An XOR-Based Pico-Physically Unclonable Function for Securing IoT Devices

Abstract

Physically unclonable function (PUF) can be applied as a lightweight way to improve the security of Internet of Thing (IoT) devices. In the existing PUF studies, reconfigurable Pico-PUF (RPPUF) is an effective solution with good uniqueness and reliability. However, it still has a limited key space and requires extra hardware resources to generate more challenge-response pairs (CRPs). Therefore, this paper improves the RPPUF and proposes a lightweight XOR-based Pico-PUF, namely XORPPUF. By replacing each NOT gate in the configurable logic with an XOR gate, the key space is effectively expanded while preserving the PUF performance. We have implemented and verified the proposed XORPPUF on Xilinx Spartan-6 XC6SLX25 microboards. The experimental results show that XORPPUF achieves 40.06% uniqueness and 99.49% temperature reliability. Compared with the RPPUF, our work improves temperature reliability by 0.26%, expands key space by 2n, and reduces hardware resources overhead by 11.3% when generating a 128-bit PUF response. In addition, the prediction rate of our XORPPUF against Decision Tree (DT) and Random Forest (RF)-based modeling attacks is 34.17% and 39.40% lower than RPPUF, respectively. This means XORPPUF performs better resistant than RPPUF in Machine Learning (ML) attack. Thus, it is more suitable for securing the IoT devices with limited resources.