Can Sensitive Information Be Deleted From LLMs? Objectives for Defending Against Extraction Attacks

Vaidehi Patil, Peter Hase, Mohit Bansal

Abstract

Evaluates methods for deleting sensitive information from trained LLMs, finding current unlearning approaches insufficient against determined adversaries.

Categories

Tags

Framework Mappings

Cite This Resource

@article{llmsec202500024,
  title = {Can Sensitive Information Be Deleted From LLMs? Objectives for Defending Against Extraction Attacks},
  author = {Vaidehi Patil and Peter Hase and Mohit Bansal},
  year = {2024},
  journal = {ICLR 2024},
  url = {https://arxiv.org/abs/2309.17410},
}