Dual-state conversion for high-entropy and reconfigurable resistive memory-based physically unclonable functions

Abstract

As embedded and connected devices proliferate across smart electronics and Internet-of-Things platforms, hardware-level security has become increasingly important. Physically unclonable functions (PUFs), which leverage intrinsic process variations to generate device-specific fingerprints, offer a promising solution. Here, we propose a PUF architecture based on resistive random-access memory (RRAM) devices integrated with ultrathin silicon nitride (SiN) interfacial trapping layers. Systematic variation of the SiN thickness from 0 to 1.5 nm identifies the 0.5 nm configuration as optimal for enhancing stochastic filament formation, resulting in increased switching variability and entropy. Broad current-state distributions in both resistance states were converted into binary maps exhibiting ideal randomness metrics, including uniformity and diffusiveness near 50 % and entropy exceeding 0.94. These characteristics were maintained across multiple bit-map sizes. Furthermore, repeated SET/RESET cycling of a single memory cell enabled the generation of multiple distinct PUF responses with consistent entropy and uniqueness. These results establish interface-engineered RRAM as a high-entropy, reconfigurable, and fabrication-compatible platform for secure key generation in edge and embedded systems. (c) 2026 Published by Elsevier Ltd on behalf of The editorial office of Journal of Materials Science & Technology.