Security Evaluation of Feed-Forward Interpose PUF Against Modelling Attacks
Weiping Xu, Lihui Pang, Yilong Tang, Mao Chen · 2024
Physical unclonable function (PUF) is a lightweight hardware-safe circuit structure for Internet of Things (IoT). The susceptibility to neural network modeling attacks stands out as a critical consideration in the design of PUF structures. Interpose PUF (IPUF) is a PUF variant that utilises the insertion topology of two levels on top and bottom. In order to enhance the security of IPUF, we propose a scheme to replace the upper and lower XOR APUFs of the IPUF using Feed-Forward XOR PUF (FFXOR PUF). The reconstruction of PUF called FF-IPUF, which will be evaluated for security using two basic neural network models: Multilayer Perceptron (MLP) and Side-channel Attacks (SCA). Experiments have demonstrated that the (4,4)-FF- IPUF, with enhanced resistance to MLP and SCA using four loops and separate loop positions, respectively, boasts the advantage of lower attack accuracy when confronted with the same training data.