Post from 2025-11-17 10:07:56

This is a fun one: LLM inference creates a timing side channel that allows identifying sensitive topics by passively intercepting encrypted traffic:

https://www.microsoft.com/en-us/security/blog/2025/11/07/whisper-leak-a-novel-side-channel-cyberattack-on-remote-language-models/

/via @jonny

permalink | main