A Video Compression and Encryption Technique Using Autoencoder and Chaotic Neural Network

This paper develops a video data compression and encryption technique by uniting autoencoder and time-delayed chaotic neural network (TD-CNN) to enhance data privacy while reducing its size. An autoencoder is a sort of deep neural network (DNN) that employs various functions and layers to decrease data dimensions across unsupervised learning leading to improved transmission delay. A TD-CNN is a cryptographic system that exhibits chaotic interactions among its elements, influenced by time delays, and is implemented through functional differential equations (FDEs). Here, 256x256 pixels of RGB images are compressed into 10x10 ones. A random binary sequence is generated of 32-bits by a single precision floating point. Another 10-bits are added for cyclic rotation and trajectory selection. Each compressed pixel is encrypted through a robust cryptographic system. Then, the designed technique offers a better solution for video data transmission than usual compression and encryption works. At last, the performance analysis shows that the developed technique not only fortifies the security of the data significantly but also strives to enhance the efficiency of data transmission by addressing temporal complexities through the utilization of TD-CNNs within the chaotic framework.