Mapping the Exploitation Surface: A 10,000-Trial Taxonomy of What Makes LLM Agents Exploit Vulnerabilities

Documenti analoghi

• Black-Box Reliability Certification for AI Agents via Self-Consistency Sampling and Conformal Calibration
  di: Mouzouni, Charafeddine
  Pubblicazione: (2026)
• GradingAttack: Exposing Security Vulnerabilities in LLM Based Educational Grading Agents
  di: Li, Xueyi, et al.
  Pubblicazione: (2026)
• Taxonomy, Evaluation and Exploitation of IPI-Centric LLM Agent Defense Frameworks
  di: Ji, Zimo, et al.
  Pubblicazione: (2025)
• Pattern Enhanced Multi-Turn Jailbreaking: Exploiting Structural Vulnerabilities in Large Language Models
  di: Nihal, Ragib Amin, et al.
  Pubblicazione: (2025)
• Relevance as a Vulnerability: How Web Retrieval Degrades Safety Alignment in LLM Agents
  di: Nawal, Aditya, et al.
  Pubblicazione: (2026)