covert-channel attack

Mentions · 7d0
Mentions · 30d0
All mentions1
First seenSep 23, 2026
Last mentionedSep 23, 2026

Timeline

1 articles · newest first

Your Model Is Leaking: Covert Information Transfer through LLM Residual Streams

A residual-stream covert-channel attack hides sensitive information in LLM intermediate activations, recoverable with a simple linear decoder without retraining or weight modification.

arXiv cs.CRSep 23, 12:27 UTC · Sep 23, 2026AI safety & securityExploited1 · 1 read

Appears with

privacy-sensitive organizationsorganization · 1
LLMai model · 1
large language models (LLMs)product · 1
large language models (LLMs)ai model · 1

Entities are extracted by the model from each article. First seen Sep 23, 2026, last mentioned Sep 23, 2026. Watching keeps it in this browser only (no account); the watchlist page and dashboard alerts use it.