 Cluster 2
a new life to the scientific fact of
machine learning which have been in
a dormant stage for a long time. Fuzzy
sets are applied in the stage of data
processing itself as they have the ability
to model incomplete and imprecise
data representation. Fuzzy extension
principle was used in mapping of data
to models. Fuzzy rule based systems
wrapped with fuzzifiers and defuzzifiers Cluster 1 Cluster 2
were used for classification and
regression. Fig 3: hard clustering and soft clustering
Fuzzy Clustering Algorithms
As for an example, in a medical
The clustering algorithms are
field different diseases have similar
mainly divided as hard clustering and
symptom or one single gene may
soft clustering. In hard clustering each
be causing multiple disorders. In all
object belongs to only one cluster. But
these cases it is really intelligent to go
in soft clustering also known as fuzzy
with fuzzy clustering. Different Fuzzy
clustering, each object belongs to
Clustering Algorithms are summarized
more than one cluster(Fig :3). In fuzzy
in Table 4
clustering, a membership value of each
Fuzzy Association Rule
Association Rule Mining is
Room considered as a powerful functionality
in data mining as it finds association
between different co-occurring events
in a data set. But the traditional
association rules found to be unfit when
solving realistic problems since they
use only binary values of 0’s and 1’s.
www.csi-india.org
22
 R E S E A R C h FRONT

Table 4: Fuzzy based Clustering Algorithms- Summary

Algorithms Descriptions References

Fuzzy C Means Proposed by Dunn in 1973 and modified by Bezdek J. C. Dunn Proposed FCM in “A Fuzzy Relative of
Clustering (FCM) in 1981. Most popular fuzzy clustering techniques the ISODATA Process and Its Use in Detecting
Fuzzy c-means clustering involves two major steps Compact Well-Separated Clusters” published
1. The calculation of cluster centers in Journal of Cybernetics in 1973 and extended
2. Assignment of points to these centers using a by. Bezdek in 1981 in his paper” Pattern
form of Euclidian distance. Recognition with Fuzzy Objective Function
The algorithm assigns a membership value to the Algorithms” published in Society for Industrial
data items for the clusters within a range of 0 to 1 and Applied Mathematics
A fuzzification parameter determines the degree of
fuzziness in the clusters
Possibilistic C Proposed by Krishnapuram and Keller Krishnapuram and Keller proposed PCM in
Means Clustering Helps to identify outliers (noise points). “The Possibilistic C-means algorithm: insights
(PCM) Results in low typicality values for outliers and and recommendations” Published in: IEEE
automatically eliminates these noise points. Transactions on Fuzzy Systems ( 1996 )
Sensitive to initializations and may generate
coincident clusters.
Fuzzy Possibilistic Proposed by Pal and Bezdek Pal,Bezdek,et.al proposed FPCM in “A Fuzzy
C-Means Integrated the features of both Fuzzy C-Means and Possibilistic c-means clustering algorithm”
Algorithm (FPCM) Possibilistic C-Means published in IEEE Transactions on Fuzzy
Fuzzy values of the FCM and typicality values of Systems(1997)
the PCM are combined in order to achieve a better
clustering model
Possibilistic PFCM solves the noise sensitivity deficiency problem Nikhil R. Pal, Kuhu Pal,et.al proposed PFCM
Fuzzy C Means of FCM algorithm. in “Possibilistic Fuzzy C Means Algorithm”
Algorithm (PFCM) PFCM helps to overcome the coincident clusters published in IEEE Transactions on Fuzzy
problem of PCM. Systems (2005)
PFCM provides an improvement to FPCM by
eliminating the row sum constraints of FPCM.
They often result in loss of information it an attractive tool in classification.
Knowledge Base
and create problem of uncertainty Fuzzy logic when combined with Neural Input
.So numeric attributes were replaced neural networks can produce more Neural Output
with fuzzy attributes. They can handle reliable output as this fuzzy enhances Neural Network Fuzzy Interface
uncertainty problems associated with the generalization capability of neural
real world and maintain information networks (Fig: 5). Large number
integrity. Numeric attributes are of neuro-fuzzy models have been Learning Algorithms
converted into fuzzy attributes using developed and they were applied in many
fuzzy apriori and various other fuzzy fields such as disease prediction, risk Fig. 5 : Classification using neuro-fuzzy
association rules (Fig 4). Strong analysis etc. Combination of fuzzy and model
association rules generated using neural networks can result in human-
fuzzy logic is used in complex decision like reasoning by producing dynamic estimate input output functions.
making situations. and parallel processing systems that
Fig 4: Learning Membership function
and mining association rules Learning
Predifined MF Deterministic MF’s Fuzzy Mining
Process
Fuzzy Classification
Classification is a supervised learning Evolution
method used to make intelligent Transaction DB Module Fuzzy Association
MF
decisions. In classification a class is Rule
defined with a specific property and all
elements within that class obeys that Transaction DB
property. The mesmerizing performance
of neural networks in dealing with Fig. 4 : learning Membership function and mining association rules
large number of parameters made

23
C S I C o mm u n I C a t I o n S | n o V E m B ER 2 0 1 7
 R E S E A R C h FRONT

Conclusion by fuzzy methods help in producing combination and fusion of information

The main contribution of fuzzy
towards machine learning is its
graduality, which is the ability to
represent gradual concepts and
properties. Fuzzy concepts are used
as modeling elements in a particular
problem domain. In machine learning,
once a problem domain is found, it can
be broken as gradual concepts. These
concepts and the relation between
them can be learned to perform proper
decision making. Another important
property, that is, granularity offered
information granules for knowledge
representation in machine learning.
Fuzzy sets offers interpretability by
providing an interface between a
numeric values and symbolic values
composed of linguistic terms. Also
these fuzzy models are more robust
than non-fuzzy models as they are
deprived of defining strong classes of
boundary. Representation of uncertainty
together with the incorporation of
back ground knowledge can result
in more sophisticated aggregation,
that are necessary for machine learning
algorithms. Ultimately, when fuzzy logic
combined with machine learning can
be applied in many real world situation
including medical decision making,
education, multimedia application,
economics, risk analysis, air craft
maintenance, gene network analysis
etc. Due to these advantages, fuzzy
logic based machine learning becomes
a promising approach towards better
decision making.
n

About the Authors

Sujamol S. is a former Assistant Professor at SNGIST Group of Institutions, N Paravur, completed M-Tech in Computer
Science and Engineering from Christ Knowledge City, Mannoor and B-tech from SNMIMT, Maliankara, Ernakulam. Currently
doing research in Data Mining at Amrita School of Arts and Sciences, Amrita University, Kochi.

Dr. Sreeja Ashok is currently working as an Assistant Professor in the Department of Computer Science, Amrita School
of Arts and Sciences, Amrita University, Kochi. She has blend experience in industry and academic fields. Her industrial
experience includes Data Analytics, Project Management, Software Development, and Quality Assurance (Wipro and Avenir).
Her research is broadly in the fields of computational biology, developing machine-learning algorithms for modeling
biological and clinical data; has a strong publication record to her credit.

Dr. U Krishna Kumar, a Life Member of CSI, is Director, Amrita School of Arts and Sciences and Dean (Arts,
Media and Commerce) of Amrita Vishwa Vidyapeetham University. He has more than three decades of
experience in teaching and research. Areas of interest include Data Mining, Medical Informatics, Media and
Science Communication. Has published research papers in various national and international journals.

Past President visit to Shambhunath Inst. of Engineering & Technology

On 28 September 2017, Sri H.R. Mohan, Past President, Computer Society of India along with Dr. P. Sakthivel, Chairman,
IEEE Computer Society, Madras Chapteit visited Shambhunath Institute of Engineering & Technology, Allahabad. They were
welcomed by Prof. D.K. Dwivedi, Director (Capacity Building & Entrepreneurship), Prof. Radha Raman, HOD, CSE, Prof.
Parikshit Agrawal, Prof. Abhishek Kumar Pandey, Student Coordinator, CSI-SIET Student Branch and other faculty members
of Department of Computer Science & Engineering, SIET, Allahabad.

www.csi-india.org
24
C S I C o mm u n I C a t I o n S | n o V E m B ER 2 0 1 7
 ARTIClE
Using statistical approach to Machine
Translation for Indian languages
M. Sasikumar
Director, Centre for Development of Advanced Computing, Mumbai (C-DAC) Mumbai
Introduction
Machine Translation (MT), i.e.,
translation of content from one
language to another language using the
help of computers, is a necessity for a
country like India. Being a multilingual
country with 22 official languages
and only small share of population
speaking English, communication
of information is a major challenge.
Human translators try to bridge this
information gap, but the inherent
limitations of manually performing a
complex knowledge intensive task like
translation restricts the height one can
reach. As the information technology
is reaching masses, we will be seeing
more focus on Indian languages.
Making all these people understand
each other at run time, is a task beyond
human capabilities. MT systems apart
from automating and speeding up the
task, also offers confidentiality, i.e., you
don’t have to expose your content to an
unintended (human) user just for the
sake of translation.
MT has been an important field
of research in computer science for
many decades. Though it’s still work
in progress, quality of translation is
improving day by day with the emerging
techniques. Despite the long gestation
period, the complexity of natural
languages has still not allowed us to
solve the problem completely. The
major source of complexity in natural
language is ambiguity, in many different
forms. One word in a language may
have multiple meanings; for example,
word ‘bank’ in the sentence “I lost
my wallet at the bank”, may mean
that, you lost your wallet at a financial
bank while withdrawing money or you
lost your wallet at a river bank while
having bath. This is called word sense
ambiguity and sense of such words in
the given sentence needs to be decided
using a technique called word sense
C S I C o mm u n I C a t I o n S | n o V E m B ER 2 0 1 7
Prakash B. Pimpale
Senior Technical Officer, C-DAC Mumbai
disambiguation. Ambiguity can even
happen at sentence level as in the
sentence, “I saw the man on the hill
with a telescope”, which has multiple
meanings depending on where ‘with a
telescope’ is attached. As MT involves
two languages, problems like these get
amplified. The involved languages may
have completely different morphology
and syntax, i.e., it may happen that, a
word in one language has no equivalent
word in another language (and so has
to translated into multiple words), or
may be one language uses different
grammatical structure than the other.
These are just a few snapshots to
demonstrate the source of complexity
for MT.
In this paper, we discuss the
Statistical Machine Translation (SMT)
approach to tackle the problem of
translation and how we at C-DAC
Mumbai have used the same in our MT
system, Anuvad, for Indian languages.
Statistical Machine Translation
The techniques for MT are being
continuously improved and researchers
have moved from the traditional Rule
Based MT (RBMT) to Statistical MT
(SMT) and recently to Neural MT (NMT/
DeepMT). Major factors which fostered
this movement were the hope of a more
accurate translation and lesser human
involvement in creating the knowledge
structures required for RBMT. Now
the problem of MT has moved from
structural and morphological transfer
of one language sentence into
another, to transfer of one pattern into
another. For SMT, the languages are
considered as patterns and learning
the transfer model from one pattern
to another became the translation
system. Availability of huge sentence
aligned parallel corpora accelerated
the development of machine-learning
based MT systems, paving way for SMT.
For NMT, the corpora requirement is
25
much higher, and where such large
corpora is available, there have been
attempts to use NMT.
A translation problem can be
construed as figuring out what patterns
in the source language correspond to
what patterns in the target language. For
a new sentence, these correspondence
rules can be used to replace the source
patterns to the corresponding target
patterns to produce a sentence in
the target language. In RBMT, these
patterns are manually identified and
coded into the system. SMT, as the
name suggests, relies on statistical
techniques to address this problem.
Main components of any SMT system are
Translation Model (TM) and Language
Model (LM). Translation Model basically
captures the pattern replacement rules
and is learnt by an SMT system from
a bilingual parallel corpus. This is a
probabilistic process and consequently
for a given pattern comes up more than
one target pattern. A language model
(LM) is used to select the right target
pattern, looking at how sentences are
constructed in the target language.
Language Model is learnt using a
target language corpus. The language
model scores the sentences generated
by TM, based on their fluency in the
target language which is represented
by the aggregate probability of word
sequences in that language.
There are variants of the SMT
model. We have used phrase based
SMT approach in Anuvad. Though
SMT can save the effort of manually
generating transformation rules, going
to a fully statistical model raises many
challenges. In practice, we need a
hybrid model, where other knowledge
sources are used to make the job of
SMT less error prone. Keeping this
in mind, along with the use of core
phrase based SMT, we have developed
and incorporated various techniques to
 ARTIClE
handle complexities involved in MT for
Indian languages. These are described
in the next section.
Anuvad - An English to Indian
language SMT system
Development of Anuvad was started
with focus on English to Hindi and
Marathi translation. Further we have
tried to extend the approach for other
major Indian languages, basically from
Indo-Aryan and Dravidian language
families and were able to achieve some
improvements there also. If you like
to read more details about our work,
please refer our publications openly
available at http://kbcs.in .
We have developed systems for
English to Hindi, Marathi, Punjabi,
Bengali, Tamil, Gujarati, Telugu,
Konkani and Malayalam. Size of
the parallel corpus available is
significantly important in determining
the performance of SMT systems, and
in most cases, large aligned corpus
is not available for Indian languages.
More work in this area is very important
to produce good quality translation
systems in our languages.
An MT system, like any other
Machine Learning system, has two
phases, the training phase and testing/
usage phase. In the training phase, the
models are created, and these are used
to produce translations in the testing
phase. Fig 1 shows block diagram of
Anuvad System at an abstract level.
We can see that, Anuvad makes use
of Reordering, Stemming, Splitting and
Fig. 1 : Anuvad Block Diagram
C S I C o mm u n I C a t I o n S | n o V E m B ER 2 0 1 7
Transliteration apart from the core SMT
training and decoding modules. These
are the improvements we worked on, as
mentioned above. These are discussed
in more detail below. The block diagram
shows English as the source language
and Hindi as the target language.
Fig 2 : Reordering Example
As we know, structure of an English
sentence is Subject – Verb – Object (SVO),
whereas for those Indian languages we
were looking at, primarily it’s Subject –
Object – Verb (SOV) and we sometimes
observe free word order also. For
example, the English sentence ‘Ram
Killed Ravan.’ (SVO) can be translated
to Hindi as ‘Raamne Raavanko maaraa.’
(SVO); translation ‘Raavanko Raamne
maaraa.’ is also frequently seen. In
the presence of adjectives, phrases,
and sub-ordinate clauses, this re-
ordering can mean movement across
very many words. Such movements are
hard to capture statistically. To ease this
difference of word order, we reorder
the English sentence to Hindi structure
in an approximate fashion. After
reordering, the above English sentence
will become ‘Ram Ravan Killed’. Fig 2
has a complete example.
Another characteristic typical of
many Indian languages is morphological
26
richness; a word takes many different
forms depending on the context it is
used in. keÀjsiee, keÀjsbies,keÀjesies, keÀ©biee etc. for
the word kar (do). By default, these
words are treated as independent
words, and makes the pattern detection
problem harder. Further, even if we
know karoonga-kartha hai (keÀ©biee –
keÀjlee nw) combination, we are unable to
convert jaoonga (peeTbiee) to jaatha hai
(peelee nw). To alleviate this problem, at
the preprocessing stage we also apply
stemming and compound word splitting
to handle morphological difference
in the source and target languages.
Stemming, splitting and any other
preprocessing, for that matter, are
applied as per the need and nature of
languages involved in translation.
There can be situations when
some of the words from new sentences,
mostly names of entities like person,
place, etc, may not get translated by our
trained system. Such words are then
passed to the transliteration system,
Xlit also developed by C-DAC Mumbai,
which essentially does an intelligent
script conversion. The final output is
presented to the user as translation
which he can choose to use as it is or
edit.
Performance of Anuvad
It’s important to know the accuracy
of a system before relying on it for the
purpose of translations. MT accuracy is
mostly measured using BLEU (Bilingual
Evaluation Understudy) score. Many
times translations are also evaluated
manually, but are time consuming and
costly.
Though BLEU scores are used
widely, averages over multiple human
scores is the best way to judge accuracy
of system for deployment in the field.
In table 1, we detail the manual scores
of Anuvad for some language pairs.
This evaluation is done by an external
agency (MTIL17) and translations were
manually evaluated by three different
language experts and then averaged. In
manual evaluation, the translations are
scored for adequacy and fluency. Thats
is, this evaluation looks at how much
information from the source sentence
is carried into the target (adequacy) and
how native/fluent the translation is in
the target language. The final score is
average of these two scores.
www.csi-india.org
 ARTIClE

