Machine Learning Vulnerability Assessment of Ring Oscillator Physical Unclonable Functions
Husam Kareem, Dmitriy Dunaev · 2023
Devices authentication and cryptographic keys production techniques are implemented in various security applications. Because of its ability to generate unique, unclonable, and unpredictable cryptographic keys, physically unclonable functions (PUFs) have been utilized in various hardware security applications. However, PUFs can be vulnerable to adversary attacks like machine learning (ML)-based modeling attacks. Input-output combinations of PUFs are called challenge-response pairs (CRPs). This article provides a vulnerability assessment of three recent Ring Oscillator (RO) PUFs approaches against ML modeling attacks. Assuming that adversary obtains part of the RO-PUF CRPs, and they can be used to predict the unrevealed CRPs. The article has a secondary objective to assess the impact of aging on the tested RO-PUF approaches in regenerating the same response.