You are on page 1of 4

2007 IEEE International Symposium

on Signal Processing and Information Technology

CAPTCHA for Blind People


Mohammad Shirali-Shahrezal, Sajad Shirali-Shahreza2
Computer Science Department, Sharif University of Technology, Azadi Street, Tehran, IRAN
Email: shraligcs.sharifedu, Website: hp ohammad.shiralir
2
Computer Engineering Department, Sharif University of Technology, Azadi Street, Tehran, IRAN
Email: shirali&e.sharifiedu, Website: http://sajad.shira1iir

Abstract - Nowadays the Internet users are from different with their children and relative using the Internet. In
ages and groups. The disabled people also use the Internet. addition to regular peoples, the disabled peoples are also
Some websites are especially created for disabled people. using the Internet.
Many Internet sites offers services for human users, but There are registration forms on many websites which
unfortunately some computer programs are designed which users should fill out them to be able to use website services.
abuse these services. As a result some systems named Unfortunately some vandalistic persons write bot programs
CAPTCHA (Completely Automated Public Turing test to tell which are filling out the registration forms with incorrect
Computers and Human Apart) have been introduced to tell information automatically. These attacks lead to interruption
apart human users and computer software. In this paper, a in websites operations and websites cannot offer services to
new CAPTCHA method is introduced which can be used by the users.
blindpeople. In this method a simple mathematicalproblem Various methods are presented to overcome above
is created according to predefined patterns and converted problem. The goal of these methods is distinguishing human
to speech using a Text-To-Speech (TTS) system. Then the users from computer programs. These methods are done
sound is played for the user and helshe must enter the automatically by computer programs since examination of a
answer of the question. Because answering this problem large amount of registration forms by human forces needs a
requires recognizing the speech, understanding the great deal of time and expense.
problem, and solving the problem, only a human user can CAPTCHA (Completely Automatic Public Turing Test
answer this question and present computer programs are to Tell Computer and Human Apart) are systems which are
unable to solve it. In addition, answering the question is used to tell human and machine apart automatically. These
easy for blind people, because the question is consists a systems are based on Artificial Intelligence (Al) topics.
number of natural language sentences and the answer is a They are similar to Turing test, but they differ in that the
number, which can be entered easily. judge is a computer [1]. The goal of these systems is to ask
a questions which human users can easily answer, while
Keywords - Blind People, Completely Automated Public current computer programs cannot. CAPTCHA systems also
Turing test to tell Computers and Human Apart have other applications such as prevention from sending
(CAPTCHA), Disabled People, Natural Language spam [2].
Processing (NLP), Text-To-Speech (TTS). CAPTCHA methods can be generally divided to two
groups:
1. INTRODUCTION 1- OCR-based: Based on the weaknesses of Optical
Character Recognition Systems.
Today lots of daily activities such as education, shopping 2- Non-OCR-Based: Mainly based on the weaknesses
and mailing are done through the Internet. By rapid growth of multimedia retrieval systems.
of the Internet and the easy access to it, a great deal of In OCR-based systems, image of a word is showed to the
private and personal information is available on the web. user after distortion and with various visual effects and
On the other hands, nowadays the Internet in not only for he/she must type it. Considering existence of various visual
special groups of people, but also peoples from any age and effects, computer programs will encounter problems in the
different groups are using the Internet. There are many recognition of these words and only a human user can
websites for children on the Internet and the children using recognize the concerned word. Examples of these methods
the Internet for various activities such as entertaining, are: Persian/Arabic Baffle Text [3] and Gimpy [4]. But
educating, etc. The elderly people communicate and chat these methods cause users' dissatisfaction. On the other