the role of an assistant for translators.  Koehn, Philipp. Statistical m a

Language Pair Expert Score/
Accuracy (in )
English-Hindi 68.64
English-Punjabi 58.34
English-Tamil 48.00
English-Malayalam 31.94
Table 1 : Manual Evaluation Results
for Anuvad
Anuvad has been tested multiple
times internally and externally. It has
been awarded best MT system for Indian
languages, twice by ‘International
Conference on Natural Language
Processing’ (ICON) and once by ‘Shared
Task Cum Workshop on Machine
Translation in Indian Languages (MTIL)’.
Our work on these languages along
with other Indian languages is still in
progress and we aim to achieve better
accuracy with improved techniques and
use of a bigger parallel corpora.
Applications of Machine Translation
Being a trainable system, an SMT
system can be used for creating a
custom machine translation system for
various domains. As the accuracy of MT
is yet to reach a level when machine
generated translations can be used as
it is, a system built with SMT can play
Following are some sample application
scenarios for the same.
 Customized MT system for
Administrative, Legal, Technical,
etc. domains with higher accuracy.
 Annual report translations
 Government agencies wanting to
make use of MT, but not willing to
share their data with third party
can have their own SMT system.
 System localisation for software
applications.
For more information about our
work in AI, Machine Learning, Machine
Translation, NLP and Data Science,
please visit our website www.kbcs.in or
write to kbcs@cdac.in.
References
 Ananthakrishnan Ramanathan,
Pushpak Bhattacharyya, Jayprasad
Hegde, Ritesh M. Shah, and
M. Sasikumar. 2008. Simple
syntactic and morphological
processing can help English-Hindi
statistical machine translation. In
International Joint Conference on
NLP, 2008.
 Jurafsky, Dan. Speech & la ngu
age
processing. Pearson Education
India, 2000.
ch
ine
translation. Cambridge University
Press, 2009.
 Philipp Koehn, Hieu H
o
ang,
Alexandra Birch, Chris Callison-
Burch, Marcello Federico, Nicola
Bertoldi, Brooke Cowan, Wade
Shen, Christine Moran, Richard
Zens, Chris Dyer, Ondrej Bojar,
Alexandra Constantin, Evan Herbst,
Moses: Open Source Toolkit for
Statistical Machine Translation,
Annual Meeting of the Association
for Computational Linguistics
(ACL), demonstration session,
Prague, Czech Republic, June
2007.
 Prakash B Pimpale, Raj N ath
Patel, Sasikumar M, “ SMT from
Agglutinative Languages: Use
of Suffix Separation and Word
Splitting”, The 11th International
Conference on Natural Language
Processing (ICON 2014) , Goa
University, Goa, India, December
18-21, 2014.
 Raj Nath Patel, Rohit Gupta,
Prakash B. Pimpale and Sasikumar
M, Reordering rules for English-
Hindi SMT, In Second Workshop on
Hybrid Approaches to Translation,
ACL 2013, Sofia, Bulgaria, August
4-9, 2013
n

About the Authors

M. Sasikumar (CSI Life Membership No. 01202409) Prakash B. Pimpale works as Senior
is Director of Centre for Development of Advanced Technical Officer with C-DAC Mumbai.
Computing, Mumbai (C-DAC Mumbai). He has been He works with KBCS division and is
involved into conceptualisation and overall design of involved in R&D projects in MT, NLP, ML
and Data Analytics.
a number of research projects in e-learning, natural
language processing, accessibility and artificial
intelligence.

“geekExpo” - Project Exhibition

“GeekExpo”- A project exhibition was held at SSIPMT Raipur on 5th October
2017 in association with CSI Raipur Chapter and Center of Excellence of
CSE & IT. In which 22 groups presented their Projects/Models/Ideas based
on IoT/Webapps/Mobile Apps. Judges are Mr. Nishant Tripathi (Chairman,
BG SSIPMT) Dr. Alok Jain (Principal SSIPMT), Dr. J. P. Patra (Secretary,
Raipur Chapter), Prof. Riju Bhattacharya (CSE), Prof. Gurudatta Verma
(CSE) and Owner of Gifttrees & Codenicely Startups. They motivate and
appreciated the efforts of students and one participant got paid internship
by Codenicely.

27
C S I C o mm u n I C a t I o n S | n o V E m B ER 2 0 1 7
 ARTIClE
Machine learning at the Edge of Internet of
Things
Mohammad Khalid Pandit, Roohie Naaz Mir and Mohammad Ahsan Chisthi
Computer Science and Engineering Department, National Institute of Technology Srinagar. Email: Khalid@nitsri.net
INTRODUCTION
Internet of Things (IoT) has recently
emerged as one of the buzzwords in
the IT world. It serves as a catalyst for
rapidly growing smart networks with the
deployment of sensors, communication
and control infrastructure at a rapid
pace. IoT envisions the concept in
which all the physical objects, generally
referred to as “things” get connected
to each other and the internet for
information sharing. IoT goes beyond
the classical machine to machine
paradigm whereas later only focuses
on the communication infrastructure
(protocols), IoT envisions not only
communication infrastructure but an
intelligent network that can be sensed
and programmed using state of art
techniques from machine learning and
neural networks. It enables new kind
of applications ranging from smart
homes, smart body implants to self-
driving cars and much more. It enables
new kind of applications ranging from
smart homes, smart body implants to
self-driving cars and much more. These
devices generate huge amounts of data,
according to IBM 2.5 exabytes (EB) -
that’s 2.5 billion gigabytes (GB) - of data
was generated every day in 2012 [1].
Estimates suggest that 90% of total data
has been produced in last 2 years, and
current data production output is almost
2.5 quintillion bytes per day. Currently
we have almost 20 billion connected
devices and IHS Markit projects 30.7
billion IoT devices by the end of 2020
[2]. In order to create a truly smart IoT
application the massive data that gets
generated from the sensors needs to be
processed and analysed into actionable
information.
The traditional approach of using
centralized architecture of cloud,
wherein the devices (end points in IoT)
are only used for data sensing, data
C S I C o mm u n I C a t I o n S | n o V E m B ER 2 0 1 7
forwarding and all the analysis happens
at the cloud which makes decisions
based on current and historical data.
These decisions are then transported
back to the edge of network for actions.
Fig 1. Big Data Statistics [11]
For the dream of IoT to be
successful, the enormous data that it
generates has to be processed in real
time and classified into actionable
information. Currently cloud is the
natural choice for large machine
learning algorithms to run and process
data. The presence of huge compute
power of cloud coupled with modern
Graphics Processing Units (GPU’s) lend
naturally towards processing these
huge amounts of data.
However with continuous incoming
data from sensors, cloud being the
only analytical engine introduces many
extra complications. For example.
connection is always required to the
cloud, unilateral trust between client
and cloud service. It can also incur loss
of privacy by releasing the personal
information to the centralized service
like cloud. Security remains the prime
concern of cloud computing as the raw
data generated by the sensors is being
send to the cloud servers. Reiter et al [3]
28
discusses the problem of high latency
of over 100’s of ms associated with the
cloud.
There has been a tremendous
increase in the computing power and
storage capabilities of the personalized
devices (smart phones, IoT gateways).
Smartphones today are million times
more powerful than all of NASA’s
combined computing in 1969. Popular
IoT gateway Raspberry pi has a 64 bit
quad core ARM processor, each of
which offers one third of computational
capabilities of Intel xeon processor
core over the cloud [4]. With such
computing capabilities around us in
form of IoT edge we feel there is an
opportunity in exploiting this compute
capability & storage power of modern
personal devices. This shift gives rise to
a new computing paradigm called Edge
computing. This new paradigm will help
in de-centralizing the computing from
cloud towards the edge of IoT. Edge
computing will complement the cloud
as more computationally intensive
tasks will be delegated to the cloud for
final outcome.
why Intelligent Edge?
Delegating more powers to the
edge than we have in the traditional
centralized approach requires that
some of the computation on the data
generated by IoT sensors to be done
on edge itself. The reasons for edge
computing becoming necessary are:
Data generation at the edge: most
of the data for analytics is produced at
the edge of network. Natural and easier
way is to process that data at the edge
itself, with the exception of historical
data that lies in the cloud.
Network and energy constraints:
the issue of latency becomes
unacceptable for some of the mission
critical applications. Edge devices
are not always connected through
www.csi-india.org
 ARTIClE
low latency network; instead they use
UMTS and LTE connections. Reiter et
al [3] shows that UMTS connections are
subject to large variations of round trip
times (RTT) of more than 100ms during
peak traffic hours.
Transferring 1 bit of data over
cellular network requires about 500
micro joules (µj) of energy. With the
amount of data that is being generated,
the energy consumption for transferring
the data will be huge and beyond our
imagination.
30 FB
Data Usage
3.7 EB
1 PB
10 GB
2000 2005 2015
Fig 2. Monthly Data Usage
Monthly data usage (fig 2) in 2005
was about 10GB, today we use more
than that amount of data in a day. There
is about 100 times increase in 2005
with 1 PetaBytes (PB) of data usage
and more than 1000 times increase in
2005 to 2015 at about 3.7 EB per month.
Estimates suggest that more than 30EB
of monthly data usage in 2020. To simply
calculate the energy requirements for
transferring these huge amounts of
data over cellular network it will require
30EB * 500µj 500 terra watt hours
It roughly estimates to about 2% of
entire world’s electricity consumption
to be used only for sending data over
cellular network [10]. This is the
staggering number as it does not even
include the energy consumption for
data processing. So, it is high time
for us to devise the solutions in which
we can do more with data at the edge
leveraging its computational power and
sending lesser amounts of data into the
network.
Human Centered Computing:
centralized architecture of cloud is
not suited for novel human centered
computing, that would allow humans
to be the owner of their data. Making
edge as computing resource will have
an advantage of using cloud as support
infrastructure but puts decision related
to trust and control back to edge and
is well suited for human centered
computing.
Machine learning at the Edge
Machine learning especially
artificial neural networks are the
natural solutions to process such
huge data that gets generated in
IoT environment. It can be used for
classification, object recognition,
natural language processing, speech
recognition etc. currently cloud is the
only choice for training and evaluating
deep neural networks due to its power
and scalability. As we have already
discussed the short comings using
cloud only solution. Therefore, the
goal is to distribute the neural network
among the local compute power (fig 3)
present in the IoT scenario (devices,
gateways, embedded GPU’s).
2020
Fig 3 : Role of Edge and Cloud services
Training a deep neural network is
computationally very expensive task, but
the trained network have to be deployed
the real world IoT environments. The
distribution of layers of neural network
among devices wills speedup the
evaluation. The layers will be distributed
among the local compute power and the
cloud. The deeper layers (that will be in
the cloud) will only be needed when the
lower layers (at the edge) will not be able
to make the classification precisely. The
local layers can be used to output the
urgent decisions and the more detailed
information could be send to cloud for
further processing.
Sam et al [8] proposed an
architecture of cascade of neural
network layers among devices in which
only one output layer is not trained
29
C S I C o mm u n I C a t I o n S | n o V E m B ER 2 0 1 7
but multiple output layers are trained
for each hidden layer. It runs the part
of network locally and part of it over
the cloud. Only those samples can be
classified locally which are processed
with sufficient confidence (above
threshold). The remaining samples
are send to the cloud for further
processing. LeCun et al [6] performed
the experiments on MNIST data set
(fig 4) which is the repository of over
70000 images of handwritten digits. The
experiments showed that the neural
network with just two layers (i.e one
hidden layer) is able to achieve 4.7%
error rate, i.e. 95.3% of the samples of
MNIST data set are classified by only one
hidden layer with sufficient confidence.
Fig 4 : MNIST Dataset [12]
This experiment clearly suggests
that more than 95% of these samples
do not require the additional deeper
layers and could be classified in single
layer of neural network which can
be easily implemented on the device
itself. Additionally we can only send the
processed form of the unclassified data
to the cloud, which can lower the energy
and cost of data transmission.
Qualcomm researchers are trying
to develop the neuromorphic chips
which can execute the neural network
in hardware, which will be clearly faster
than the software simulated neural
networks. Neuromorphic hardware
closely mimics the working architecture
of human brain [9]. They are trying
to come up with neural processing
units (NPUs) that will replace or work
alongside the traditional CPUs and
GPUs that are powering computers
today. Qualcomm’s Zeroth machine
intelligence engine provides a software
development kit that will allow
developers to perform many of the
same machine learning capabilities
 ARTIClE

