Implementing hardware primitives based on memristive spatiotemporal variability into cryptography applications

Implementing hardware primitives into cryptosystem has become a new trend in electronic community. Memristor, with intrinsic stochastic characteristics including the switching voltages, times and energies, as well as the fluctuations of the resistance state over time, could be a naturally good entropy source for cryptographic key generation. In this study, based on kinetic Monte Carlo Simulation, multiple Artificial Intelligence techniques, as well as kernel density map and time constant analysis, memristive spatiotemporal variability within graphene based conductive bridging RAM (CBRAM) have been synergistically analyzed to verify the inherent randomness of the memristive stochasticity. Moreover, the random number based on hardware primitives passed the Hamming Distance calculation with high randomness and uniqueness, and has been integrated into a Rivest-Shamir-Adleman (RSA) cryptosystem. The security of the holistic cryptosystem relies both the modular arithmetic algorithm and the intrinsic randomness of the hardware primitive (to be more reliable, the random number could be as large as possible, better larger than 2048 bits as NIST suggested). The spatiotemporal-variability-based random number is highly random, physically unpredictable and machine-learning-attack resilient, improving the robustness of the entire cryptosystem.