Research

Paper

TESTING March 02, 2026

VidDoS: Universal Denial-of-Service Attack on Video-based Large Language Models

Authors

Duoxun Tang, Dasen Dai, Jiyao Wang, Xiao Yang, Jianyu Wang, Siqi Cai

Abstract

Video-LLMs are increasingly deployed in safety-critical applications but are vulnerable to Energy-Latency Attacks (ELAs) that exhaust computational resources. Current image-centric methods fail because temporal aggregation mechanisms dilute individual frame perturbations. Additionally, real-time demands make instance-wise optimization impractical for continuous video streams. We introduce VidDoS, which is the first universal ELA framework tailored for Video-LLMs. Our method leverages universal optimization to create instance-agnostic triggers that require no inference-time gradient calculation. We achieve this through $\textit{masked teacher forcing}$ to steer models toward expensive target sequences, combined with a $\textit{refusal penalty}$ and $\textit{early-termination suppression}$ to override conciseness priors. Testing across three mainstream Video-LLMs and three video datasets, which include video question answering and autonomous driving scenarios, shows extreme degradation. VidDoS induces a token expansion of more than 205$\times$ and inflates the inference latency by more than 15$\times$ relative to clean baselines. Simulations of real-time autonomous driving streams further reveal that this induced latency leads to critical safety violations. We urge the community to recognize and mitigate these high-hazard ELA in Video-LLMs.

Metadata

arXiv ID: 2603.01454
Provider: ARXIV
Primary Category: cs.CV
Published: 2026-03-02
Fetched: 2026-03-03 04:34

Related papers

Raw Data (Debug)
{
  "raw_xml": "<entry>\n    <id>http://arxiv.org/abs/2603.01454v1</id>\n    <title>VidDoS: Universal Denial-of-Service Attack on Video-based Large Language Models</title>\n    <updated>2026-03-02T05:11:47Z</updated>\n    <link href='https://arxiv.org/abs/2603.01454v1' rel='alternate' type='text/html'/>\n    <link href='https://arxiv.org/pdf/2603.01454v1' rel='related' title='pdf' type='application/pdf'/>\n    <summary>Video-LLMs are increasingly deployed in safety-critical applications but are vulnerable to Energy-Latency Attacks (ELAs) that exhaust computational resources. Current image-centric methods fail because temporal aggregation mechanisms dilute individual frame perturbations. Additionally, real-time demands make instance-wise optimization impractical for continuous video streams. We introduce VidDoS, which is the first universal ELA framework tailored for Video-LLMs. Our method leverages universal optimization to create instance-agnostic triggers that require no inference-time gradient calculation. We achieve this through $\\textit{masked teacher forcing}$ to steer models toward expensive target sequences, combined with a $\\textit{refusal penalty}$ and $\\textit{early-termination suppression}$ to override conciseness priors. Testing across three mainstream Video-LLMs and three video datasets, which include video question answering and autonomous driving scenarios, shows extreme degradation. VidDoS induces a token expansion of more than 205$\\times$ and inflates the inference latency by more than 15$\\times$ relative to clean baselines. Simulations of real-time autonomous driving streams further reveal that this induced latency leads to critical safety violations. We urge the community to recognize and mitigate these high-hazard ELA in Video-LLMs.</summary>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.CV'/>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.AI'/>\n    <published>2026-03-02T05:11:47Z</published>\n    <arxiv:primary_category term='cs.CV'/>\n    <author>\n      <name>Duoxun Tang</name>\n    </author>\n    <author>\n      <name>Dasen Dai</name>\n    </author>\n    <author>\n      <name>Jiyao Wang</name>\n    </author>\n    <author>\n      <name>Xiao Yang</name>\n    </author>\n    <author>\n      <name>Jianyu Wang</name>\n    </author>\n    <author>\n      <name>Siqi Cai</name>\n    </author>\n  </entry>"
}