of cloud on the device (edge) through
snapdragon 820 System on Chip (SoC).
Elias et al [7] proposed a big little
architecture for distributed neural
networks for internet of things. The
neural network is partitioned into big
and small (little) networks. The little
network is housed in the edge device
itself and the big network is put in the
cloud. The aim here is to use the local
compute power first (little network)
before turning up to the cloud. This will
drastically save the network, energy and
cost requirements. [7] proposes the use
of one vs. all or one vs. one classifiers to
speed up the execution of small neural
network which is distributed among the
local compute power.
Microsoft researchers [5] proposed
the use of tree based learning
procedure ‘Bonsai’ for resource
constrained devices which are typical to
IoT scenario. Bonsai aims at providing
the alternative paradigm where in IoT
devices can make decisions locally
without the compulsory connection
to the cloud. Also sometimes the
energy requirements for executing the
instruction may be much lower than
the energy required to transmit the data
thereby extending the battery life.
Conclusion
This article outlines the need for
the computing at the edge to solve the
energy and performance issues. It will
be compliment to the cloud paradigm
whereby the ML algorithms will run
on the small scale devices driving the
localised decision making. With the
amount of data being generated and IoT
device population surpassing human
population, the energy requirements
highlighted makes it imperative to do
a lot more with the data at the edge
itself and sending far less data into the
network.
There is a need for investigating the
opportunities of distributing the deep
neural networks and other machine
learning algorithms across the edge,
fog and cloud devices. This seems
the only way by which we can actually
realise the dream of IoT, making devices
truly smart.
References
[1] IBM Annual Report 2013 https://www.
ibm.com/annualreport/2013.
[2] https://spectrum.ieee.org/tech-talk/
telecom/internet/popular-internet-of-
things-forecast-of-50-billion-devices-
by-2020-is-outdated
[3] Reiter, Andreas, Bernd Prünster,
and Thomas Zefferer. “Hybrid Mobile
Edge Computing: Unleashing the Full
Potential of Edge Computing in Mobile
Device Use Cases.” Proceedings of
the 17th IEEE/ACM International
Symposium on Cluster, Cloud and Grid
Computing. IEEE Press, 2017.
[4] Yogesh Simmhan, “IOT Analytics Across
Edge and Cloud Platforms.” IEEE
Internet of Things, May 2017.
[5] Kumar, Ashish, Saurabh Goyal, and
Manik Varma. “Resource-efficient
Machine Learning in 2 KB RAM for
the Internet of Things.” International
Conference on Machine Learning. 2017.
[6] LeCun, Yann, et al. “Gradient-
based learning applied to document
recognition.” Proceedings of the
IEEE 86.11 (1998): 2278-2324.
[7] De Coninck, Elias, et al. “Distributed
neural networks for Internet of Things:
the Big-Little approach.” Internet of
Things. IoT Infrastructures: Second
International Summit, IoT 360° 2015,
Rome, Italy, October 27-29, 2015,
Revised Selected Papers, Part II.
Springer International Publishing, 2016.
[8] Leroux, Sam, et al. “Resource-
constrained classification using
a cascade of neural network
layers.” Neural Networks (IJCNN),
2015 International Joint Conference on.
IEEE, 2015.
[9] Monroe, Don. “Neuromorphic
computing gets ready for the (really) big
time.” Communications of the ACM 57.6
(2014): 13-15.
[10] David Brooks “Hubweek Symposium
09/30, David Brooks: Taking
Machine Learning to Edge Devices”
6-october-2016.
[11] Data Never Sleeps 3.0 https://
www.domo.com/blog/data -never-
sleeps-3-0/.
[12] MNIST data set http://csc.lsu.
edu/~saikat/n-mnist/.

About the Authors

Mohammad Khalid Pandit is currently pursuing his PhD from National Institute of Technology Srinagar in the
Department of Computer Science and engineering. He received Bachelor of Engineering (B.E.) and M.E. in Computer
Science and Engineering from Anna University Chennai. He worked as Assistant Professor in department of CSE at NIT
Srinagar from September 2013 to May 2017. He has served as student member of CSI, and is the member of IACSIT and
IAENG.. His research interests include Artificial Intelligence, Machine learning and Edge computing.
Roohie N. Mir is a professor in the Department of Computer Science &amp; Engineering at NIT Srinagar, INDIA. She
received B.E. (Hons) in Electrical Engineering from University of Kashmir (India) in 1985, M.E. in Computer Science &
Engineering from IISc Bangalore (India) in 1990 and Ph.D from University of Kashmir, (India) in 2005. She is a Fellow
of IEI and IETE India, senior member of IEEE and a member of IACSIT and IAENG. She is the author of many scientific
publications in international journals and conferences. Her current research interests include AI and Machine learning,
reconfigurable computing and architecture, mobile and pervasive computing and network security.
Dr. Mohammad Ahsan Chishti [CSI - I1501576] has done his Doctor of Philosophy (Ph.D.) from National Institute
of Technology Srinagar. He has completed Bachelor of Engineering (B.E.) and M.S. in Computer and Information
Engineering (MSCIE) from International Islamic University Malaysia. Presently he is working as Assistant Professor
and Head of the Department of Computer Science &amp; Engineering, National Institute of Technology Srinagar. He
has more than 50 research publications to his credit and 10 patents with two granted International Patents. He has
successfully completed Department of Science &amp; Technology sponsored research project. He has been awarded
“IEI Young Engineers Award 2015-2016” in the field of Computer Engineering for the year 2015-16 by the Institution of
Engineers (India) and “Young Scientist Award 2009-2010” from Department of Science &amp; Technology, Government
of Jammu and Kashmir for the year 2009-2010. He has been Executive Member, Institute of Engineers India (IEI) of
J&K Chapter.

www.csi-india.org
30
C S I C o mm u n I C a t I o n S | n o V E m B ER 2 0 1 7
 S E CU R IT Y C OR N ER

Empirical Survey on Advances of Network

Forensics in the Emerging Networks
A. R. Jayakrishnan
Research Scholar, Rathinam College of Arts and Science,
(affiliated to Bharathiar Univ.), Coimbatore, Tamilnadu, India.
E-mail: anilakkad.jk@gmail.com
V. Vasanthi
Research Supervisor, Rathinam College of Arts and Science,
(affiliated to Bharathiar Univ.), Coimbatore, Tamilnadu, India.
E-mail: vasanthi.cs@rathinamcollege.com

In the modern world, where most of the population is connected to the internet via smart devices,
the surface for committing crimes and being attacked by hackers have been increased drastically.
Number of cybercrimes keep increasing and may ramp up when IoT, M2M, 5g and other new network
technologies become more popular in coming years. The electronic crimes that involve the use of
computer networks are considered the most complex category. After an attack is identified on a
network, there is a necessity for computer forensics to investigate and gather digital evidence to bring
to justice. Network forensics is dealt with the dynamic and volatile information instead of static/stored
data. hence, highly sophisticated techniques are required for instantaneously tracing down criminals
network activities. A honeypot tool can be used to trap intruders and collect essential information on
the crime activities. Classification of network forensics, reviews on emerging network areas, intrusion
and prevention methods, attack detection process, network forensic process, reviews on various
NFATs, survey on network forensic approaches are incorporated in this paper.

Keywords: Cyber Forensics, Network Forensics, Digital Security, Live Forensics Mechanisms, NFAT, Honeypot

1. Introduction
Computer forensic experts
often face difficulties during data
collection from a network to aid in
forensic investigation. The crime scene
constantly changes and as the traditional
crimes reduce in number, the cyber
crimes incidents are occurring more
frequently. To comprehensively solve the
emerging network problems through
network forensics, understanding the
context of forensic investigation and the
general nature of emerging networks is
the key issue.
Network forensics is a scientific
field that deals with the collection,
recovering and the examinations of
network events for the purposes of
establishing sources of security attacks.
This phenomenon is useful in the
identification of unauthorized access
to computer network systems and is
an investigative tool in case of such
occurrences. According to Alzaabin [1],
network forensics plays an investigative
role at network level, the issues that
take place or have occurred within an
information technology system. The
major components of network forensics
include intrusion detection through
logging and correlating with the process.
Network forensics plays a primary role
in the provision of sufficient evidence
to allow the perpetrators of crime to
be successfully persecuted. Practical
application of network forensics in
emerging networks could be present
in areas such as email investigation,
hacking, insurance companies, fraud
detection, drug trafficking, software
piracy, credit card cloning, electoral law,
discrimination and sexual harassment
[2].
This study aims at conducting
a literature survey to ascertain the
emerging network forensic areas
and identify effective architectures
for maintaining network security. The
process for network attacks detection,
network forensic process and the
essential tools for network forensic
analysis shall also be analyzed to
compile the study. Questions of network
security issues and attacks and the
investigation mechanisms used in a
network security accident will also be
discussed in the study. Concerns as
whether network forensic mechanisms
can be used in eliminating network
security issues and attacks will also be
considered in the research.
2. Background
The word computer crime was first
used in 1976 by Donn Parker in a book
“crime in computer” [12], [13] that it is
referred as dealing with unauthorized
modification or deletion of data in a
computer system [14]. Nevertheless, FBI
was established first actual computer
analysis and response team in 1984
to conduct advanced digital forensic
investigation of the crime scenes [15].
It was in 1986 the first complicated
digital forensic investigation case was
executed, chasing a hacker named
Markus Hess [16]. Mr Hess had

