Discrete memristor-coupled dual-neuron chaotic model and its application in ROI selective facial privacy encryption

Le résumé fourni par la source

Abstract The need to protect the privacy of facial images has gained much significance in the context of the increased usage of cloud storage, smart surveillance, and biometric identification. Nonetheless, most of the current chaotic image encryption techniques either encode the image as a whole ignoring any semantic information or use chaotic systems with few dynamical complexities thus decreasing the computational and practical effectiveness. In order to solve these problems, a discrete memristor-coupled dual-neuron chaotic model is proposed in this paper and an region-of-interest selective facial privacy encryption scheme has been developed. Two bidirectionally coupled Chialvo neurons are coupled to a discrete flux-controlled memristor, with a multiplicative periodic modulation mechanism added to increase the nonlinearity and randomness of the system. The analysis of the dynamical behavior of the proposed model is presented in systematic fashion by the nonlinear dynamical analysis, showing that it has rich chaotic behavior, high sensitivity to initial conditions and good pseudo-randomness. These chaotic sequences are further combined with facial landmark localization to perform permutation and diffusion only within sensitive facial regions, while preserving non-sensitive image content. Experimental results demonstrate that the proposed scheme provides effective facial privacy protection, strong statistical security, robustness against noise attacks, and pixel-level lossless reconstruction under the correct secret key. The proposed method offers an efficient and reversible solution for semantic-aware facial privacy encryption.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.