Side-channel assisted modeling attacks on feed-forward arbiter PUFs using silicon data

Abstract

Physically Unclonable Functions (PUFs) are used for generating unique signatures from the complex manufacturing variations in integrated circuits. However, the majority of PUFs have been shown to be vulnerable to modeling attacks. In this work, we take a closer look at the vulnerability of feed-forward arbiter PUFs towards a combined side-channel and modeling attack using data measured from our 32nm test chips. The side-channel information from the feed-forward PUF construction is used as a catalyst for improving the performance of modeling attacks. This hybrid attack helps to push the prediction accuracies to very high limits (>98.5%), especially for larger PUF circuits. The hybrid attack yields around 7% improvement in prediction rates when compared to a conventional modeling attack mounted under the presence of error-inflicted challenge-response pairs.