33
C S I C o mm u n I C a t I o n S | n o V E m B ER 2 0 1 7
 S E CU R IT Y C OR N ER
intruded to the LBL, the Lawrence-
Berkeley-National-Laboratory. It was
detected and investigated by Mr. Clifford
Stoll. There wasn’t any standard digital
forensic investigation framework in
place at the time of the incident, so
Stoll had to carry out the investigation
on his own. As Stoll’s objective was
discovering the identity of the intruder,
he didn’t alter anything in the system
and only collected the possible traces.
He finally managed to discover the
identity and location of the attacker by
tracking the hacker for many months
using alarms which send notification
when the attacker was active with the
cooperation of FBI and Telco Company.
The case became a big issue since it
was involved academic and individual
bodies in U.S, Germany and military
bodies [17]. Day by day improvement
of digital devices makes digital crime
more complicated than it was in 1986.
Our augmented dependency on latest
networked technologies and the
prevalent inter-connectivity of systems
used in our networked modern world
are potentially be exploited for criminal
purposes (e.g. malware, hacking,
phishing, theft of trade secrets &
intellectual properties etc.). There are
basically two approaches for network
protection. Defensive Mechanism is used
to defend from intruders by blocking
malicious communications on network.
Tools like Firewall, IDS are used by
this method. Firewalls control network
traffic according to the rules set, but has
many limitations. IDS is primarily a real
time system for detecting and reporting
attacks as it happens on the fly, however
it has no evidence gathering capability.
Preventive Mechanism will not block
the network attack, but tries to prevent
it by access control or authentication.
Reliable and comprehensive network
security and forensics architecture are
essential in maintaining security of the
emerging networks.
2.1 Definition
Network forensics is a science that
monitors and analyzes network traffic,
aiming at detecting malicious network
activities and preserving the trace to
produce as legal evidence at court. It
deals with capturing, recording and
analyzing of network traffic in order to
discover the source of attack and the
C S I C o mm u n I C a t I o n S | n o V E m B ER 2 0 1 7
criminals behind it. From the existing
network security products and tools,
network log data are collected, analyzed
for attack classification and investigated
to trace through the perpetuators. This
process can bring out deficiencies in
security products which can be utilized
to guide deployment and improvement
of these security tools as well. The scope
of network related incidents should be
quickly determined and also be able
to rewind the breach by the forensics
investigator so that they can detect
where to look for the potential evidence.
Network forensic investigator will
require the same thought process and
skill set as hackers in order to detect
why and what has happened and also
to identify the motivation for the crime.
Different networks may be spanned
across multiple geographical zones and
various jurisdiction, which demands
the use of absolute trusted timestamps
to make sure the integrity and
authentication of timestamps for each
portion of network evidence ensuring
that all jurisdictions collaborate. The
network data will be available in both
real time and off line modes, the former
requiring the ability to pick and analyze
data on the fly. The amount of data could
be massive due to the increasing size of
network bandwidth and also it could
involve various different protocols. A
protocol could in turn involve multiple
layers of signals. The current available
set of forensics tools may not be able
to handle the real-time and huge data
volume. Various techniques are required
for instantaneously tracing criminal’s
network activities on computers and for
mapping the networks topology.
For any cyber-attack, an evidence
trajectory would be left in IDS - intrusion
detection systems, firewalls, proxies,
routers and within the network traffic.
Hence, there is an increasing need
for investigators to analyze network
traffic, events, logs, netflow and digital
device log, in order to discover how an
attack was carried out. Such activities
could lead in reconstruction of a
crime and eventually the identification
of the criminal(s). Rigorous
methodology and set of procedures
for conducting network forensics and
investigations are very much required
in place. However, the increasing use of
32
networked technologies, fast advancing
networking technologies and the need
to deal with volatile and dynamic live
data, complicates the efforts to identify,
collect and preserve evidence data in a
timely fashion.
In order to keep up to the pace
with the modern changing face of
criminal activities and its growth, It is
important for the forensic research
and practitioner communities to have
an in-depth understanding of the types
of artefacts that are likely to remain on
network and on networked devices as
well as the capability to undertake data
collection and acquisition in well-paced
and in a forensically sound manner.
Network security is not another
term for network forensics. Network
forensics is an extended phase of
network security. Network security
protects systems against attacks
whereas network forensics focuses
on capturing evidences of the attacks.
Network forensics is considered as
an extension of computer forensics.
Monitoring is a continuous process.
However, network forensics involves
post mortem investigation of the attack
and is initiated post crime notification.
It is case specific since each crime
development is different and the
process is time sensitive.
2.2 Classification
Network forensics can be classified
into two types: “Catch-it-as-you-can”
systems, whereby all the packets that
are passing a specific traffic point are
captured and analyzed subsequently in
batch mode. This method requires lots
of storage space as it captures volumes
of data. Also, searching and finding
required information from the stored
data is a big overhead. “Stop, look and
listen” is another network forensics
system that analyzes information in a
rudimentary manner in memory and
only certain information is saved for
future analysis [3]. Fig. 1 depicts the
classification of forensics.
Network Forensics
Catch-it-as-you-can Stop, look and listen
Fig. 1 : Classification of forensics
www.csi-india.org
 S E CU R IT Y C OR N ER
It is often a long and tedious
process that requires standard
operational procedures for the
management. In order to come up
with a standard operating procedure
for network forensics in emerging
networks, it is recommended to have
a standard investigation image for
system. This image should not be
of a bit stream and do not contain
all the standard applications used.
When an intrusion incident occurs,
the vulnerability of the system should
be fixed to prevent any chances of
attacks. All the volatile data should be
acquired through live acquisition before
the system is switched off during the
development of the standard procedure
for network forensics [5]. Additionally, a
compromised drive should be acquired
and an image of it is taken. Comparison
of forensic images should be done
and hash values properly checked to
ascertain if there are any changes in
them [5].
Traditional vs Live - Live forensics
(forensics on a system that cannot be
switched off, as in the critical systems)
and the attribution root question (linking
the criminal activities to the convicts
behind it) are examples of issues that
requires additional research. The
traditional forensics and Live forensics
have common features, where both of
them are looking for similar artifacts
on a system. However, the differentiator
with live response is that the artifacts
are discovered on a live running system
against an active adversary. With
traditional forensics, images are taken
of volatile memory and hard disks before
being analyzed. Imaging alone can take
many hours and then the images should
get processed and indexed to allow for
keyword searches. Obtaining and then
processing the image could take a day
or longer with large capacity disks.
With live response, there won’t be any
scope for imaging or processing that
has to occur, everything will be real
time. This would dramatically improve
the response time in identifying and
quantifying a threat. The quicker the
threat is identified, the quicker it can be
contained and remediated.
3. literature Review
3.1 Emerging Network Forensic
Areas
Network forensics plays an
C S I C o mm u n I C a t I o n S | n o V E m B ER 2 0 1 7
important role in the emerging network
areas such as Machine-to-machine
(M2M), 5G and Internet of Things (IoT).
M2M - The Internet in achieving
intelligent interactions between
machines having different terminals
uses the Machine-to-Machine networks.
This emerging network is composed
of equipments, back-end systems,
transmission links and front-end
sensors. The front end sensors helps in
the collection and transmission data to
the back end systems that is overhauled
and the back end nodes transmits
information to back end control system.
When the M2M networks are attacked,
the terminals it is associated with
becomes affected and the network
is eventually paralyzed and cannot
function. The security issues of M2M
are attracting many concerns as the
encounter of several attacks, hence, the
need for an elaborate forensic expertise
arises. Network security architectures
and protocols are evolving to counteract
the challenges experienced by emerging
networks. This evolution is key in the
unauthorized access prevention leading
to security improvement [18].
5G – Even though, the specifications
of the 5G is still not finalized, but
the technology is expected to be the
fastest speed as high as 10Gbps. The
new technology would lead to ultra-
low latency applications, high volume
data rate applications, network of
network applications. Real time threat
detection and network isolation are
going to be a big challenge for 5G
networks. Interoperability and diversity
in networks increase attack surface
and motivation of attacks. Forensics
of such a superfast network would be
a nightmare for the experts in coming
years. [31]
IoT - The development and
advancement of the Internet and other
smart electronic systems have led to
the emergence of computing prototypes
such the Internet of Things (IoT). The
IoT has been considered as the future
of Internet works with other emerging
networks such as M2M communication
and Radio Frequency Identification
(RFID). IoT has the primary mandate
of adequately securing data exchange
between real world applications
and devices. The paradigm of digital
forensics in line of Internet of Things
is challenging and diverse and the
33
traditional model of forensics does
not adequately address the recent IoT
environment [19].
3.2 Preliminaries of Network Security
and Forensics.
Intrusion detection and prevention
systems are used to monitor network
traffic and activities of the network
system for any malicious activities.
Intrusion prevention system takes
action by sending an alarm, resetting
connection, dropping malicious packets
or blocking traffic from any offensive
IP addresses. Intrusion prevention
System (IPS) is also essential in
cyclic redundancy check (CRC) errors
correction, eliminating extra features
and network layer options as well as
unnecessary packet streams [18].
Honeypots and Forensics - With
the rise of cyber-attacks, Honeypots
have been considered one of the best in
network crime forensics, warm attack
detection and other aspects of active
defense methods. A honey pot is a trap
network that is static and is effective
for attackers who are reckless. When
the attackers are however aware of
the existence of the honey pot, their
functionality becomes degraded. In
order to increase the functionality and
persuasiveness of the Honeypots, the
behaviors of the attackers need to be
collected and stored in a safe place [11].
Honeypots are copies of the real
servers and are often used to simulate
real service through the coordination
of various modules. These modules
include data acquisition module,
remote storage module, detection
module and implementation module.
Data acquisition module captures
and collects the activities of intruders
or attackers. The remote storage
module in the other hand backs up the
network system remotely and captures
data to the log servers and keeps
the information without any further
modification. Detection module protects
the host and records the invasion
process. Implementation module on
the other hand is used in the simulation
and protection of the host and prevents
attacks from any intruders.
Surveillance and Vulnerability
Scanning - During a forensic event,
the surveillance and vulnerability
scanning systems are usually not stored
through the SIEM system. Surveillance
 S E CU R IT Y C OR N ER
& vulnerability systems are potential
parameters in attacks minimization via
firewall rule adaptations or alterations.
The above mentioned preliminaries of
network security and forensics helps in
the detection of all network anomalies
and are useful in the identification of all
the attackers. After the attackers have
been traced by these systems network
forensics then follows. This process
experiences challenges such as the loss
lag of network forensics and to some
extent even loss of evidence. Attack
detection needs to be integrated with
network forensics in order to improve
the timeliness of forensics after the
detection of attackers.
Attack Detection Process - During
the attack, the network defense system
sends out attack alerts. Such alert
information needs to be collected in
order to identify the exact location.
Access authority is then sought and
the behaviors of the attackers analyzed
according to their path and finally
their operations intercepted and the
information of the attackers recorded.
The main attack detection process
includes sending a request to the URL
that has been accessed to request the
collection of data information, recording
the results of traffic anomalies and the
actual suspicious traffic, analysis of the
obtained suspicious traffic and then
accessing the relevant static resources.
The behaviors of the attackers are then
analyzed and finally the interception and
recording of the information will help in
the network criminal forensics process
[4].
3.3 Network Forensics Process
Demand for evidences collection
is changeable in the process of
network forensics. For one to meet
such demand, composition of network
forensic modules has to be designed.
This can be done through network data,
which records the process of network
communication evidence collection
and binary network data packets
preservation. Content crawling network
is one module that can be designed in
meeting the demand of data collection.
This layer helps in the preservation
of documents relevant to a specific
network and that, which is readable
in terms of images and text data. With
such a module, it is easy to extract
C S I C o mm u n I C a t I o n S | n o V E m B ER 2 0 1 7
information for network forensics.
Another module that can be
designed is the data analysis layer. It
analyzes network contents from the
content crawling layer. It extracts and
displays information that is useful and
related to the attackers and finally
helps in the compilation of a report of
evidence. Forensics implementation
layer is an additional module that
is significant in the adjustment of
the evidence collection stability and
flexibility and it modifies the acquisition
of information and filters them in order
to simulate the login rules [7].
During network forensics process,
after the design and compilation of
the layers in the forensics analysis
modules, the network forensic analysis
needs to finish the collection of data,
collection of information, analysis of
the collected data and ultimately and
most importantly the implementation
of the forensics process [8]. A waterfall
model which has been commonly
followed in forensics has various steps
– Preparation, Detection, Collection,
Preservation, Examination, Analysis,
Investigation, Presentation and Incident
response. Feedback from steps
examination, analysis and presentation
can be used for improvement of network
security, tools used etc. Fig. 2 depicts
the process of network forensics.
Preparation
Improvement of Security Tools
Detection
Collection
Preservation
Examination
Analysis
Iterative
Investigation
Process
Presentation
Fig. 2 : Network forensics process
3.4 Network Forensics Analysis Tools
(NFAT)
Network Forensic Analysis Tools
(NFATs) are essential forensic tools used
in the collection of data and aggregation
of data form various security tools. They
function in IP security provision, inside
34
and outside network attack detection,
data recovery, risk analysis, attack
patterns detection, future attacks
prediction and detection of anomalies
in the network system. The main
properties of network forensics analysis
tools include collection of information
[6].
Some of the network forensics
analysis tools include but are not limited
to NetIntercept, which collects and
analyzes bundles of traffic, Iris, which
collects data from the Internet, then
reassembles them and reconstructs
the actual text from the session.
NetDetector is another NFATs that
capture the attack, integrates signature
anomalies and reconstructs the actual
text from the session [10]. NetDetector
also supports network interfaces and
analyzes report on network traffic.
Silent Runner is a network forensic tool
that primarily focuses on inside threats
and analyzes the threats in a three
dimension on the network in order to
monitor all the packets passing through
a specific network. Other network
forensic tools include NetWitness,
NetworkMinor, and NetStumbler etc [9].
3.5 Literature survey evaluation on
Network Forensic approaches
Evidence Graphs for Network
Forensics Analysis [21], is a technique
for network forensics analysis which
includes manipulation, effective
evidence presentation and automated
Incident Response reasoning. Also included is an evidence
graph which facilitates the presentation
and manipulation of intrusion evidence.
Automated evidence analysis has
an ordered hierarchical reasoning
framework that includes local and
global reasoning. Local reasoning
aims to infer the roles of suspicious
hosts from local observations. Global
Reasoning aims to identify group of
strongly correlated hosts in the attack
and derive their relationships. Most
comprehensive and sophisticated step
is the analysis step. More realistic
experiments and investigate methods to
automate the process for hypothesizing
missing evidence and validating
hypotheses are needed in the model in
local and global reasoning process as
mentioned by the authors.
Step by step framework [22], has
detailed about merging the previous
www.csi-india.org
 S E CU R IT Y C OR N ER
frameworks to form a reasonably
complete framework that groups
all the existing processes into three
stages - preparation, investigation and
presentation. They are implemented
as guidelines in network forensics. The
aim of the framework is to establish
guideline of what steps should be
followed in a forensic process. However,
it looks to be difficult in understanding
how the Forensics Zachman (FORZA)
[23] depicts a framework that focuses
on the legal rules and participants in
the organization rather than technical
procedures. The framework tries to
solve complex problems by integrating
answers for the queries why (the
motivation), what (the data attributes),
who (the people involved), where (the
location), how (the procedures) and
when (the time) questions. Framework
also includes eight rules namely case
leader, legal advisor, system or business
owner, security or system architect
or auditor, digital forensic specialist,
digital forensic analyst, digital forensic
investigator or system administrator
or operator and legal prosecutor.
Framework is human dependent and
Table 1: literature survey on Network Forensic approaches
Author Proposed work
Stephenson et. al Forensics Zach man
2003 (FORZA)
Kohn et. al 2006 Step-by-step framework
Khatir et. al 2008 Two-dimensional evidence Consists of 16 sub phases and are grouped No standard method of detection and
reliability amplification them in to five main phases. This methodology collection of evidence exists, it affects the
process.
Yong-Dal et. al Digital forensics
2008 investigation procedure
model
Rahayu et. al 2008 categorization of
investigation process
Wei et. al 2010 Evidence graphs for
Network Forensics
Analysis
C S I C o mm u n I C a t I o n S | n o V E m B ER 2 0 1 7
requires more tools to conduct network reduction and organization), and
forensic analysis in order to provide post analysis (report and resolution).
accurate results in investigation phase. Standard methods of detecting and
Two-dimensional evidence collecting evidence doesn’t exist, which
reliability amplification process produces insignificant evidence and
model [24], it consists of sixteen sub- it affects the accuracy of the incident
phases and they are then grouped in response as well.
to five main phases namely initializing, Digital forensics investigation
evidence collection, evidence procedure model [25], it consists of ten
examination or analysis, presentation phases with investigation preparation,
and case termination. All phases of classifying cybercrime & deciding
the model are explained in detail by investigation priority, investigating
identifying the roles of the inspector damaged (victims) digital crime
and manager in each phase. The model scene, criminal profiling consultant
aims to provide answers to common and analysis, tracking suspects,
cybercrime queries, however it doesn’t investigating injurer digital crime scene,
focus on intention and strategic analysis beckoning suspect, writing criminal
(why and how questions). It has much profiling and reporting. Model presents
similarity with incident response and block diagram without any technical
computer forensics. Both presents a details or process to manipulate with
common process model for incident all these phases. Main focus was on
response and computer forensics to the number and the type of the network
improve the investigation phase. The forensics phase rather than how they
model includes steps grouped into three works and how they conduct outcomes.
main phases consisting of pre analysis Categorization of investigation
(detection of incidents, initial response procedure was done [26] to group and
and formulation of response strategy), merge the similar processes in five
analysis (live response, data recovery, phases that provide the same outcome.
forensic duplication, harvesting, The phases lists: Phase-1 (preparation),
Advantages Disadvantages
The framework could solve complex problems It is human dependent. Requires many tools
by integrating answers for the queries why to conduct network forensic analysis in order
(motivation factor), what (data attributes), to deliver accurate results in investigation
who (people involved), where (location), how phase.
(procedures) and when (time) questions.
It reasonably has complete framework which Clarity of understanding how framework
groups all existing processes in to three stages addresses all the phases of network forensics
like preparation, investigation and presentation, in main stages is a bit difficult.
which are implemented as guidelines in
Network forensics.
aims to provide answers to most of cybercrime accuracy of incident response.
queries.
Contains ten phases. This model shows block Main focus deviated from how the phases
diagram without technical details or methods to work and how they organize the outcomes.
manipulate with these phases. Rather it focused on the volume and the type
of network forensics phases.
proposes a mapping process of digital forensic Little bit of unclarity on grouping and
investigation process model to eradicate the merging the various process and arriving at
redundancy of the process involved in the model 5 processes.
and standardize the terms used in achieving the
investigation goal.
It includes effective evidence presentation, The model needs to be refined in both native
automated reasoning and manipulation of and global reasoning process with more
intrusion evidence with a graph. representative experiments.
33
 S E CU R IT Y C OR N ER
Phase-2 (collection & preservation),
Phase-3 (analysis & examination),
Phase-4 (presentation & reporting)
and Phase-5 (publishing the case).
Also proposes a mapping process of
digital forensic investigation process
model to eradicate the redundancy of
the process involved in the model and
standardize the terms used in achieving
the investigation goal.
4. Conclusion
Network forensics is the scientific
process that ensures investigation of
attacks that are performed in network
or network devices. The significantly
surfacing problem in emerging
networks is the constant change of
the crime scene. An increase in the
advancement of technology attracts
network revolution that calls for
effective methods for monitoring
and control of security and forensics.
This paper effectively addressed how
network forensic mechanisms can be
used to eradicate network security
attacks. The researches failed to
deal with the concerns raised by the
emergence of new networks such as
M2M and IoT as per the advancement of
improved network forensics essential in
handling the situation of attacks. It is of
essence to comprehensively analyze the
emerging networks to help with forensic
investigation. Network forensics plays
a crucial role in new and developing
areas of network and those related to
data mining, social networking, data
visualization and data imaging.
Future Work - With the empirical
survey on advances of network
forensics in emerging networks like IoT,
5G, it has been identified that there is
a need in research to be conducted on
the upcoming IoT (Internet of Things)
forensics, to capture the network
attacks and identify attack patterns
using honeypots.
5. References
[1] Alzaabin, Marc. “CISRI: a Crime
Investigation System Using the Relative
Importance of Information Spreads
in Network Depicting Criminals
Communications” published in
Information Forensics and Security, IEEE
Transaction, vol. (10), pp. 2196-2211,
2015.
[2] Almulhem, John and Iven Traore.
“Experience with Engineering a
Network Forensics System”, Lecture
C S I C o mm u n I C a t I o n S | n o V E m B ER 2 0 1 7
Notes in Computer Science, vol. 3391,
pp.62–71, Jan. 2005.
[3] Anchit, Bijalwan, and Wazid,
Mohammad, Emmmanuel S.Pilli.
“Forensics Of Random-UDP Flooding
Attacks”, published in Journal of
Networks, vol:10, 2015, pp.287-293.
[4] Cato Networks: Global Industry Report,
Top Networking and Security Challenges
in the Enterprise, November 2016
[5] Changwei, Liu, Anoop Singhal and
Duminda Wijesekera: A Logic Based
Network Forensics Model for Evidence
Analysis. IFIP Int. Conf. Digital
Forensics, 2015.
[6] Dhishan, Dhammearatchi. ‘Use
of Network Forensic Mechanisms
To Formulate Network Security’,
International Journal of Managing
Information , NoTechnology (IIJMIT)
Vol.7 No.4, November 2015
[7] Shrivastava, Gulshan and B. B. Gupta.
“An Encapsulated Approach of Forensic
Model for Digital Investigation.”,
Consumer Electronics (GCCE), 2014 IEEE
3rd Global Conference on, pp. 280-284.
IEEE, 2014.
[8] Nik, Mariza, Nik Abdull Malik, Saadiah
Yahya and Mohd Taufik Abdullah,
“Critical Phases in Network Forensic-
A Review”, International Conference On
Digital Security And Forensic, pp 68-75,
2014.
[9] Pilli, S.E., R. Joshi and R. Niyogi,
“Network Forensic Frameworks: Survey
And Research Challenges,” Digital
Investigation, pp. 14-27, 2010.
[10] Saad, Sam and Iven Traore, “Method
Ontology for Intelligent Network
Forensics Analysis”, Eighth Annual
International Conference on Privacy
Security and Trust (PST), pp. 7 – 14, 2010.
[11] Wei, Wang and Thomas E. Daniels,
”Network Forensics Analysis with
Evidence Graphs”, Digital Forensic
Research Workshop, 2005.
[12] M. Pollitt, “A History of Digital
Forensics,” in Advances in Digital
Forensics VI, vol. 337, K.-P. Chow and
S. Shenoi, Eds. Berlin, Heidelberg:
Springer Berlin Heidelberg, 2010, pp.
3–15.
[13] D. B. Parker, Crime by computer.
Scribner, 1976.
[14] E. Casey, Digital Evidence and Computer
Crime. Academic Press, 2004.
[15] P. Sommer, “The future for the policing
of cybercrime,” Computer Fraud &
Security, vol. 2004, no. 1, pp. 8–12, Jan.
2004.
[16] S. L. Garfinkel, “Digital forensics
research: The next 10 years,” Digital
Investigation, vol. 7, Supplement, pp.
S64 – S73, 2010.
[17] C. Stoll, “Stalking the wily hacker,”
36
Communications of the ACM, vol. 31,
no. 5, pp. 484–497, May 1988
[18] Kun Wang, Miao Du, Yanfei Sun, Alexey
Vinel, and Yan Zhang, “Attack Detection
and Distributed Forensics in Machine-
to-Machine Networks”, IEEE Network •
November/December 2016
[19] Javeria Ambareen, Pritam Gajkumar
Shah and M. Prabhakar, “A Survey
of Security in Internet of Things –
Importance and Solutions”, Indian
Journal of Science and Technology, Vol
9(45), December 2016
[20] Joseph, Jane and Peter, “A Proactive
Approach in Network Forensic
Investigation Process” International
Journal of Computer Applications
Technology and Research Vol 5– Issue
5, pp 304 - 311, 2016
[21] Wei, R, “Modeling the network forensics
behaviors” In Security and Privacy for
Emerging Areas in Communication
Networks. Workshop of the 1st
International Conference in 2005.
[22] Kohn, M. J, “Framework for a digital
forensic investigation”, Information
Security South Africa (ISSA). South
Africa: Insight to Foresight, 2006
[23] Stephenson, “A Comprehensive
Approach to Digital Incident
Investigation”, Information Security
Technical Report,E.A. Technology,
Editor pp 42-54, 2003
[24] Khatir, M. S, “Two-Dimensional
Evidence Reliability Amplification
Process Model for Digital Forensics”. In
Digital Forensics and Incident Analysis.
WDFIA ‘08. Third International Annual
Workshop in 2008.
[25] Yong-Dal, “New Digital Forensics
Investigation Procedure Model”. In
Networked Computing and Advanced
Information Management,.NCM - 2008.
[26] Rahayu Selamat, “Mapping Process
of Digital Forensic Investigation
Framework”. IJCSNS International
Journal of Computer Science and
Network Security, Vol.8(No. 10): pp 163-
169, 2008.
[27] Sherri Davidoff, Jonathan Ham,
“Network Forensics – tracking hackers
through cyberspace” Prentice Hall
publications ISBN-13 978-0-13-
256471-7, 2012
[28] Charles Pfleeger, Shari Lawrence,
“Security in Computing” fourth edition,
ISBN 978-81-317-2725-6, 2013
[29] Ankit Fadia, “Network Security, a
hackers perspective” second edition,
ISBN 1403-93088-0, 2009
[30] William Stallings, “Cryptography and
Network Security – Principles and
practice” sixth edition, ISBN 978-93-
325-1877-3, 2014
[31] IEEE 5G Summit Trivandrum – “IoT &
Cyber security” 2017.
www.csi-india.org
S E CU R IT Y C OR N ER