978-1 -4244-1 835-0/07/$25.00 2007 IEEE 995


hand some methods have been innovated to attack OCR- Unlike OCR-based CAPTCHA methods, the users of any
based methods [5]. language in any age group can use this program and run it
In contrast, there are Non-OCR-Based methods which on devices with limited resources.
are more comfortable for users than OCR-based methods.
These methods mainly based on the features of multimedia B. Collage CAPTCHA [7]
such as pictures and sounds and usually using methods like
small puzzle games. Examples of these methods are PIX [6] This method is a CAPTCHA method which is developed
and Collage CAPTCHA [7]. by authors. In this method, the images of some different
However, some of the Internet websites are especially objects (for example six objects such as airplane, car, apple,
designed for disabled persons. Also these websites need orange, pineapple and ball) are chosen. Then some effects
protection against computer programs which try to use such as rotation are done on the images and they merged to
website resources. But common CAPTCHA systems are create a single image (Fig. 1). This image is shown to the
usually difficult for disabled people. In next section, some user and he/she is asked to click on a certain object (for
of the Non-OCR-Based methods which can be used by example on the image of the apple).
disabled people are described. Like PIX [4], Collage CAPTCHA requires a database of
The suggested method in this paper is a Non-OCR-based labeled images. Creating this database is expensive and
CAPTCHA method which is designed for blind people. In requires a lot of time. It is an easy CAPTCHA method for
our suggested method, a simple mathematical problem is users, because in this method the user must find the object
created according to predefined patterns and converted to image which its name is shown. In addition, this method
speech using a Text-To-Speech (TTS) system. Then the may have high rate of random passing. If the images of
sound is played for the user and he/she must answer the different objects can be easily become separated, and the
question. A computer requires the following abilities to number of different objects is N, the probability of passing
answer the question: the test with random answer can be 1/N.
1- Recognition of the question using Speech Disabled people such as hearing or sight impaired
Recognition systems. persons can use this method because it uses images without
2- Understanding the meaning of the question. any distortion. In addition, mobility impaired persons can
3- Solving the problem and answering the question. use this method easily because it requires only one click.
Since it is difficult for computer program to succeed in
doing any of these operations, only a human user can C. ASIRRA [10]
answer the question. In section 3 we will explain this
method and its experimental results. This method is developed by Microsoft Research with
To answer in this way the only thing the user has to do is the help of "Petfinder.com". ASIRRA is the abbreviation of
to enter a number (which is the answer of the problem) and "Animal Species Image Recognition for Restricting
we don't have to type long phrases. Answering the question Access". In this method, a number of different images (for
is easy for blind people, because the question is consists of example 12 images) of cats and dogs are shown to the user
some natural language sentences and the answer is a and the user must select the images of cats.
number, which can be entered easily.
In the final section, the conclusion will be made after Collage CAPTCHA
investigating and studying some advantages of this method. W mtteoby:
Mohamad Shirali-Shahreza
2. RELATED WORKS Sajad Shirali-Shahreza
Here, we study and examine some Non-OCR-Based
CAPTCHA methods which can be used by disabled people.

A. Drawing CAPTCHA [8]

This method is for devices like PDA (Personal Digital


Assistant) which uses stylus. In this method, numerous dots
are drawn on a screen with noisy background and the user is
asked to connect certain dots to each other.
In view of the problems that computers face in
recognizing the dots from the noise, only a human user can Sel:ect the cat imrage.
easily identify the special dots and connect them to each
other. Fig. 1. A sample of Collage CAPTCHA [9]

996
The random pass rate of this method is low. This method As you see, the user has only to enter a number. We can
is similar to Collage CAPTCHA [7], because the name of increase the diversity and variety of the questions by
the object (which is always "cat") is shown to the user. designing different and various patterns and even make
This method has a large database (two million images), them more difficult and more sophisticated, too.
which is created manually [10]. Now the question is converted to an audio file. This is
One of the problems of ASIRRA method is that done using a Text-To-Speech system. At last the question
distinguishing between cats and dogs cannot be done easily audio file will be played for the user.
for all images. For example, deciding whether the Figure 2 As we specified earlier, it is easy for human users to
is cat or dog is not easy. This image is selected from the answer these questions and the only thing they must do is to
ASIRRA website [10]. enter a number. Consequently, little time is required to
Disabled people such as hearing or sight impaired answer and the user can answer these questions quickly.
persons can use this method because it uses natural images But answering these questions is difficult for a computer
without any distortion. program because the computer requires at least three
abilities to answer these questions:
3. OUR SUGGESTED METHOD 1- Computer must recognize phrase said through a
speech recognition system.
In this section, our suggested method has been described 2- After recognition of sound, the computer should be
and implementation of an example of it has been explained. able to understand the meaning of the question.
In this method, at first, on the basis of a series of pre- 3- At last and even if a computer does the two above-
designed patterns, a question is prepared. In these patterns, mentioned stages successfully, it must be capable
some of the elements of the problem are variable and to answer the question.
changeable and they are chosen from some items randomly. The percentage of success of computers is low in each of
For example a pattern of the problem is as follows: the mentioned stages. As a result, if the user can answer the
There are a1 numbers of b1, a2 numbers of b2 and a3 question correctly, we can conclude that the user is a human
numbers of b3 on a table. How many c1s are there on the user.
table? To implement this project, at first we made some patterns
In this pattern, b, is name of the object, for example according to the structure which was previously mentioned.
apple, ball, pencil, cat, etc. and ai > 0 is the number of b, Then a program was designed for making a sample question
objects. Each object belongs to a specific group which is based on these patterns. The program is written in Java
showed with cj. For example cat, dog, and mouse are among programming language.
pets which are showed for example with c2. Or pencil, At first this software chooses one of the patterns
paper, and book are among stationery. As a result, if a randomly, and then it makes a choice from among the
problem is made with this pattern it may be as follows: existent and available choices for each of the variables and
"There are 5 cats, 3 apples, and 4 dogs on a table. How puts them in the pattern.
many pets are there on the table?" After the question was made, it is converted to an audio
The answer is 7 (3 cats+4 dogs=7 pets). file. We use the FreeTTS. FreeTTS is a speech synthesis
Or it may be a problem like the following one: system written entirely in the Java programming language.
"There are 2 pencils, 3 books, and 1 mouse on a table. It is based upon Flite, a small, fast, run-time speech
How many fruits are there on the table?" synthesis engine, which in turn is based upon University of
Here the answer is 0 because there is no fruit on the Edinburgh's Festival Speech Synthesis System and Carnegie
table. Mellon University's FestVox project [11].
Now the audio file is played for the user. If the user
enters the answer (which is a number) correctly, the
software will play a successful message and allows the user
carry out the concermed operations.
4. CONCLUSION
In this paper a new CAPTCHA method has been
222224~~~I&I ~presented which is designed for blind people. In this method
a simple mathematical problem is created according to

