Computer use agents can be hijacked through visual deception, indirect prompt injection leading to code execution, and leakage of chain-of-thought reasoning.
Osworld: Benchmarking multimodal agents for open-ended tasks in real computer environments,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
dataset 1
citation-polarity summary
fields
cs.CR 1years
2025 1verdicts
CONDITIONAL 1roles
dataset 1polarities
support 1representative citing papers
citing papers explorer
-
A Systematization of Security Vulnerabilities in Computer Use Agents
Computer use agents can be hijacked through visual deception, indirect prompt injection leading to code execution, and leakage of chain-of-thought reasoning.