Paper Accepted at ICCAD 2026

I am happy to share that our paper, “BREACH: Breaking Resilient NVM-based PUFs via Tailored ML Attacks,” has been accepted at the 2026 IEEE/ACM International Conference on Computer-Aided Design (ICCAD 2026).

The paper is co-authored by Mohamed Alsharkawy, Hassan Nassar, Kuan-Hsun Chen, and Jörg Henkel.

NVM-based physical unclonable functions have been considered particularly resistant to machine-learning modeling attacks because of their nonlinear multi-level-cell behavior. In this work, we introduce BREACH, a set of tailored modeling attacks that exploit architectural knowledge of the underlying PUF designs.

Across three state-of-the-art NVM-based PUFs, BREACH achieves modeling accuracies between 96% and 99.4%, substantially outperforming conventional machine-learning attacks.

The work shows that architectural complexity alone does not necessarily guarantee security when the same design knowledge can be used to construct more effective attacks.

Congratulations especially to Mohamed for leading this work, and I look forward to seeing it presented at ICCAD 2026.