VIGIL: Defending LLM Agents Against Tool Stream Injection via Verify-Before-Commit

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Lin, Junda, Zhou, Zhaomeng, Zheng, Zhi, Liu, Shuochen, Xu, Tong, Chen, Yong, Chen, Enhong
Format:	Preprint
Published:	2026
Subjects:	Cryptography and Security Artificial Intelligence
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

Internet

https://arxiv.org/abs/2601.05755