Discrete memristive cellular neural network and its application in real-time video encryption

Abstract

With the rapid growth of multimedia data transmission over potentially insecure channels, secure video encryption has become increasingly important. A five-dimensional discrete memristive cellular neural network is constructed by embedding voltage-controlled and magnetically-controlled discrete memristors into a three-cell neural network with heterogeneous activation functions, forming a novel class of discrete maps governed by difference equations. Systematic analysis via Lyapunov exponent spectra and bifurcation diagrams reveals that the system exhibits rich dynamical behaviors including hyperchaos under the variation of synaptic weight and electromagnetic induction parameters. Exploiting the pseudorandomness of hyperchaotic sequences, a three-stage encryption architecture is devised, comprising Josephus-ring dynamic permutation of the keystream, XOR-based pixel masking with forward diffusion, and chaotic dynamic S-box substitution. The complete system is validated on a field-programmable gate array platform, achieving real-time video encryption at high-definition resolution and frame rate.