Machine Learning based Modeling Attacks on a Configurable PUF

Physical Unclonable Functions (PUFs) are one of the most advanced and secure hardware based security primitive for system authentication. However, with the introduction of Machine Learning (ML) modeling techniques, even the strongest PUF designs can be compromised. In this paper, we examine the resistance of a strong configurable RO-PUF design controlled by Programmable XOR gates (PXOR) towards machine learning based modeling attacks.