The evidence behind the guides

Sources

Find a study, check what it supports, and follow it back to a concern.

TopicCollection

Sources for Weight protection in inference

3 sources

Guide citations are still under editorial review.

Zhu and colleagues · USENIX Security 2021 Research study

Hermes Attack: Steal DNN Models with Lossless Inference Accuracy (original source, new tab)

Demonstrates extraction through unencrypted bus traffic on tested GPU platforms; does not establish exposure of all modern protected hardware.

Horváth and colleagues · USENIX Security 2024 Research systematization

SoK: Neural Network Extraction Through Physical Side Channels (original source, new tab)

Examines platform-dependent attacks, assumptions and practical limits, with experimental validation on an accelerator.

NVIDIA reference architecture Official technical documentation

Trust and Threat Model: Deploying Proprietary Models Securely with NVIDIA Confidential Computing (original source, new tab)

Lists trust assumptions and exclusions for a specific architecture, including application logging, compromised trusted roles and side channels.

Sources retain their original publishers’ rights. Inclusion does not imply endorsement. Saved sources stay in this browser on this device. Dates and source descriptions come from the existing editorial records.