A true random number generator based on nanowire-edge crossing memristors for image encryption.

True random numbers, owing to their inherent unpredictability, play a crucial role in information security and cryptography applications. In this work, a true random number generator (TRNG) circuit was built based on the reliable stochastic switching behaviors of conductive filaments (CFs), formed within a SiO2 dielectric at the crossing locations between n-type doped Si nanowires (SiNWs) and the edge of Ag pads. The ultrathin SiNWs, grown as an orderly array via an in-plane solid-liquid-solid mechanism with a typical diameter of ∼50 nm, can help in confining the individual CFs only at the crossing points. The controlled formation and rupture of nanoscale CFs enable stable memristor switching for generating long-term true random numbers. The generated random numbers pass all 15 NIST statistical tests without post-processing. This NW-edge crossing TRNG device was further utilized to produce random image keys, and accomplish rather high uniqueness and non-reproducibility. Evaluation with the state-of-the-art YOLO model showed that encrypted images retained visual fidelity with a 15% average mAP@0.5-0.95 decrease, while decrypted images recovered almost fully with less than 1.2% deviation. These results show that the NW-edge crossing TRNG strategy offers a secure and stable hardware encryption solution for image privacy and secure transmission.