Exploiting Quasi-Volatility in Silver-Doped RRAMs for Physical Unclonable Functions Toward Robust Security Primitives

Resistance random access memory (RRAM)-based physical unclonable functions (PUFs) have emerged as promising security primitives owing to their ability to generate reliable, unpredictable maps. In this study, we propose a novel PUF architecture that leverages the quasi-volatility of temporal and spatial randomness in a $32\times 32$ crossbar array composed of silver-doped RRAM devices. The proposed PUF constructs a unique identifier based on the intrinsic switching variability and temporal retention characteristics of the RRAM devices. This dual source of randomness ensures high unpredictability and robustness against environmental noise and common modeling attacks. In optimized experiments, the PUF achieved uniformity, inter-hamming distance (inter-HD), and intra-hamming distance (intra-HD) metrics of 49.9%, 50.11%, and 0%, respectively, demonstrating strong stability and resistance to bit errors. The architecture incorporates a temporal majority voting (TMV) scheme to correct bit errors without additional hardware, achieving a bit error rate (BER) of 0%. These results highlight the potential of this RRAM-based PUF for secure key and identification-generation applications, especially in demanding internet of things (IoT) and embedded systems. Future work will focus on further evaluating the resilience against advanced attacks and integrating the PUF into complete security systems.