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

2026-03-06

AI Summary

This study investigates whether multimodal diffusion models genuinely leverage all modalities in backdoor attacks. The authors uncover a prevalent phenomenon termed “backdoor modality collapse,” wherein attacks often degenerate into reliance on only one or a few modalities. To quantify modality contributions and interactions, they introduce two novel metrics: Trigger Modality Attribution (TMA) and Cross-Trigger Interaction (CTI). These reveal counterintuitive behaviors in multimodal backdoor attacks, including insufficient inter-modal collaboration and even detrimental interference among modalities. Extensive experiments across diverse training configurations demonstrate that a “winner-takes-all” dynamic is widespread, with high attack success rates frequently masking excessive dependence on specific modalities. This exposes a critical blind spot in current security evaluations of multimodal systems.

Technology Category

• Machine Learning: Multimodal Learning
• Computer Vision: Multi-modal Vision
• Natural Language Processing: Language Grounding & Multi-modal NLP

Application Category

• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Web Mining and Content Analysis: Mining multimedia, multimodal, multilingual, cross-lingual Web data

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.

Problem

• Backdoor Modality Collapse
• Multimodal Diffusion Models
• Trigger Modality Attribution
• Cross-Trigger Interaction
• Backdoor Attacks

Innovation

• Backdoor Modality Collapse
• Multimodal Diffusion Models
• Trigger Modality Attribution
• Cross-Trigger Interaction
• Adversarial Vulnerability