analiza aiqbench
This research exposes a critical availability vulnerability where the advanced reasoning capabilities of LLM guardrails can be weaponized into denial-of-service attacks via extended inference loops. By demonstrating significant token amplification and latency spikes across major model backbones, the study highlights that safety mechanisms may inadvertently become system bottlenecks when targeted by sophisticated prompt engineering. The findings underscore an urgent need for developing cost-bounded, reasoning-robust guardrails to prevent shared infrastructures from being saturated or paralyzed in real-world agent deployments.
From Shield to Target: Denial-of-Service Attacks on LLM-Based Agent Guardrails
arXiv:2606.14517v2 Announce Type: replace-cross Abstract: LLM-based guardrails have emerged as a highly effective defense against prompt injection and jailbreak attacks in autonomous agents. However, we reveal that the very reasoning and task-following capabilities enabling this protection introduce a novel vulnerability: attackers can inject crafted data to trap the guardrail in extended reasoning loops, effectuating a systematic denial-of-service (DoS) attack. To systematically expose this threat, we design a beam-search optimization framework that crafts natural-language payloads to maximize guardrail reasoning length, utilizing an LLM proposer guided by a strategy bank. Based on the observation of guardrail's schema-following nature, we also provide another attack framework driven by mechanism-aware structural mutations with less computational load. The attack efficacy is systematically evaluated in two parts. First, in standalone evaluations, the attack generalizes across diverse guardrail architectures, safety templates, and agent benchmarks. Payloads optimized on a single open-source surrogate successfully transfer to eight leading model backbones (e.g., Claude, GPT, Gemini, DeepSeek, and Qwen), achieving a 13--63$\times$ token amplification. Second, in end-to-end real-world agent deployments (web, desktop, code, and multi-agent systems), the attack reveals up to a 148$\times$ latency amplification. We show that a single poisoned document can saturate shared guardrail infrastructures, effectively starving co-located agents and paralyzing the entire system. By uncovering this availability flaw, our work underscores the urgent need to develop cost-bounded, reasoning-robust guardrails.
Abstract:LLM-based guardrails have emerged as a highly effective defense against prompt injection and jailbreak attacks in autonomous agents. However, we reveal that the very reasoning and task-following capabilities enabling this protection introduce a novel vulnerability: attackers can inject crafted data to trap the guardrail in extended reasoning loops, effectuating a systematic denial-of-service (DoS) attack. To systematically expose this threat, we design a beam-search optimization framework that crafts natural-language payloads to maximize guardrail reasoning length, utilizing an LLM proposer guided by a strategy bank. Based on the observation of guardrail's schema-following nature, we also provide another attack framework driven by mechanism-aware structural mutations with less computational load. The attack efficacy is systematically evaluated in two parts. First, in standalone evaluations, the attack generalizes across diverse guardrail architectures, safety templates, and agent benchmarks. Payloads optimized on a single open-source surrogate successfully transfer to eight leading model backbones (e.g., Claude, GPT, Gemini, DeepSeek, and Qwen), achieving a 13--63$\times$ token amplification. Second, in end-to-end real-world agent deployments (web, desktop, code, and multi-agent systems), the attack reveals up to a 148$\times$ latency amplification. We show that a single poisoned document can saturate shared guardrail infrastructures, effectively starving co-located agents and paralyzing the entire system. By uncovering this availability flaw, our work underscores the urgent need to develop cost-bounded, reasoning-robust guardrails.
Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI)
arXiv:2606.14517 [cs.CR]
(or arXiv:2606.14517v2 [cs.CR] for this version)
https://doi.org/10.48550/arXiv.2606.14517 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 311 cuvinte
Subiecte conexe
From Shield to Target: Denial-of-Service Attacks on LLM-Based Agent Guardrails
arXiv · 15 iun. 2026
Weekly Cyber Security Newsletter Bulletin – Entra ID RCE, Claude Code Ransomware, T-Mobile Cable, Azure Credential Theft +20 Stories
Cybersecurity News · 23 aug. 2026
Hegseth AutoWarCom: who builds it, what the money buys, and why its hardest problem after Minab is the record, not the model
livetradingnews.com · 2 oct. 2026
News Archive 2026
Quantum Computing Report · 25 iul. 2026