USENIX Security Test of Time Award Honors Decade-Old ML Model Stealing Paper
thegautamkamath · x · 2026-08-14
Florian Tramer and colleagues won the USENIX Security Test of Time Award for their paper Stealing Machine Learning Models via Prediction APIs.
This was the authors' first ML security project, investigating attacks that extract model parameters via prediction APIs. Tramer noted that he never predicted how this 10-year-old work would shape his career and remain highly relevant today.
More from Safety
- Anthropic Rewrites Claude's Biology Classifier, Cutting False Positives by ~85% — dl_weekly · 2026-08-14
- Hidden Prompt Injection Found in CT Court Filing Leads to Sanctions — 404 Media · 2026-08-14
- AI Safety Memes Hit NYT: 'Frankenstein Shit' in SF Labs — ZeroStateReflex · 2026-08-14
- Agentic Risk Treatment Planning Paper Wins Best Paper Award at IEEE ISI 2026 — Fowe · 2026-08-14
- Arcforge: Open-Source Governance Layer for Autonomous Agents — TravelBusy8117 · 2026-08-14
- Friend Cloned My Personality into ChatGPT Using Secret Recordings and Chats — Previous_Foot_5328 · 2026-08-14