Report of ‘3rd IT Excellence Awards 2017’

by CSI Ghaziabad Chapter
Ghaziabad Chapter organized ‘3rd IT Excellence Awards 2017’ on 9th
September 2017 at Auditorium of Inderprastha Engineering College,
Sahibabad. IT Excellence Awards were presented in various categories to
individuals and organisations working in promotion of IT. This was the 3rd
edition of awards which was originally instituted for spread and awareness
of IT in Ghaziabad region but this year it covered nominations from Northern
India (CSI Region 1) as well. Some award recipients were marquee achievers
at National level.
Award recipients were:
 All India Institute of Medical Sciences (AIIMS) for patient experience
improvement in OPD
 Mr Rajesh Dogra (National Head of Passport Operations) for excellence
in Government-to-Citizen services
 Dr Anil Pandey & Mrs Usha Pandey – For dedicating their life to
wholesome education to slum children
 Purple Arrow Consultancy Services – Special Startup award for
innovation & Entrepreneurship
 Aptron Solutions – IT Training institute for designing relevant courses
for Engineering students
A panel discussion was also organized during the event on the topic – ‘Can
initiatives like Start-up India and Digital India generate employment by
creating more entrepreneurs?’
Panelists were Dr A K Chawla (HR Data Privacy Lead, TCS and Former HR Head for North India, TCS), Ms Sarita Chauhan (Business Power
Coach) and Prod Manoj K Pandey (Director, SRM University NCR Campus). Discussion was quite interesting which was excellently moderated
by Ms Charu Garg (Former anchor, Zee News).
CSI awards were also presented to Student members, member Engineering Colleges and academicians. Prof A K Nayak (Hony National
Secretary, CSI) was the Chief Guest on the occasion. He emphasized the importance of Digital India and how it is critical for today’s society.
Mr Saurabh Agrawal (Patron CSI and Chair – Awards Jury) briefed on the essence of IT Excellence Awards and how they are being recognized
at the regional level.
Er. Anil JI Garg (Chairman CSI Ghaziabad Chapter) briefed about the growth and evolution of Ghaziabad chapter over last 13 years and how
it has been the best performing chapter in the entire region. Chapter’s performance can be gauged from the fact that out of 12 student
branches in the state of U.P., 10 are in Ghaziabad alone. Dr Kavita Saxena (Secretary, CSI Ghaziabad Chapter) presented her annual report on
chapter activities for 2016-17. Dr B C Sharma (Director, IPEC) and Shri Piyush Goyal (Founder Secretary, CSI Ghaziabad Chapter & DGM SAIL)
also spoke on the occasion and shared his thoughts with the delegates and students. Also present were former RVP(1), Shri Shiv Kumar and
UP SSC, Dr Sunil K Pandey.
Next morning on 10th September 2017, Prof Nayak had a breakfact think-tank meeting with the chapter patrons and OBs. Hony Secretary
congratulated Ghaziabad chapter on the excellent show and on its continuous good work in the region. Meeting generated several ideas and
Prof Nayak promised to present them in the upcoming National Executive Committee meeting.

About the Authors

Mr. A. R. Jayakrishnan [Member No: I1503330] is currently pursuing PhD in Cyber Forensics from Bharathiar
University. He pursued MCA from MG University in 2001, PGDBA from Pondicherry University in 2005 and has
industry experience of 16 years in IT companies in India and UK. His area of interests are Network Forensics,
Information Security, IoT Forensics, Honeypots. He can be reached at anilakkad.jk@gmail.com

Dr. (Mrs.) v. vasanthi pursued M.Sc(CS), M.Phil, and Ph.D Computer Science from Karpagam University,
Coimbatore in 2014. She is currently working as a Professor in Department of Computer Science at Rathinam
College of Arts and Science, Coimbatore since 2015. She has published many research papers in reputed
international journals including Thomson Reuters (SCI & impact factor) and conferences including IEEE &
Springer. Her main research work focuses on Adhoc and Sensor Networking. She can be reached at vasanthi.
cs@rathinam.in

37
C S I C o mm u n I C a t I o n S | n o V E m B ER 2 0 1 7
 A REPORT
TM
Computer Society of India
CSI Young IT Professional (YITP) Awards 2017
Final Round at Nashik on 7th Oct. 2017

Back row – SB Bagal, Principal GN Sapkal college of Engg., person as mentioned above.
Manas Ranjan Pattnaik CSI treasurer, Shri.Gautam Mahapatra  Motivate upcoming professionals and enhance the
CSI Vice President, Prof. A K Nayak Hon Secretary CSI, prestige of the profession by enabling them to exercise
Sanjay Mohapatra CSI President, Ravindra Sapkal CMD their skills and competencies in the best possible and
Sapkal Knowledge Hub, Prof. E. Vayunandan VC YCMOU, innovative manner
Nitin Mahajan GM BSNL, Dr. Anirban Basu CSI IPP, Jury Mr.  Directly or indirectly contribute towards the improvement
Hussain Dahodwala, Shri SR Karode, Dr. RS Tiwari of professional effectiveness in different spheres of life
 Motivate young men and women to imbibe the
Front row – Prabhakar & team, Shreyasi Datta winner, Dr.
entrepreneurial spirit and thereby encourage more and
SS Sane RVP region VI & YITP convener, Jury Rohit Kulkarni,
more people to venture into entrepreneurship
Sandeep Karkhanis Chairman CSI Nashik Chapter
 Encourage contribution in boosting the economy of the
CSI had announced “CSI Young IT Professional (YITP) country and promote research
Awards” competition for Individual IT Professionals and the IT
The Criteria:
Professional Teams below 35 Years Age, with objectives to:
The most outstanding technology project related to any
 Recognize the significant contribution(s) made by a
theme mentioned below & completed during the year 2016/
The regional winners:
Region Competition Winner Winner’s Project Runner-up Runner’s Project
Location
I Lucknow Mr. Ashutosh Kumar Route Finder and Communicator Mr. Manish Kumar Design and Development of a
for Blind, Deaf and Dumb Web Crawler for Hidden Web
II Kolkata Ms. Shreyasi Datta Detection of AF and Other Mr. Purnendu Debnath Smart BIO Secure Server Login
Arrhythmia Using Single Lead
ECG
III Vadodara Ms. (Dr.) Deepshikha Fire Detection in Visible Spectral Mr. Rushi Trivedi Modem Trace Tool & Automation
(Baroda) Tiwari Range of Carrier Acceptance Testing
IV Bhubaneswar Mr. Sanjaya Kumar Dynamic Data Normalization in Mr. Santosh Kumar Online Query Tracking System
Panda Cloud Computing Satapathy
V Bengaluru Mr. Shanmuga Vehicle Clutch Health Ms. Shresta Anatha Xcelerator - Education, Project
prognostics and Vehicle Fuel and Career Guidance Portal
Consumption calculation
VI Pune Mr. Prabhakar My Robot – Personal Assistant Mr. Sudhir Gorade, & Bandhan / MahaBandhan – A
Gadupuddi Team Self-Help-Group of Women
VII Kovai / Mr. S. Sathish “Vibo” - An Artificial Mr. V. Gokula An Intelligent Smart Energy
Kovanputhur Kumar, & Team Intelligence based Machine Krishanan, & Team Saver
(Coimbatore) Learning ROBOT for Assisting
Physiotherapists

