analiza aiqbench
This research exposes a critical security vulnerability where inference optimization techniques inadvertently activate hidden backdoors in large language models, creating a novel attack surface distinct from standard model poisoning. The study demonstrates that numerical inconsistencies introduced during optimization can bypass existing detection pipelines while maintaining high clean accuracy, posing significant risks for deployed systems relying on latency-reduction tools. To address this, the authors propose specific defense mechanisms that effectively neutralize these triggers without compromising model performance or integrity.
Trusted Weights, Treacherous Optimizations? Optimization-Triggered Backdoor Attacks on LLMs
arXiv:2605.20641v2 Announce Type: replace-cross Abstract: Inference optimization aims to minimize the latency and resource consumption of LLM inference while preserving output quality, making large-scale deployment practical and cost-effective. However, optimized execution can introduce small numerical inconsistencies from the original model. We reveal that this inconsistency not only causes the model's outputs to diverge, but more critically can introduce hidden backdoors. The backdoor remains dormant under standard unoptimized execution and is activated only when inference optimization is enabled, allowing it to evade existing backdoor detection pipelines. We first introduce the Input-Specific Optimization Backdoor (ISOB) to demonstrate that optimization-induced differences can cause wrong predictions. However, ISOB remains input-specific and cannot establish a universal optimization-triggered backdoor. To overcome this limitation, we design the Universal Optimization Backdoor (UOB). The backdoored model stays benign under unoptimized execution but activates when inference optimization is enabled. We conduct extensive experiments across seven mainstream open-source LLMs, four tasks, and three optimization backends. UOB reaches up to 100\% attack success while largely preserving clean accuracy. To mitigate this vulnerability, we design three defense methods that reduce the backdoor ASR to at most 0.02 while preserving clean accuracy. These results reveal inference optimization as a new LLM security attack surface and motivate defenses against test-deployment disagreement.
Title:Trusted Weights, Treacherous Optimizations? Optimization-Triggered Backdoor Attacks on LLMs
Abstract:Inference optimization aims to minimize the latency and resource consumption of LLM inference while preserving output quality, making large-scale deployment practical and cost-effective. However, optimized execution can introduce small numerical inconsistencies from the original model. We reveal that this inconsistency not only causes the model's outputs to diverge, but more critically can introduce hidden backdoors. The backdoor remains dormant under standard unoptimized execution and is activated only when inference optimization is enabled, allowing it to evade existing backdoor detection pipelines. We first introduce the Input-Specific Optimization Backdoor (ISOB) to demonstrate that optimization-induced differences can cause wrong predictions. However, ISOB remains input-specific and cannot establish a universal optimization-triggered backdoor. To overcome this limitation, we design the Universal Optimization Backdoor (UOB). The backdoored model stays benign under unoptimized execution but activates when inference optimization is enabled. We conduct extensive experiments across seven mainstream open-source LLMs, four tasks, and three optimization backends. UOB reaches up to 100\% attack success while largely preserving clean accuracy. To mitigate this vulnerability, we design three defense methods that reduce the backdoor ASR to at most 0.02 while preserving clean accuracy. These results reveal inference optimization as a new LLM security attack surface and motivate defenses against test-deployment disagreement.
Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
arXiv:2605.20641 [cs.CR]
(or arXiv:2605.20641v2 [cs.CR] for this version)
https://doi.org/10.48550/arXiv.2605.20641 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 · 309 cuvinte
Subiecte conexe
ToxScreen: Detecting Whether an LLM Has Been Poisoned
arXiv · 30 iul. 2026
Concrete Evaluations to Investigate the OpenAI Model That Hacked Hugging Face
AI Alignment Forum · 3 aug. 2026
Informe de riesgos de la IA de frontera (febrero–marzo de 2026)
METR · 19 mai 2026
Frontier AI Risk Report (February–March 2026)
METR · 19 mai 2026