A little bit of DL-SCAhttps://nathan-rousselot.com/enhttps://nathan-rousselot.com/writing/gradient-plateau/https://nathan-rousselot.com/writing/gradient-plateau/Previous research linked the plateau effect to relying on SGD-based training. Can gradient-free DL-SCA overcome the plateau?Sun, 11 Oct 2026 00:00:00 GMThttps://nathan-rousselot.com/writing/is-it-really-broken/https://nathan-rousselot.com/writing/is-it-really-broken/How the way the DL-SCA community computes guessing entropy leaves the door open to false positives. And an introduction to a robust distinguisher.Sun, 01 Feb 2026 00:00:00 GMThttps://nathan-rousselot.com/writing/scoop-explained/https://nathan-rousselot.com/writing/scoop-explained/Scoop is a novel optimization algorithm, aimed at reducing the plateau effect. To achieve this, we made a theoretical study of what happens with gradient-based optimization against higher-order leakages.Tue, 01 Apr 2025 00:00:00 GMT