Back to AI News
arxiv

ChronoPatch: Black-Box Energy-Latency Attacks on Vision-Language Models

Researchers present ChronoPatch, a black-box attack that prolongs vision-language model generation using only the final response. It combines response-length-guided evolutionary search for visual delay anchors with subspace-calibrated decision-based optimization to reduce perturbation size. Across four open-source VLMs, it reportedly raises inference latency by 47.86%–293.34% and GPU energy use by 41.44%–249.70% under a 5% spatial budget. On three commercial VLM APIs, response length rose 37.6%–81.5% and client-observed latency increased 1.82x–2.29x within 500-query budgets per API.

AI Brief

It exposes a practical availability and energy-cost risk for deployed generative VLM services and argues for generation-aware safeguards.

arXiv:2611.26117v1 Announce Type: cross Abstract: Vision-language models (VLMs) are increasingly deployed as generative inference services, where autoregressive decoding makes response length a key determinant of latency and computational cost. This creates an availability risk: adversarial visual inputs can prolong generation and increase serving overhead. Existing energy-latency attacks typically rely on gradients, logits, or token-level feedback, leaving open whether such risks persist when an attacker observes only the final response. We present ChronoPatch, a black-box attack that induces prolonged VLM generation requiring only the final response. It discovers visual delay anchors through response-length-guided evolutionary search, then minimizes perturbation magnitude through subspace-calibrated decision-based optimization. Across four open-source VLMs, ChronoPatch increases inference latency by 47.86%--293.34% and GPU energy consumption by 41.44%--249.70% under a 5% spatial budget. It also induces more repetitive outputs, as reflected by increased n-gram repetition ratios. On three commercial VLM APIs, ChronoPatch increases response length by 37.6%--81.5% and client-observed latency by 1.82x--2.29x under query budgets of 500 per API. These results reveal a generation-based availability risk that persists under final-response-only access and motivate generation-aware safeguards for deployed generative AI services.
Be the first to rate this article.

Comments & Questions (0)

Captcha
Please be respectful — let's keep the conversation friendly.

No comments yet

Be the first to comment!