www.csi-india.org
38
C S I C o mm u n I C a t I o n S | n o V E m B ER 2 0 1 7
 A REPORT
2017 (project duration could be of 2-3 years from the start
date) within an organization were judged for following criteria
by the selection committee:
Criticality of IT Usage, Improvement of Customer Service,
Innovation, Return on Investment, Quality of Management,
Impact on Organization and/or Society & Presentation Skills
The completion was carried first at each of the 7 Regions
of CSI at the following Centers where the Winners and Runners
up of the regional contest from each region were selected for
the Final Round Competition at Nashik.
The winners of the regional contest were qualified for
competing in the final national round.
Each Participant / Team Leader at the Final Round
was given 30 Minutes for the presentation, followed by the
questions by the Juries & the Audience.
However, all 14 Regional Winning Teams, 2 from each of
the 7 Regions, could not appear for the Final Round due to
various reasons.
The 4 Juries appointed for the evaluation of the Final Round
Teams were eminent & very well experienced IT Professionals
from Nashik who cross-questioned the contestants to really
judge them and bring-out on the front their IT Skills.
The Jury team comprising of 1] Shri. S. R. Karode, 2]
Dr. Anand Tiwari, 3] Mr. Rohit Kulkarni and 4] Mr. Hussain
Dahodwala were keen to understand from the participants
about their contributions.
The Program started off with introduction speeches, and the
inauguration of the function by performing puja of goddess
of knowledge, Ma Saraswati Devi, followed by the dignitaries
on the dais lighting the traditional oil lamp, symbolizing the
removal of the darkness of ignorance and lighting-up the
knowledge.
The competition was rightly set for -
Category Final Round -
National Awards (`)
Winner 25,000/-
1st Runner-up 15,000/-
Special mention 10,000/-
All teams presented their creative and innovative solutions
and product developed by them for their organisations. This
difficult task of selecting the final Winner, the 1st Runner-up
and the Special mention was ably performed by the jury team.
The winner Ms. Shreyasi Datta on her
project - Detection of Atrial Fibrillation
(AF) and Other Arrhythmia Using Single
Lead Electro-Cardio-Gram (ECG).
The subject itself was highly technical,
combining the technologies of Sensor,
Electronic Signal Processing, Medical
Electronics / Diagnosis, Machine Learning, and Intelligent
Algorithms & Validation Schemes etc.
The motivation behind this project is to enable continuous
monitoring of cardiac conditions of the elderly through a
simple, easy to use, smart-phone based attachment by
constant analysis of ECG data by doctors remotely, which is
39
C S I C o mm u n I C a t I o n S | n o V E m B ER 2 0 1 7
well achieved in this research. The innovative Sensing is the
Low-Cost “Single Lead” but at the same time, recognizing and
accepting that “Every irregularity in the pattern is NOT noise!!”
Therefore, the Noise in the Signal cleaning (removing) and the
Marker Detection utilized the Spectrogram Method of Digital
Signal Processing which is analyzing the Signal Frequency
Components in Hz with respect to time in Seconds (Time
Domain) for the Frequency band of 0 to 150 Hz (higher power
at higher frequency means noisy window, to be rejected) in the
time frame of 0 to 30 Seconds. Traditional systems generally
record a longer duration of ECG (more than 1 Hour) using
12-Leads Medical-Grade Device.
The automated feature extraction techniques made the
Invisible features, visible to the doctors!
More than 200 ECG Features can be extracted which are
domain specific which are impossible for the doctors to do.
So, the Computers catch them fast and accurately.
The Arrhythmia Screening Process Block Diagram showed a
Patient / Person using the Single-Lead Sensor for his left-hand
& right-hand fingers. The Sensor provided an Ultra-Sound
Electro-Acoustic Coupling to a side-by Mobile Microphone.
The ECG acquired in the Mobile Memory is transmitted to a
Server (Cloud). The Signal Processing done on the Server are
Noise Cleaning from the ECG waveform, Feature Extraction,
Classification based on the Inputs from Pre-trained Models.
This Project Team has obtained a Patent for this project. The
Team won an International Competition of the Physionet
Challenge in 2017. The Team has excellent support from the
medical doctors.
Overall, involvement, enthusiasm and the devotion noted
on this project was the highest and highly technical and
thoroughly professional. This IT Professional would continue
to demonstrate even brighter results in time to come.
The second prize was bagged by team lead by Ms. Shresta
Anatha on their Team’s development & Implementation of the
Education, Project and Career Guidance Portal (Excelerator).
She highlighted the gap between the Industry Requirements
and the Academic Output, addressing the serious problems
of large number of Un-employable professionals with un-
adequate skills which are demanded by the Industry.
The Industry needs, apart from the basic knowledge & the
domain knowledge, soft-skills, ability to analyze the problems,
and the problem solving approaches. She stated that formal
learning constitutes only 10% where as 20% is the informal
learning and the most significant part, that is 70% is by way of
the on-the-job experience.
So, the 3 Buckets of making the potential candidate employable
are: Education, Exposure to Industry and Hands-on Experience
from 2 Weeks up to the required duration.
To address the situation, her team cleverly came-up with the
Solution, “Xcelerator” as a Conceptual Architecture which
holistically considers the Recommendations, the Skills
Framework, and helps in building the Candidate’s Digital
Portfolio suitable for Recruitment.
 A REPORT
A Recruitment Dashboard on the Portal provides for Job
Description, going in details of Use Case, Recommendations,
Clarity on the Job Role, Skills Frames covering all aspects of
Education, Exposure and Experience. The Portal Technology
Stack was explained in terms of layered & robust architecture
in terms of Top Layer as Presentation Layer utilizing the
ReactJS & NginX (both acting as the web server) and Sub-
Layers as Navigational Feeds, the Application Layer based
on Phoenix Web Development Framework written in Elixir,
deriving it’s strength from Erlang (concurrent, fault-tolerant
and real-time system – A Virtual Machine); the Data Layer of
Redis (mainly used for caching and to manage job queues) &
Postgres (it runs Relational DB and NoSql DB in the same
server); and Infrastructural Layer of Docker (to manage the
whole application stack by containers being run), AWS (Cloud
Server).
The Candidate’s Attitudinal Problems too are attending
including a Blogger Forum.
The Return on Investments were discussed on 4 Revenue
Streams of Workshops, platform subscriptions, mentoring and
guidance fees, recruitment service fees, affiliate agreements,
etc.
Finally, Impact on Society was presented in detail as the Team
want to give students pursuing a degree, a sense of purpose,
to help them choose a goal, make a plan to achieve the goal,
and help them stick to the plan to achieve their goal. The
platform empowers students to articulate and pursue their
aspirations and ambitions. The Teams’ Future Plans as
explained include expansion to an even wider college base
– more students on the platform will help them make their
recommendation algorithms more robust.
A similar effort on the industry side to expand the number
of companies on the platform is also underway. Their URL:
https://www.xcelerator.ninja
The special mention prize was awarded to Mr. Prabhakar
Gadupuddi and his Team on the Project - My Robot – Personal
Assistant.
The Project summary explained “My Robot – Personal
Assistant” as a mobile application and end user(s) can talk to
their mobile phones using natural language and assign the
task for automation. Users can talk to their mobile phone and
mobile phones will control/ monitor user’s daily activities,
meetings, reminders, personal computers and IT assets,
virtual machines and etc.
In this application, artificial intelligence module is enabled.
Mobile phones virtual think and respond to users for the
task completion. All devices (computers, servers, mobiles)
are inter connected for sharing the device status and info for
monitoring/ controlling.
40
C S I C o mm u n I C a t I o n S | n o V E m B ER 2 0 1 7
Scope of the solution as presented has the high level features
as Personal assistant, IT Assets management and Device
management.
This project realized well their goal of providing a single
application which automates IT assets management, Personal
assistant, Device management as well as it helps individuals
to simplify their daily routine tasks.
Features/ Functionality successfully lively demonstrated
were Voice enabled features for E-mail integration, Device(s)
management, Artificial intelligence enabled Chat-bot
integration, IT Assets/ compliance management, and Personal
assistance.
The Project’s Software and hardware requirements for 25 users
are, Cloud Server (Hard Disk 80 GB, RAM 2 GB); Software:
Android V 6.0+, Python required libraries Version 2.7+, MySQL
DB v5.0+, TensorFlow -, Google Speech recognizer – Natural
Language Tool Kit (NLTK); Mobile application – Android
application V6.0+
The Development tools and technologies utilized were, Front
end – Android layout model (XML); Back end – MySQL DB,
Rest APIs and the Development environment – Android Studio
/ Eclipse environment (Java)
The System Implantation Level Block Diagram showed the
Mobile as a Voice Input & Output Device which sends digital
information to the Gateway which includes IMEI Operation and
Model Token. The Services APIs are Loader, Main, Audit, Device,
Services, Chatbot, Remind. The Request generated is filtered
and bundled. The cloud Storage provides the databases.
This project showed effectively how “Time is Money!!” and
how this application reduces the Total Cycle Time. A Physical
Method which would take say about 14 Minutes, the “My Robot”
can do the same in just 2 Seconds.
On the whole, the entire presentation was most attentively
watched by the audience and for the Non-IT personnel; it
would have given the feel of a Sci-Fi Movie. The Juries queries
were most successfully fielded by the presenters.
It was enjoying moments as to watch and experience what
today’s Information & Communication Technologies can do for
the mankind.
The event got overwhelming response from IT professionals
from Nashik, Academia, Students and was appreciated
by participants, guest and executive committee of CSI for
arrangements done by Nashik Chapter & Sapkal Knowledge
Hub under guidance of Prof. Shirish Sane, RVP region VI and
YITP 2017 convener.
n
www.csi-india.org
Pilgrims in the Digital World
By Sandeep godbole
Reviewed by Sunil Bakshi, CISA, CISM, CGEIT, CRISC, AMIIB,
BS 25999 LI, CEH, CISSP, ISO 27001 LA, MCA, PMP, Visiting
Faculty and Industry expert at the National Institute of
Banking Management (NIBM), consultant and trainer in IT
governance and information security. Sunil has served as
President at CSI Pune and is a life member.
‘Tomorrow is not what we were told it would be’ goes an
English adage and the world is busy experiencing it each
day. Proliferation of information and digital technology
and at a blistering pace has left users bewildered. The
change - though unexpected and largely organic, has found
acceptance in the global community. The role of technology
is central to the changes that have permeated multiple
spheres. It is important that the role of technology the
changes and its implications be discussed for the benefit of
the common users. Sandeep Godbole’s book has achieved
this with elegance, simplicity and ease.
The title of the book is apt to the theme of the book. The
book describes the journey, the digital transformation and
presents what the future could possibly hold for us. Robert
E Stroud (past President, ISACA) when discussing this book
41
C S I C o mm u n I C a t I o n S | n o V E m B ER 2 0 1 7
Book Review
has stated that we are only at the beginning of this journey
and tomorrow may be different than what we may imagine it
to be. This is a book on digital technology for users who may
not want to dive into the technical details.
The book is organized into four parts, each representing a
theme relevant to the digital world. Part one aptly named
as ‘Sunrise’ introduces digital world and “town square” of
our era – the Internet. This part also discusses the change
that has been brought about in India by harnessing the
power of digital technology. It makes a case for the need
for digital technology given the country’s demographics
and aspirations. The chapter on data explosion and ‘Big
Data’ reveals the importance of data for analytics and our
own progress. The chapter on the application of the digital
technology provides us glimpses into the digital aspects
related to health and education.
The second part is focused on the Issues that arise in the
digital world. IT awakens us to the reality that though digital
technology is immensely beneficial, there are aspects
that are a serious concern. Complex issues are explained
in simple terms by the author. The issues discussed are
freedom in digital world, data privacy, and the intrusion and
encroachment of personal space that can even impact the
safety and wellbeing of individuals.
The third part talks about the new threats and dangers
that afflict the digital world. Cyber security has become an
essential to the digital world. Technology is a double edged
sword, it can bring us immense benefits but can also be a
vehicle to inflict serious damage. Technology in the hands of
criminals, enemy nations and adversaries can be exploited
by them to cause damage. This part discussed threats like
cyber war, cybercrime, hacks malware enabled attacks like
advanced persistent threats, Botnets etc.
Last part provides the means and response to the threats
that are discussed in previous part. This part explains
security techniques and approaches in simple terms. One
of the chapters is devoted to encryption, which is perceived
as complex and technical. Sandeep has made is simple and
explained the technical concepts in a language that readers
can understand. This part also discussed Governance of IT
and how it can help in creating value to the society.
The reader is treated to an elegantly written and logically
composed book. It stands testimony to Sandeep’s writing
skills. The quotes, references and data provided by Sandeep
to argue and put forth his views is indicative of a well-
researched writing. Edward teller, the father of hydrogen
bomb has once said that “Science today is technology
tomorrow”. Yesterday’s Computer science is Digital
technology of the day and we need to master its use, while
being cognizant of the benefits and limitations. Sandeep’s
book takes you on an interesting pilgrimage that leaves you
better informed and enriched.
F ROM C h AP TERS & DI v I S I O NS

COIMBATORE ChAPTER Business at Dr Damodaran College of Science on 6th October

2017. The audience of the seminar entitled “Disrupting
Digital Business” were enlighten by the knowledge shared
by the resource person Mr. S. Ravi Shankar, Chief Technical
Officer, Kovai Medical Centre and Hospital, Coimbatore.
He started the session with an introduction to digital
transformation and insisted the need of accepting digital
transformation in this technology arena for a successful and
consistent business. He also discussed about the evolution
of hardware and software technology which paved way
for a clean and groomed business environment. This also
The Chapter organized Young IT Professional (YITP) Contest supports in improving the management and accessibility
2017 - Regional Round for Region VII on 17th September 2017. of data. The environment of the current generation is
Dr R Nadarajan, Professor & Head, Dept of Mathematics & highly surrounded by the digital media. Online shopping
Computer Applications, P S G College of Tech., Coimbatore, websites have reduced the traditional purchase, affecting
Dr. N K Karthikeyan, Coimbatore Institute of Technology, a family outing and social communication. He pointed out
and Mr. S Manoharan, Head IT, the drivers driving those disruptions as globalization, data
Elgi Ultra Industries Limited, Coimbatore were the selection explosion, mobility, rise of social business, etc. He listed out
committee members. Out of 13 papers received, the team and discussed the key areas to govern the future technology.
selected 3 papers and called for presentation. The Winners/ He insisted students to equip themselves with any of the
Runners are as listed below industry requirement and update their knowledge base to
shine in their future endeavors.
Winner
Mindnotix Technologies, Dr. S Balamurugan hARIDwAR ChAPTER
Coimbatore Mr. S SathishKumar
Mr. R Deepak Engineers day was celebrated by CSI Haridwar Chapter on
15th September 2017. This very special day dedicated to all
Runner Mr. S Kaviarasan engineers, Online quiz competition for students by students
Panimalar Inst of Tech, Mr. A IrudayaPaulraj on students created platform. In this competition students
Vinod Mr. V Gokulakrishnan took part and Mr Japneet Singh won the award of ` 1,000/-.
Special mention Ms. Ramyaa R In continuation, an expert lecture on Big Data and Hadoop
Sri Ramakrishna Engg College, Ms. Sivaranjani N was delivered by Mr Mohit Kaushik, who worked for Oracle
Coimbatore Mr. N Deepak and Mr. Vivek Chaudhary, Director, Arise Data. This event
was organized by Mr Suyash Bhardwaj.
Mr. Sivabalan, Vice Chairman announced the results.
Trophy, Certificate and cash award winner ` 5,000/-, Runner NAShIK ChAPTER
` 3,000/- Special Mention ` 2,000/- Dr. R Nadarajan
distributed trophy and certificate to winners and Dr. M
Sundaresan, RVP-VII distributed trophy and certificate to
runners. Mr. N Valliappan distributed trophy and certificate
to special mention and Dr. G Radhamani, Hon. Secretary
proposed the Vote of Thanks.

