Video Encryption And Compression Algorithm Based On Neural Networks

With fast development of computer network and information technology, the applications for multimedia information is being used in wilder fields, Multimedia information can be expediently and fast transmitted in network. However, the globality and openness of Internet severely threat the information security. Video is a special expression of multimedia information, which has many advantages such as visual, dramatic and intuitive performances. When problems of storage and transmission caused by a huge amount of data have been overcome, video has become a primary communication method. So the security of video information is becoming the focus on public interest, which has become a worldwide problem.Most of the video encryption algorithms are based on H.26X or MPEG.X, all of these method meet the requirement of video encryption technology more or less. To further enhance their security level, new theories and new methods need to be explored to give new scheme for encrypting video. Neural network has been applied to many fields. However, there have been only a few reports about its application to video encryption so far.This paper explores the feasibility of neural network-based video encryption, the main researching job of this dissertation are as follows:By studying the existed algorithms, a novel method for video encryption and compression integration has been proposed. Because of the structure of bottle-neck, BP neural network is suitable for image compression. Based on the characteristic of non-liner of BP neural network, this paper applies BP neural network to the method of video encryption and compression integration. In order to meet the requirement of real-time video transferring, we proposed the algorithm of combining the on-line learning and off-line learning, the new method reduces the number of training. The test showed that this method exhibits nice security, low computational complexity, has no impact on encoding efficiency and can enhance the processing speed of video data.Based on the existing entropy encoding encryption algorithm, a new chaotic video encryption algorithm based on the video's vector quantization. This scheme chooses SOM neural network as the compression platform, the chaotic sequences are generated by the Logistic chaotic mapping, then encrypt the sets of indices generated by the vector quantization using the chaotic sequences. This encryption method is so easy that it has little impact on the real-time transferring. Experimental results and analysis show that this algorithm exhibits so strong security level that it can resist any attacks.