[Paper] Detecting Hidden ML Training With Zero-Overhead Telemetry
Source: arXiv - 2606.19262v1
Overview
Hardware-enabled monitoring of GPU workloads underpins many proposals for AI compute governance, but if developers can defeat monitoring mechanisms, such schemes are unworkable. We evaluate the adversarial robustness of GPU workload classification using only zero-overhead, privacy-preserving NVML telemetry: content-agnostic signals that observe physical effects of computation without accessing model weights, training data, or hyperparameters. Across 5 rounds of monitor-evader iteration, we evaluate 20 evasion strategy families on 9 GPU models spanning 4 architecture generations. We develop a classifier that achieves 98.2% binary accuracy at identifying training workloads across the whole corpus, and 43-87% accuracy against the most challenging unexpected workloads even when they are adversarially disguised.
Key Contributions
This paper presents research in the following areas:
- cs.LG
Methodology
Please refer to the full paper for detailed methodology.
Practical Implications
This research contributes to the advancement of cs.LG.
Authors
- Robi Rahman
- Sabiha Tajdari
Paper Information
- arXiv ID: 2606.19262v1
- Categories: cs.LG
- Published: June 17, 2026
- PDF: Download PDF