CSI Nashik organized a seminar on 64-bit processor

The Chapter and Dr. Damodaran College of Science,
Coimbatore organized a seminar on Disrupting Digital
technology from AMD on 22nd September 2017. AMD is
one of the leading x86 processor & graphics organization in
the world today. As the originator of 64-bit x86 technology,
integrated memory controller, integrated graphics and
compute cores (APU) AMD leads the world in cutting-edge
technology globally. The purpose of the seminar was to
create awareness amongst SMB, Education sector across
Nashik periphery. The presenters were Shri Ajay Koul,
Regional Sales Manager(West), Enterprise Division, Shri

www.csi-india.org
42
C S I C o mm u n I C a t I o n S | n o V E m B ER 2 0 1 7
F ROM C h AP TERS & DI v I S I O NS
Manmit Desai, Business Development Manager (India),
Professional Graphics for AMD. Ajay presented on AMD
processors, which mainly include, AMD Godavari, AMD
Kaveri, AMD Bristol Ridge, & New Processor Ryzen, Server
platform EPYC. He explained AMD presence and stake
in India as well as global. Manmit explained about AMD
graphics. Bahubali movie on AMD pro graphics. Radeon
Pro Graphics and applicable industry. Testimony given
by M/s Minitek and AMD core architecture was explained
by Shri. Jaykrushna Puranik committee member and
experienced in microprocessor based controllers and
systems development. Session was interactive and followed
by technical Q and A. The participants expressed that the
session was indeed a nice and informative one.
RAIPUR ChAPTER
GeekExpo, a project exhibition was held at SSIPMT Raipur on
5th October 2017 in association with the Chapter and Center
of Excellence of CSE & IT. In which 22 groups presented
their Projects / Models / Ideas based on IoT / Webapps /
Mobile Apps. Judges are Mr Nishant Tripathi, Chairman, BG
SSIPMT, Dr Alok Jain, Principal SSIPMT, Dr J P Patra, Hon
Secretary, Raipur Chapter, Prof Riju Bhattacharya (CSE),
Prof Gurudatta Verma (CSE). They appreciated the efforts of
students and one participant got paid internship.
TIRUChIRAPPAllI ChAPTER
The Department of CS and Engg. has organized a Two
Days AICTE Sponsored National Seminar on Multimedia
Information Retrieval- Research Issues on 25th and 26th
of September 2017 in association with CSI Tiruchirappalli
Chapter. As Multimedia Information Retrieval has become
one of the most active sub-fields of multimedia research, this
two days seminar has provided a platform for researchers
and academicians to network and exchange different
techniques to retrieve multimedia data and focused various
topics such as Multimedia Information Retrieval, Basic
Concepts and Issues, Multimedia Indexing Research Issues
and Tools, Social Media Information Retrieval, Information
43
C S I C o mm u n I C a t I o n S | n o V E m B ER 2 0 1 7
Retrieval using Opinion Mining, Machine Learning and
Data Analytics in Multimedia IR and Information Retrieval
– Research Perspective. There were 113 participants who
have registered for the seminar, out of which 54 members
were from the various colleges in Tamil Nadu. The speakers
were leading researchers and academicians in the field of
Information Retrieval from top engineering institutions and
industry. The session speakers were Dr S Swaminathan,
Anna University, Chennai, Dr Janet, NIT, Tiruchirappalli, Prof
J Naveen, Christ University, Bangalore, Dr E Kirubhakaran
, BHEL, Tiruchirappalli, Dr U Srinivasalu Reddy, Assistant
Professor and Associate Dean, NIT, Tiruchirappalli and Dr
S Nicolas, NIT, Tiruchirappalli. Dr S Ravimaran, Principal
& Chairman of CSI Tirchirappalli chapter rendered the
welcome address and explain the objectives of this two days
seminar to the participants. He congratulated the faculty
members for their support and stated how the two-days
seminar will be helpful for the academicians and research
scholars.
MAM College of Engineering in association with CSI
Coimbatore Chapter organized TRIGGER 2017 - 7th District
Level Inter School Technical Competition – TRIGGER 2017
on 14th October 2017. About 532 students from 26 various
schools participated. Initially, the events started off with a
colorful and grand inaugural ceremony headed by Director,
M.A.M. College of Engineering Prof. U.S. Saravadayaparan
and Dr S.Ravimaran, Principal of the college and
Immediate Past Chairman of CSI Tiruchirappalli Chapter. A
lot of enthusiasm was seen in all the student participants.
Director and Principal of the college distributed the prizes
and trophies to the winner and runners of various events.
vADODARA ChAPTER
The Chapter hosted the Regional Round of Young IT
Professional (YITP) Award 2017 for CSI Region III. The event
was scheduled on Sunday, 24th September 2017. Followed
by the event, there was a Meeting with Dr. Vipin Tyagi,
Regional Vice President, Region-III the MC Members of CSI,
F ROM C h AP TERS & DI v I S I O NS
Vadodara Chapter, Chairpersons of the Chapters under the
purview of CSI Region III and Ex-Chairmen of CSI, Vadodara
Chapter
vEllORE ChAPTER
The Chapter in association with Apple Developer Group VIT
Student Branch organized a workshop on Virtual Reality and
Augmented on 16th & 17th Sep 2017 at VIT University. Mr
A. Mani, General Manager, Wipro Technologies, Bangalore
explained the basics of augmented reality and virtual reality
with the help of tool called “Oculus Rift”. The speaker
explained the different features and functions of the tool
to develop application in real life scenario. Around 200
members participated in workshop, organized by Dr. R
Rajkumar, Prof. K Govinda.
FR OM S T U D E N T B R A N C h ES
REgION-I
Maharaja Agrasen Institute of Technology, Delhi
31-8-2017 - CSI Orientation programme
44
C S I C o mm u n I C a t I o n S | n o V E m B ER 2 0 1 7
The Chapter organized its fifth edition of Laser Tag. It was
a three days event, starting on 22nd September to 24th
September 2017 at VIT University. The arena was dimly
lit and was set up with neon obstacles and UV lights. The
participants were divided into teams of four and were
provided with laser guns and vests. The objective of the
game was to the tag the vest of the opponent player with
the laser beam to score points. The equipment for the
game was provided by Talk to the Hand- a Bangalore based
company. It was a very interactive game which required well
thought through tactics and teamwork. Around 300 students
participated in the event and organized by Dr. R. Rajkumar,
Prof. K Govinda and Prof. Sendhil.
Jaypee University of Information Technology, Solan
15-9-2017 - workshop on Massive Open Online Courses
(MOOCS) by Prof Mangala Sunder Krishnan, IIT Madras
www.csi-india.org
FR OM S T U D E N T B R A N C h ES

REgION-I
Dronacharya College of Engineering, gurgaon

12-10-2017 - Guest Lecture on Information Security by 13-10-2017 - Workshop on CISCO Networking

Mr. Narasimha Rao, C-DAC Hyderabad by Mr. Siddharth Tiwari
REgION-III
Sagar Institute of Science and Technology(SISTec), Bhopal

27-9-2017 - Expert Lecture on Data Warehousing and 7-10-2017 - Expert Lecture on C/C++ : An Expert’s View
Business Intelligence by Mr. Tarang Gupta
Jaypee University of Engineering & Technology, guna

23-9-2017 & 24-9-2017 - Workshop on C Programming 27-9-2017 – Event on KODEATHON

g h Patel College of Engineering & Technology, vallabh vidyanagar

20-9-2017 & 21-9-2017 - Event on QRIOSITY’ 17 23-9-2017 Workshop on Introduction to Python Programming
- A quizzing contest

45
C S I C o mm u n I C a t I o n S | n o V E m B ER 2 0 1 7
FR OM S T U D E N T B R A N C h ES

REgION-III
The lNM Institute of Information Technology, Jaipur

4-8-2017 - Introduction of CSI to the freshers & short quiz 4-9-2017 - SIP (Summer Internship Project)

REgION-v
Scient Institute of Technology, hyderabad

16-9-2017 - Workshop on Mathematical Applications 9 -10-2017 to 14-10-2017 - Workshop on Android Hackathon

CMR Technical Campus, hyderabad

29-9-2017 - Guest Lecture on Education Counselling 5-10-2017 –Seminar on Full Stack Development by Mr. Rohit
by Mr. Amarnath Goud

NBKR Institute of Science and Technology, Nellore

5-10-2017 to 6-10-2017 - Workshop on 10-10-2017 - Training Program on Website Designing

Cyber Security and Malware Analysis

www.csi-india.org
46
C S I C o mm u n I C a t I o n S | n o V E m B ER 2 0 1 7
FR OM S T U D E N T B R A N C h ES

REgION-v
geethanjali Institute of Science & Technology, Nellore vasavi College of Engineering (Autonomous), hyderabad

15-9-2017 to 17-9-2017 - Workshop on 9-9-2017 – Contest on Connect It

Google Android Developer Workshop
Anurag group of Institutions, hyderabad

15-9-2017 to 17-9-2017 - Short Term Training program on 12-10-2017 - Orientation Program about CSI by Mr Mohan
Ruby on Rails by Mr Pavan Manne Raidu, Chairman, CSI Hyderabad Chapter
REvA UNIvERSITY, Bengaluru

1-9-2017 – Event on Best Idea Competition 7-10-2017 - Inauguration of CSI Student Branch
by Dr. Anirban Basu, IPP, CSI
St Joseph Engineering College, Mangaluru JSS Academy of Technical Education, Bangalore

18-9-2017 - CSI Student Branch Inauguration 23-9-2017 - Technical talk on Information and storage issues in
Modern Intelligent environments by Mr Hanu Ramasanjeeva

47
C S I C o mm u n I C a t I o n S | n o V E m B ER 2 0 1 7
FR OM S T U D E N T B R A N C h ES

REgION-v
K S Institute of Technology, Bangalore

28-8-2017 - Seminar on Overseas Education 11-9-2012 to 12-9--2017 – Hands-on workshop DBMS

gSSS Institute of Engineering & Technology Maharaja Institute of Technology, Mandya
for women, Mysore

22-9-2017 National Level Coding Competition : 16-9-2017 - Technical Talk on Cloud Computing and Mobile
CODE IGNITER – 2017 Computing by Mr Manu Kamath
REgION-vI
Yeshwantrao Chavan College of Engineering, Nagpur Pillai hOC College of Engineering and Technology, Rasayani

16-8-2017 & 17-8-2017 - Two Days National Level Technical 16-9-2017 - Workshop on Web Application Development
Fest on iTECHROOTS 8.0 Using MVC & SERVLET
late g. N. Sapkal College of Engineering, Nashik

25-9-2017 - Seminar on CCNA Certification 26-9-2017 - Workshop on Project to Entrepreneurs

www.csi-india.org
48
C S I C o mm u n I C a t I o n S | n o V E m B ER 2 0 1 7
 FR OM S T U D E N T B R A N C h ES

REgION-vII
Panimalar Institute of Technology, Chennai

23-9-2017 to 7-10-2017 – Training Program on SQL 2016 and C# 24-9-2017 - Runner up team with the organizers in the CSI
.NET by Mr Joel S Davis Young IT Professional (YITP) Award 2017
Rajalakshmi Engineering College (Autonomous), Chennai

11-9-2017 - Workshop on Web Application Development 13-9-2017 - Seminar on Bootstrapping in Mobile Application
Using Hibernate with Spring MVC by Mrs Banu Manoharan Development by Mr Ashwin Irudayarajan
Nehru Institute of Engineering and Technology, Coimbatore Avinashilingam Institute for home Science and
higher Education for women, Coimbatore

21-9-2017 – Dr. Sundaresan, RVP-VII, 22-9-2017 - Workshop on Arduino Controller and its
CSI Inaugurated CSI Student Branch Applications by Mr. Madhan Kumar
Toc h Institute of Science & Technology, Ernakulam vIT University, vellore

18-8-2017 – Dr Rajesh Maliyeckal, Past Chairman, CSI Cochin 30-9-2017 - Workshop on coding “Send CodeZ”
Chapter Inaugurated CSI Student Branch

49
C S I C o mm u n I C a t I o n S | n o V E m B ER 2 0 1 7
 FR OM S T U D E N T B R A N C hES

REgION-vII
Jeppiaar Institute of Technology, Sriperumpudur

13-9-2017 - Guest Lecture on 15-9-2017 - Guest Lecture on Theory of Computation

Computer Graphics and Multimedia by Dr. Aju by Dr. Anand Mahendran

CSI CA l EN D AR 2 0 1 7 - 1 8

gautam Mahapatra, Vice President, CSI, Email: vp@csi-india.org

Date Event Details & Contact Information

NOvEMBER State Level Student Convention for Andhra Pradesh organized by CSI Student Branch of GITAM Inst. of Tech., GITAM
24-25, 2017 University, Visakhapatnam. Contact : Prof. R Sireesha, 9848503365, sireesha.rodda@gitam.edu

DECEMBER Regional Level Student Convention for Region-VI organized by CSI Student Branch of Bansilal Ramnath Agarawal
19-20, 2017 Cheritable Trust’s Vishwakarma Institute of Information Technology, Pune. Contact : Prof Snehal Rathi, 9850604785,
snehal.rathi@viit.ac.in

21-23, 2017 State Level Student Convention for Tamil Nadu by Karpagam College of Engineering, Coimbatore. Contact : Dr. S. Angel
Latha Mary, 9842242882, angellathamary.s@kce.ac.in

28-29, 2017 Regional Level Student Convention for Region-V “Malwares in Social Media” organized by CSI Student Branch of
NBKR Institute of Science and Technology (Autonomous), Nellore. Contact : Dr. S. Maruthuperumal, 9848772088,
maruthumail@gmail.com

