Whisper Leak: a side-channel attack on Large Language Models

Geoff McDonald, Jonathan Bar Or

Large Language Models (LLMs) are increasingly deployed in sensitive domains including healthcare, legal services, and confidential communications, where privacy is paramount. This…