Synchronization of chaotic neural networks for secure communications

Methods for synchronizing discrete time chaotic neural networks are presented with applications in secure communications. Chaotic neurons, characterized with a piecewise linear transfer function, are connected into Hopfield-like networks. The networks are used as transmitter and receiver in chaotic communications. The first algorithm is a modification of simple chaotic masking which makes synchronization robust and insensitive to the perturbation from the added information signal. A mathematical proof and simulation results of the scheme are shown for small networks. We have verified the method experimentally, using single-neuron circuits. The second algorithm utilizes modulation of the transmitting chaotic network and detection of the corresponding synchronization error at the receiver.