Modeling-Attack-Resilient Lightweight Strong PUF Based on Add-Rotate-XOR
A 3.22-pJ/bit, 3.5-MF² Modeling-Attack-Resilient Lightweight Strong PUF Based on Add-Rotate-XOR
Abstract
Strong physically unclonable functions (PUFs) provide an exponentially large challenge–response pair (CRP) space for authentication. However, conventional single-bit-response designs must be evaluated repeatedly to assemble a protocolsized response, increasing communication, latency, and energy overhead. Supporting wider responses reduces this evaluation burden but often involves elaborate transformation logic and substantial area overhead. This article presents a lightweight strong PUF architecture that supports 128-bit CRPs, aligning with standard cryptographic hardware interfaces. By leveraging add-rotate-XOR (ARX)-based iterative mixing, the proposed PUF combines modular addition with carry propagation, bit rotations, and roundwise XOR-based diffusion and entropy injection to achieve high diffusion, strong nonlinearity, and area efficiency. Implemented in a 180-nm CMOS process, the fabricated prototype occupies an area of 3.5 MF2 , achieves a measured throughput of 202 Mbps at 0.8 V within the reliability-characterized operating range, consumes as little as 3.22 pJ/bit, and exhibits a bit error rate (BER) below 5E-7. Statistical evaluations, including the NIST SP 800-22 test suite, the NIST SP 800-90B evaluation, and additional analyses, confirm high randomness and low interbit correlation across the 128-bit responses, even under correlated Gray code challenges. Modeling attacks trained on 40 million CRPs show that the prediction accuracy is close to random guessing, reflecting the high nonlinearity of the challenge–response mapping and its resistance to the evaluated machine learning attacks. These results indicate that the proposed PUF provides an efficient and secure authentication solution for resource-constrained IoT systems.
Thank you for visiting this project page. Register this project to explore its novelty and implementation support.
