Hardware Security for Edge Computing Via CMOS-Compatible Multi-Level Flash Memory with Hash-Based Key Generation

Abstract

Ensuring secure and energy-efficient authentication of resource-constrained devices has become a critical challenge owing to the rapid expansion of the Internet of Things (IoT) ecosystem. Physically unclonable functions (PUFs) that exploit inherent manufacturing variations to generate device-unique keys have emerged as a promising hardware-based security primitive. In this study, a PUF architecture is proposed that utilizes multi-level programming of flash memory capacitors. Reliable and reproducible binary responses are generated across multiple programmed states by extracting flat-band voltage variations from capacitance–voltage measurements. Performance is evaluated using the uniformity, P-value, inter-Hamming distance, and intra-Hamming distance, which indicated strong randomness, uniqueness, and stability. To further enhance the entropy and cryptographic strength, a hash-based message authentication code (HMAC)-based key derivation function is integrated, which transformed the PUF outputs into high-entropy pseudorandom keys. The final architecture supported key generation with zero standby power and is compatible with commercially available memory structures, thereby enabling low-cost and scalable deployment in secure IoT systems. These results highlight the potential of multi-level memory-based PUFs as a lightweight and robust solution for next-generation hardware security.