Research

Paper

AI LLM March 06, 2026

When One Modality Rules Them All: Backdoor Modality Collapse in Multimodal Diffusion Models

Authors

Qitong Wang, Haoran Dai, Haotian Zhang, Christopher Rasmussen, Binghui Wang

Abstract

While diffusion models have revolutionized visual content generation, their rapid adoption has underscored the critical need to investigate vulnerabilities, e.g., to backdoor attacks. In multimodal diffusion models, it is natural to expect that attacking multiple modalities simultaneously (e.g., text and image) would yield complementary effects and strengthen the overall backdoor. In this paper, we challenge this assumption by investigating the phenomenon of Backdoor Modality Collapse, a scenario where the backdoor mechanism degenerates to rely predominantly on a subset of modalities, rendering others redundant. To rigorously quantify this behavior, we introduce two novel metrics: Trigger Modality Attribution (TMA) and Cross-Trigger Interaction (CTI). Through extensive experiments across diverse training configurations in multimodal conditional diffusion, we consistently observe a ``winner-takes-all'' dynamic in backdoor behavior. Our results reveal that (1) attacks often collapse into subset-modality dominance, and (2) cross-modal interaction is negligible or even negative, contradicting the intuition of synergistic vulnerability. These findings highlight a critical blind spot in current assessments, suggesting that high attack success rates often mask a fundamental reliance on a subset of modalities. This establishes a principled foundation for mechanistic analysis and future defense development.

Metadata

arXiv ID: 2603.06508
Provider: ARXIV
Primary Category: cs.LG
Published: 2026-03-06
Fetched: 2026-03-09 06:05

Related papers

Raw Data (Debug)
{
  "raw_xml": "<entry>\n    <id>http://arxiv.org/abs/2603.06508v1</id>\n    <title>When One Modality Rules Them All: Backdoor Modality Collapse in Multimodal Diffusion Models</title>\n    <updated>2026-03-06T17:42:08Z</updated>\n    <link href='https://arxiv.org/abs/2603.06508v1' rel='alternate' type='text/html'/>\n    <link href='https://arxiv.org/pdf/2603.06508v1' rel='related' title='pdf' type='application/pdf'/>\n    <summary>While diffusion models have revolutionized visual content generation, their rapid adoption has underscored the critical need to investigate vulnerabilities, e.g., to backdoor attacks. In multimodal diffusion models, it is natural to expect that attacking multiple modalities simultaneously (e.g., text and image) would yield complementary effects and strengthen the overall backdoor. In this paper, we challenge this assumption by investigating the phenomenon of Backdoor Modality Collapse, a scenario where the backdoor mechanism degenerates to rely predominantly on a subset of modalities, rendering others redundant. To rigorously quantify this behavior, we introduce two novel metrics: Trigger Modality Attribution (TMA) and Cross-Trigger Interaction (CTI). Through extensive experiments across diverse training configurations in multimodal conditional diffusion, we consistently observe a ``winner-takes-all'' dynamic in backdoor behavior. Our results reveal that (1) attacks often collapse into subset-modality dominance, and (2) cross-modal interaction is negligible or even negative, contradicting the intuition of synergistic vulnerability. These findings highlight a critical blind spot in current assessments, suggesting that high attack success rates often mask a fundamental reliance on a subset of modalities. This establishes a principled foundation for mechanistic analysis and future defense development.</summary>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.LG'/>\n    <published>2026-03-06T17:42:08Z</published>\n    <arxiv:comment>Accepted to the ICLR 2026 Workshop on Principled Design for Trustworthy AI. The first two authors contributed equally</arxiv:comment>\n    <arxiv:primary_category term='cs.LG'/>\n    <author>\n      <name>Qitong Wang</name>\n    </author>\n    <author>\n      <name>Haoran Dai</name>\n    </author>\n    <author>\n      <name>Haotian Zhang</name>\n    </author>\n    <author>\n      <name>Christopher Rasmussen</name>\n    </author>\n    <author>\n      <name>Binghui Wang</name>\n    </author>\n  </entry>"
}