Efficient image encryption scheme based on 7D-FHN neural network with bit-Level permutation and bidirectional column cascaded diffusion

Abstract In response to the prevalent challenge of balancing security and efficiency in current image encryption schemes, this study proposes a novel image encryption approach based on a seven-dimensional FitzHugh-Nagumo neuron model (7D-FHN) integrated with bit-level scrambling and column-wise bidirectional diffusion. The research aims to offer new perspectives for applying neural dynamics in the field of information security and to establish a feasible pathway for designing efficient and secure image encryption systems. By utilizing memristors to simulate biological synapses for coupling heterogeneous neurons, a 7D-FHN neural network is constructed, and its rich chaotic dynamics are systematically analyzed through numerical simulations. Building on this foundation, an image encryption mechanism that combines bit-level scrambling with column-wise bidirectional diffusion is introduced. This mechanism employs two independent chaotic sequences to perform forward and reverse column-wise diffusion operations on the scrambled image, thereby significantly enhancing security while improving encryption efficiency. Finally, a series of reliability and security tests are conducted, along with comparative analyses against existing typical image encryption schemes. Experimental results demonstrate that the proposed scheme exhibits superior robustness against various attack methods, such as statistical and differential attacks, while also achieving higher time efficiency, making it suitable for image encryption applications with stringent real-time requirements.