Fig. 2. A sample of images used in ASIR1RA [10]

997
Considering the computer's difficulty in recognition of REFERENCES
the question from audio, understanding of the problem, and
solving the problem, only a human user can answer the [1] H.S. Baird and J.L. Bentley, "Implicit CAPTCHAs,"
question. Proceedings SPIEIIS&T Conference on Document Recognition and
Retrieval XII (DR&R2005), San Jose, 2005, pp. 191-196.
The question is said as a natural language phrase and
there is no need to read anything, so blind people can also [2] N. M. Boers and P. Gburzynski, "An Automation of Mail
use this method easily. this method easily. Channels," Proceeding of International Conference on Internet and Web
Applications and Services/Advanced International Conference on
Unlike OCR-based CAPTCHA methods, this method Telecommunications (AICT-ICIW 06), Washington, DC, USA, February
requires only typing of a number as the answer. Therefore it 2006, pp. 210.
is easy to use, saves user time and does not cause any [3] M.H. Shirali-Shahreza and M. Shirali-Shahreza, "Persian/Arabic
discomfort for users. Typing a number is not difficult for Baffletext CAPTCHA," Journal of Universal Computer Science (J. UCS),
vol 122, no.
o112, December
eebr06p.18-76
2006, pp. 1783-1796.
blind people and is very easier than typing a series of
characters. [4] L. von Ahn, M. Blum, and J. Langford, "Telling Humans and
By using different TTS systems, this method
Byuingdiferem
m d can be
, ti be
cn
Computers
pop, Apart Automatically", Communications of the ACM, vol. 47,
feray20,p.5-0
no. 2, February 2004, pp. 57-60.
localized for different languages. For example, the question
lcalbaizd
can be said in dFrentFh
for
French.
[5] K. Chellapilla and P. Simard, "Using Machine Learning to
Break Visual Human Interaction Proofs (HIPs)", Advances in Neural
We can increase the degree of difficultness of this Information Processing Systems (NIPS'2004), MIT Press, Vancouver,
method and as a result the degree of the resistance against Canada, December 2004.
the attacks through changing pattern of questions. But doing [6] M. Blum, L. von Ahn, and J. Langford, The CAPTCHA Project,
this will cause the program to be more difficult for human (Completely Automatic Public Turing Test to tell Computers and Humans
users, too. Apart), School of Computer Science, Carnegie-Mellon University,
In this method it is not necessary to have a keyboard and November 2000, http:!!www.captcha.net.
we have only to enter a number. Thereforewewe canonly
have [7] M.Proceedings
usetoetraubrTeeCAPTCHA,"
this Shirali-Shahreza and S. Shirali-Shahreza, "Collage
of the 20th IEEE International Symposium Signal
method on devices which don't have a keyboard or on Processing and Application (ISSPA 2007), Sharjah, United Arab Emirates,
devices in which it is difficult to use a keyboard, such as February 12-15, 2007.
mobile phones and Pocket PCs. Furthermore we can use [8] M. Shirali-Shahreza and S. Shirali-Shahreza, "Drawing
them not only on web pages but also on computer programs CAPTCHA," Proceedings of the 28th International Conference Information
as well. Technology Interfaces (ITI 2006), Cavtat, Dubrovnik, Croatia, June 19-22,
If we change question patterns we can put forth other 2006, pp. 475-480.
problems than mathematical problems. For example we can [9] M. Shirali-Shahreza and S. Shirali-Shahreza, Collage
design a medical question such as "How many bones are CAPTCHA website, 2007, http://www.hipir/Co1lageCAPTCHA/.
there in a human's jaw?" [10] Microsoft, Asirra (Animal Species Image Recognition for
We can customize this method for special websites. For RestrictingAccess), 2007, http://research.microsoft.comasirra/.
example we can use medical questions like the above- [11] W. Walker, P. Kwok, and P. Lamere, FreeTTS Open Source
Speech Synthesis System, Speech Integration Group, March 2005,
mentioned question for medical websites. http://freetts.sourceforge.net .

998

You might also like