← all datasets

InjecAgent

Emerging
2papers using it
2024first seen

InjecAgent is a benchmark containing 1,054 test cases that evaluates the vulnerability of tool-integrated large language model agents to indirect prompt injection attacks, focusing on two primary types of attack intentions: direct harm to users and exfiltration of private data.

Papers using InjecAgent (2)

InjecAgent β€” datasets β€” cybersecurity