Pseudo-Random Generator Based on a Photonic Neuromorphic Physical Unclonable Function

In this work we provide numerical results concerning a silicon-on-insulator photonic neuromorphic circuit configured as a physical unclonable function. The proposed scheme is enhanced with the capability to be operated as an unconventional deterministic pseudo-random number generator, suitable for cryptographic applications that alleviates the need for key storage in non-volatile digital media. The proposed photonic neuromorphic scheme is able to offer NIST test compatible numbers with an extremely low false positive/negative probability below 10-14. The proposed scheme offers multi-functional capabilities due to the fact that it can be simultaneously used as an integrated photonic accelerator for machine-learning applications and as a hardware root of trust.