2018 Regional Level Student Convention for Region-VII organized by CSI Student Branch of Jeppiaar Engineering College,
JANUARY Chennai. Contact : Dr P Jesu Jayarin, 9444053102, jjayarin@gmail.com
23-24, 2018

24-25, 2018 State Level Student Convention for Maharashtra State organized by CSI Student Branch of St.Vincent Pallotti College of
Engineering & Technology, Nagpur. Contact : Prof. Abhishek D. Pathak, 9890919330, abhishekpathak81@rediffmail.com

FEBRUARY State Level Student Convention for Gujarat State organized by CSI Student Branch of G H Patel College of Engineering &
09, 2018 Technology, Vallabh Vidyanagar. Contact : Dr Moulika S Patel, 9428488563, maulikapatel@gcet.ac.in

www.csi-india.org
50
C S I C o mm u n I C a t I o n S | n o V E m B ER 2 0 1 7
 52nd Annual Convention of Computer Society of India
In Technical Collaboration with
(CSI 2017)
host: Kolkata Chapter | venue: Science City
19-21 January, 2018
On enormous request the last Date of Submission has been extended up to 15th November 2017
Theme: Social Transformation - Digital way
Proceedings will be published by Springer Nature Singapore in CCIS Series indexed by SCImago, Scopus, DBlP
52nd Annual Convention of Computer Society of India (CSI 2017) will be held at Science City, Kolkata during 19th January to
21st January 2018. The CSI Annual Conventions are held in different cities across India. The CSI Annual Conventions, usually
attracting 2000 plus participants have been held since 1965. Apart from technical sessions, tutorials, panel discussions and
Exhibition are main features of the convention. The Conventions held at Kolkata include CSI 66 (First Convention), CSI 78, CSI
86, CSI 90, CSI 94, CSI 2001, CSI 2006, CSI 2012.
The theme of the CSI 2017 is “Social Transformation - Digital Way”. It will cover all aspects of digital technology and machine learning in
computer science and engineering, general science, educational research and management science. This convention will provide a platform
to the participants to share their views and ideas on latest technological developments in an inter - and intradisciplinary perspective. There
are total eight tracks in the call for papers. Tracks of the international Conference are (but not limited to): Signal Processing, Microwave and
Communication Engineering; Electronic Devices, Circuits and Systems; Data Science and Data Analytics; Security and Forensics; Digital
Mapping in Environment; Bio Computing and Biomedical Engineering; Mobile, Nano and Quantum Computing; e-Governance, e-Learning
and confidence based Learning.
Participations in the 52nd Annual Convention of Computer Society of India (CSI 2017) are solicited from Industry Houses, Acad emicians,
Corporate Houses, Exhibitors, Faculty Members, Researchers and Students.

Submission guidelines:
The publication for the International Conference will be in SCI Mago, Scopus DBLP Indexed CCIS Series of Springer (Confirmed) and the
article should be within 15 pages in the Springer CCIS one-column format. All submissions are subject to screening for plagiarism by
iThenticate®. Double blind review system will be followed and each paper will be reviewed by at least three reviewers. Formatting style is
available in http://www.springer.com/computer/lncs?SGWID=0-164-6-793341-0. Each paper should be registered separately with at least
one full registration fee and papers only of student authors are eligible for student registrations (proof needed). Research scholars will not
be considered as any category of student.
Submission Link: https://easychair.org/conferences/?conf=csi20170

Contact Details
Organizing Chair: Dr. Phalguni Muherjee (+91 9433477369) Exhibition Chair: Mr. Sib Daspal (+91 9830333781)
(phalguni.mukherjee@gmail.com) (sibdaspal@gmail.com)
Organizing Co-Chair: Mr. Subir Lahiri (+91 9432284548) Exhibition Co-Chair: Mr. Gautam Hajra (+91 7980739842)
(subirlahiri49@gmail.com) (computerhom@gmail.com)
Technical Program Committee Chairs: Event Chair: Dr. Samir K. Bandopadhyay (+91 9831030823)
Dr. Kalyanmoy Deb, Michigan State University, USA (1954samir@gmail.com>)
Dr. J. K. Mandal (+91 8240323373) (jkm.cse@gmail.com) Event Co-Chair: Dr. A. K. Mukhopadhyay(+919434385282 )
Program Co-Chair: Dr. Devadatta Sinha (+91 9830269278) (askm55@gmail.com)
(devadatta.sinha@gmail.com) Registration Chair: Mr. Aniruddha Nag (+91 9830876463)
Tutorial Chair: Dr. Amlan Chakrabarti (+91 9831129520) (ani_nag123@rediffmail.com)
(achakra12@yahoo.com) Registration Co-Chair: Mr Abhijit Sarkar(+919831611033)
Tutorial Co-Chair: Dr. Debasis De(+918617256060) (abhijit_sarkar_1@rediffmail.com)
(dr.debashis.de@gmail.com) Publicity Chair: Mr. Subhobrata Roy Chaudhuri
Tutorial Convener: Dr. Somnath Mukhopadhyay (+919475413463) (sabsrc@yahoo.co.in)
(som.cse@live.com) Publicity Co-Chair: Dr. Kaushik Roy (+919432841740)
Finance Chair: Mrs. Alakananda Rao (+91-98300-58313) (kaushik.mrg@gmail.com)
(alka.rao@alvarisys.com) Hospitality Chair: Dr Manas Kumar Sanyal(+919434451520)
Finance Co-Chair: Mr. Sourav Chakraborty (sourav_ch@rediffmail.com) (manas_sanyal@rediffmail.com)
Convener: Mr. Devaprasanna Sinha (+91 9830129551) Hospitality Co-Chair: Mr Subhranshu Roy(+919051377068)
Subhranshu_81@yahoo.com
Interested persons are requested to contact the respective Chair/Co-chair for further details for participating in this mega event to be held
at the City of Joy, Kolkata. The weather during January at Kolkata will be very much conducive for this mega event. All are requested for wide
publicity of the event so that the information about the event can reach to each corner of the globe at the earliest for their comprehensive
planning to visit Kolkata during the event.

Convention Secretariat
Computer Society of India, Kolkata Chapter, 5, Lala Lajpat Rai Sarani (Elgin Road), 4th Floor, Kolkata - 700 020
E-mail : csical@gmail.com, chairoccsi2017@gmail.com, csi2017kolkata@gmail.com
 Registered with Registrar of News Papers for India - RNI 31668/1978 If undelivered return to :
Regd. No. MCN/222/20l5-2017 Samruddhi Venture Park, Unit No.3,
Posting Date: 10 & 11 every month. Posted at Patrika Channel Mumbai-I 4th floor, MIDC, Andheri (E). Mumbai-400 093
Date of Publication: 10th of every month

Call for Papers

International Conference on
Social Transformation - Digital Way
Proceedings of the 52nd Annual Convention of Computer Society of India (CSI 2017) In Technical Collaboration with
Host: Kolkata Chapter | 19-21, January, 2018
Venue : Science City, Kolkata
Proceedings will be published by Springer Nature Singapore in CCIS Series indexed by SCImago, Scopus, DBLP

Conference Chair: Dr. Goutam Mahapatra, DRDO,GOI, VP,CSI Computer Society of India
Convenor: Mr. D. P. Sinha, RVP-II, CSI
The Springer Nature Sponsored International Conference on Social Transformation - Digital Way, is being organized by the CSI
Organising Chair: Dr. Phalguni Mukherjee, CSI
Kolkata Chapter at the Science City during 19th-21st January 2018. Proceedings will be published by the Springer in CCIS Series
Organising Co-Chair: Mr. Subir Lahiri, CSI
(approved) indexed by SCImago, Scopus, DBLP etc. Extended versions of presented papers will be submitted for publication in S CI
Technical Programme Chairs: and Scopus indexed journals.
Dr. Kalyanmoy Deb, Michigan State University, USA The event is also technically sponsored by the IEEE Young Professionals, IEEE Computational Intelligence Society, BCC & I etc.
Dr. J. K. Mandal, University of Kalyani, India
Technical Programme Co-Chair: Dr. D. D. Sinha, Calcutta University The articles for the International Conference should be within 15 pages in the Springer CCIS one-column format. All submissions
Finance Chair: Ms. Alakananda Rao, Alvari Systems are subject to screening for plagiarism by iThenticate®. Double blind review system will be followed and each paper will be reviewed
Finance Co-Chair: Mr. Sourav Chakraborty, CSI by at least three reviewers.
Exhibition Chair: Mr. Sib Daspal, The Bengal Chamber & Sasha Important Dates :
Group Last date of submission: 15.11.2017; Date of Notification:15.12.2017; Camera Ready Submission and Registration: 31.12.2017.
Exhibition Co-Chair: Mr. Gautam Hajra, CSI
Registration Chair :Mr. Aniruddha Nag, CSI Submission link: https://easychair.org/conferences/?conf=csi20170
Registration Co-Chair: Mr. Abhijit Sarkar, JOBexo Call for Papers
Event Chair: Dr. Samir K Bandopadhyay, CU The theme of the CSI 2017 is “Social Transformation - Digital way”. It will cover all aspects of digital technology and machine
Event Co-Chair: Dr. A K Mukhopadhyay,CSI learning in computer science and engineering, general science, educational research and management science. This convention
Publicity Chair: Dr. Subhobrata Roychaudhuri, Techno University will provide a platform to the participants to share their views and ideas on latest technological developments in an inter- and intra-
Publicity Co-Chair: Dr. Kaushik Roy, WBSU disciplinary perspective. There will be total eight tracks for the international conferences.
Hospitality Chair:Dr. Manas Kumar Sanyal, KU
Tracks of the Conference (not limited to) • Digital Mapping in Environment
Hospitality Co-Chair: Mr. Subhranshu Roy, KU
Tutorial Chair: Dr. Amlan Chakrabarti, CU • Mobile and Wireless Communication • Intelligent Bio Computing
Tutorial Co-Chair: Dr. Debasish Dey, MAKAUT • Nano Device Modelling • Digital Judiciary
Tutorial Convenor: Dr. Somnath Mukhopadhyay, Hijli College
• Data Analytics • e-Learning and confidence based Learning
& RSC II
• Security and Forensics • Intelligent Data Mining
Advisory Committee(Partial)
Mr. Snajay Mahapatra, President, CSI Conference highlights
Dr. Anirban Basu, Immd. Past President, CSI Keynote/Invited/Banquet Talks, Technical Paper Sessions, Student Research Forum, Panel Discussions, Demos & Exhibits,
Dr. A. K. Nayak, Secretary, CSI Mentoring Sessions
Mr. Manas Ranjan Pattnaik, Tresurer, CSI Important Dates :
Dr. Aditya Bagchi, RKM University, India Formatting Guidelines: http://www.springer.com/computer/lncs?SGWID=0-164-6-793341-0
Dr. S Bandopadhyay, Calcutta University Submission link: https://easychair.org/conferences/?conf=csi20170
Mr. R. N. Lahiri, CSI WikiCFP Link: http://www.wikicfp.com/cfp/servlet/event.showcfp?eventid=66231
Dr. Kalyanmoy Deb, MSU, USA
Dr. Aynur Unal, Stanford University, USA Registration Fees :
Dr. Prith Banerjee, Schneider Electric, USA
Category of Delegates/ Authors Early Bird (On or before 10th After 10th October,2017 Spot Registration(Only in
Dr. Shikharesh Majumdar, CU, Canada October, 2017) Cash)
Dr. Amiya Nayak, OU, Canada
Dr. Sajal Das, UTA, USA Indian Authors & Delegates(in INR) including GST
Dr. Santosh Mohanty, TCS, India CSI Members General CSI Members General CSI Members General
Dr. Zbigniew Michalewicz, UA, Australia
Full time Students 1,800/-* 2,000/-* 2,300/-* 2,500/-* 2,300/-* 2,500/-*
Dr. Y Narahari, IISC Bengaluru, India
Dr. Hyeona Lim, MSU, USA Academician/Research Scholar 6,500/- 7,000/- 7,000/- 7,500/- 8,000/- 8,500/-
Dr. Rajkumar Buyya, UM, Australia
Industry/Govt./PSU 7,000/- 7,500/- 7,500/- 8,000/- 8,500/- 9,000/-
Dr. Somnath Mukhopadhay TU, USA
Dr. A. Kaykobad, BUET, Bangladesh One Day Registration 3,000/- 3,500/- 3,500/- 4,000/- 4,000/- 4,500/-
Dr. B. B. Chaudhuri, ISI Kolkata, India Pre-convention Tutorial INR 1,000 for authors and INR 1,500 for others
Dr. S. Satapathy, PVPSIT, India
Foreign Delegates $300 for Registration and $ 100 for Tutorial
Dr. Atulya Nagar, LHU, UK
Dr. B. K. Panigrahi, IIT Delhi, India *Including Tutorial Sessions
Dr. Basabi Chakraborty, IPU, Japan
Dr. Subarna Shakya, TU, Nepal Sponsorship Platinum Gold Silver Knowledge Partner Academic Conference Kit Conference Cultural Program
Dr. Girijasankar Mallik UWS, Australia Partner Dinner
Technical Program Committee Members: Amount (`) 15 Lakh 10 Lakh 5 Lakh 2.5 Lakh 1 Lakh 2 Lakh 2 Lakh 1 Lakh
Dr. S. K. Basu, BHU, India
(I) Each paper should be registered separately with at least one full registration fee.
Dr. P. K. Jana, ISM Dhanbad, India
Dr. A. Damodaram, TU, India (II) Papers only of student authors are eligible for Student registrations (proof needed).
Dr. Subir Sarkar, JU, India (III) Research scholars will not be considered as any category of student.
Dr. S. P. Maity, IIEST, India
Email: csi2017kolkata@gmail.com Accommodation: To be made on payment basis Convention Website: www.csi2017.in
Dr. Vipin Tyagi, CSI
Dr. Durgesh Nandi, CSI Group Registrations:
Dr. Paramartha Dutta, Viswa Bharti University, India Student Author Delegate
Dr. P. K. Roy, APIIT, India
Group of 4 or 5 Group of 6 or more Group of 4 or 5 Group of 6 or more Group of 4 or 5 Group of 6 or more
Dr. D. K. Pratihar, Indian Inst. of Technology Kharagpur, India
Dr. B. B. Pal, University of Kalyani, India ` 1,800 Per Head ` 1,600 Per Head ` 6,500 Per Head ` 6,000 Per Head ` 7,000 Per Head ` 6,500 Per Head
Dr. Dhananjay Bhattacharyya, SINP, Kolkata, India