Professional Documents
Culture Documents
INTRODUCTION :
SPEECH encryption seeks to perform a completely reversible operation on a portion of speech, such that it s totally unintelligible to an unauthorized listener. The three most important criteria used to evaluate speech scramblers are: The scramblers ability to produce encrypted speech with low residual intelligibility. The extent to which the encryption and decryption processes affect the quality of the speech recovered by the intended recipient.
Several firsts were The first realization of enciphered telephony The first transmission of speech by Pulse Code Modulation (PCM) The first examples of multilevel Frequency Shift Keying (FSK) The first useful realization of speech bandwidth compression.
The Bandsplitter
CRYPTANALYSIS
Cryptanalysis of any incoming speech data may be carried out in two steps:
Identification of the scrambling/ciphering scheme i.e. system identification. Descrambling/deciphering using the various parameters of speech to obtain intelligible speech.
Time domain parameters: Short time energy function Short time autocorrelation function Short time average magnitude difference function
Frequency domain parameters: FFT of spectrum estimation Filter bank methods Homophonic speech processing Model based techniques LP-based spectral estimation Maximum entropy spectral estimation
Transform Domain : These scramblers operate on speech which has been sampled, digitized partitioned into frames of equal length, each containing N speech samples. Transformation is applied on them and then permuted.
Plaintext/Ciphertext Attack: Ciphertext Only Attack on a Fixed Permutation System: Ciphertext Only Attack on a Varying Permutation System :
Plaintext/Ciphertext Attack: Ciphertext Only Attack on a Fixed Permutation System: Ciphertext Only Attack on a Varying Permutation System :
Authen_Tag = SHA(Payload_Dataencrypted)
DIGITAL WATERMARKING
A secret digital watermark, speech_watermark, used for Non Repudiation and Authenticity. Both the sender and receiver share one reference watermark (comm_watermark) from an independent trustworthy party.
SPEECH WATERMARKING
Regular Pulse Excitation, Long-Term Prediction (RPELTP) is slightly varied to include the watermark
A GSM 610 coder generates 52 RPE pulses of 3 bits each for a speech frame of 20 ms duration. An RPE pulse pair in one frame is chosen to embed one bit of the digital watermark, and the pulse pair is selected according to key.
ADVANTAGE