Reproduce “Power side-channel leakage localization through adversarial training of deep neural networks” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Jimmy Gammell, Anand Raghunathan, Kaushik Roy

AdvMask localizes timesteps in power traces by training a classifier to predict a sensitive AES intermediate while a learned, data-independent mask blends trace coordinates with Gaussian noise. Alternating updates make the mask disrupt prediction while an L1 penalty limits masking. In synthetic Boolean-masked traces, it identifies most or all leaking points when at least six are present; real-device reliability remains unresolved.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.