A chaotic neural network-based encryption algorithm for MPEG-2 encoded video signal

Abstract

In this paper, a cipher algorithm based on chaotic neural network CNN is used and integrated inside MPEG-2 video codec system to encrypt and decrypt the quantised coefficients and the motion vector data. This symmetric cipher algorithm was used to transform the plaintext into an unintelligible form under the control of the key. Chaos theory property and its effect on cipher algorithm have been investigated. Result shows that a minor-key modification of the receiver side will lead to unclear video scene with very low PSNR value of -18.363 dB. To reduce the required execution time for CNN cipher algorithm; a motion vector of video signal was selected for encryption and decryption instead of the quantised coefficients. Results indicate little execution time for motion vector encryption and decryption process of 5.498 and 5.381 seconds respectively, but the entropy value decreases to 7.645 as compared to the entropy value of the quantised coefficients encryption. The whole system model can control bit rate and video quality depending on the available bandwidth channel. It can be shown from results that by increasing video quality value the PSNR and the compressed bit rate values will increase also, but with penalty of compression ratio decreasing.