GPT-4 LLM agents autonomously exploit 87% of tested one-day vulnerabilities when given CVE descriptions, far outperforming other models and tools.
Automated vulnerability detection in source code using deep representation learning
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CR 1years
2024 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
LLM Agents can Autonomously Exploit One-day Vulnerabilities
GPT-4 LLM agents autonomously exploit 87% of tested one-day vulnerabilities when given CVE descriptions, far outperforming other models and tools.