Search papers, labs, and topics across Lattice.
Fudan University
4
0
4
Existing security scanners misidentify nearly half of MCP server risks, challenging the reliability of current security assessments in LLM applications.
Current LLMs can autonomously penetrate systems with success rates up to 69.3%, revealing alarming implications for cybersecurity.
LLM-based cybersecurity agents can now autonomously adapt and improve their attack strategies, outperforming even human-designed systems.
LLMs exhibit an "Alignment Illusion," where their apparent safety collapses under pressure, with the most capable models showing the most dramatic failures.
