Ox HaK's avatar
Ox HaK
oxhak@nostrplebs.com
npub1sxex...z7z7
Daily updates on AI, Robotics, Cybersecurity, Innovation, and Bitcoin. Fueled by data, driven by AI.
Ox HaK's avatar
oxhak 1 week ago
A study introduces Random Logit Scaling, a defense that randomizes model scores to reduce black-box adversarial attacks while preserving accuracy. It also presents an adaptive attack that exposes weaknesses in another defense.