Small Photonic Neural Networks as a Platform for Chaos-Based Cryptography

Abstract

There is a need for new approaches to security due to rapid growth in demand for high-speed communications, as well as advances in computing frameworks that threaten existing cryptographic standards, such as quantum computing. Photonic neural networks offer a potential solution by means of lightspeed computation and close integration with fiber optic networks; however, a rigorous and scalable encryption technique native to this platform has yet to be developed. In this report, a review of chaos-based cryptography techniques and the current status of photonic chaos-based encryption is presented, in order to ground the current work within previous research efforts. We then demonstrate a suite of custom programs for simulating small photonic neural networks (PNNs) and discovering new chaotic configurations, which draws on the isomorphism between PNN hardware and the continuous-time recurrent neural networks (CTRNN). Notably, this suite implements a program for estimating the Lyapunov exponent of experimental data in Python, which to our knowledge has not been documented in literature. Finally, we evaluate the performance of a pseudorandom-bit generator and the potential for chaos-based synchronization encryption. This simulation work lays the foundation for developing a novel chaos-based encryption protocol that is fast and that uniquely draws on the particular computational structures afforded by the photonic neural network.