LLM Safeguard is a Double-Edged Sword: Exploiting False Positives for Denial-of-Service Attacks

Documenti analoghi

• Rethinking Latency Denial-of-Service: Attacking the LLM Serving Framework, Not the Model
  di: Wang, Tianyi, et al.
  Pubblicazione: (2026)
• From Stealthy Data Fabrication to Unsafe Driving: Realistic Scenario Attacks on Collaborative Perception
  di: Zhang, Qingzhao, et al.
  Pubblicazione: (2026)
• Semantic Denial of Service in LLM-controlled robots
  di: Steinberg, Jonathan, et al.
  Pubblicazione: (2026)
• ThinkTrap: Denial-of-Service Attacks against Black-box LLM Services via Infinite Thinking
  di: Li, Yunzhe, et al.
  Pubblicazione: (2025)
• ReasoningBomb: A Stealthy Denial-of-Service Attack by Inducing Pathologically Long Reasoning in Large Reasoning Models
  di: Liu, Xiaogeng, et al.
  Pubblicazione: (2026)