A discrete memristive heterogeneous neural network with grid multi-windmill hyperchaotic attractors and application in secure communication

Dynamic interactions between different neural networks can yield more complex dynamical behaviors. Nevertheless, research on systems composed of heterogeneous neural networks remains insufficiently explored. This study constructs a high-dimensional discrete memristive heterogeneous neural network (DMHGNN) which integrates two distinct neural networks leveraging a discrete memristor as a synaptic connection. Theoretical and numerical simulation results demonstrate that DMHGNN exhibits countless fixed points, different numbers of grid multi-windmill hyperchaotic attractors, and bidirectional initial offset-boosting characteristics. By adjusting the network parameters, the system possesses multiple positive Lyapunov exponents, revealing a more intricate hyperchaotic state and reflecting remarkable dynamical complexity. Furthermore, grid multi-windmill hyperchaotic attractors generated by DMHGNN have been successfully implemented on an FPGA platform. Finally, a DMHGNN-based image secure communication system is designed and evaluated, exhibiting good security performance in experimental validations.