Speech Encryption Algorithm Based On Blind Source Separation

The huge revolution of internet and information technology brought with it a high demand to develop new cryptosystems that provide adequate data security. For speech transmition, there are many cryptosystems exist, which are classified either as analog encryption or digital encryption (sometimes is called scrambling) such as chaos-based encryption.This thesis proposed a new speech encryption algorithm based on Blind Source Separation (BSS). Blind source separation is statistical approach used to separate independent source signals from their mixture without prior knowledge to the mixing coefficients. The basic idea of the proposed algorithm is to utilize the difficulty of solving the underdetermined BSS problem where the number of source signals is greater than the observations. First step and According to BSS mathematical principles, inseparable underdetermined mixing matrix was formed. Second, the mixing matrix then used to encrypt the segmented speech signals by mixing it with key signals which are pseudorandomly generated by Pseudo Random Number Generator (PRNG) and has the same number and size of the speech signals. To form the mixing matrix, Kronecker product of matrices and matrix convolution were used to construct two different forms of mixing matrix with high sensitivity to initial value changes. This important property of the mixing matrix increases the sensitivity of the encryption key thus increases the encryption algorithm security against plaintext attack, chosen plaintext attack, chosen ciphertext attack and other modern attacks. Third, to recover the speech signals back, Blind source separation by Fast lndependent Component Analysis "FastICA" algorithm vere used to separate the speech signals from the key signals. Finally, to realize this work, simulate experiments using Matlab R2009a software was done to simulate the algorithm then the results were analyzed and the efficiency and quality of the proposed speech cryptosystem is proved.