Answers are grounded in this paper and cite their sources.
Novel PUF architecture enhances security against ML attacks, indicating strong IoT authentication potential.
Strong Physical Unclonable Functions (PUFs) are critical for lightweight authentication in the Internet of Things (IoT). However, traditional Strong PUF designs are susceptible to advanced Machine Learning (ML) modeling attacks. In this paper, we propose a novel modeling-attack-resilient Strong PUF architecture that transforms the challenge-response mapping from a statically approximable function into a mathematically rigorous Keyed Pseudo-Random Function (Keyed-PRF). The proposed architecture features two core innovations: a Sponge-Based Configuration Mechanism (SCM) that utilizes reliable Weak PUFs to dynamically configure the feedback polynomial of a Linear Feedback Shift Register (LFSR), effectively creating a device-specific secret key; and a Response-Modulated State Evolution Mechanism (RM-SEM), where the instantaneous physical responses of the underlying Arbiter PUF determine the evolution step size of the LFSR. This creates a deep temporal feedback loop that transforms the system into a Hidden Markov Model (HMM), blocking gradient-based learning strategies. We also introduce a reliability screening strategy based on a strict bit-error-rate threshold to ensure stability. Experimental results on Xilinx Artix-7 FPGAs demonstrate that the proposed PUF maintains a prediction accuracy of approximately 50% against four mainstream modeling attacks even with 1 million training Challenge-Response Pairs (CRPs). Furthermore, the design exhibits excellent uniformity, uniqueness, and reliability, achieving these security properties with minimal hardware overhead.
No takes yet. Share an insight, caveat, or question.
Chen et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: