Watch, Listen, Understand, Mislead: Tri-modal Adversarial Attacks on Short Videos for Content Appropriateness Evaluation

Mohammad Islam, Jishanul Islam, Sahid Mustakim, Sadia Ahmmed, Syed Tasdid, Azam Dhrubo, Swakkhar Shatabda, Ummay Muna, Montasir Chowdhury, Tashfia Sikder

arXiv · PDF · Open in the Atlas

Code that ran

We lifted 1 functions out of this paper's own repositories and ran 0 of them in a sandbox. "Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results.

Repositories linked to this paper

• sahidmustakim/ChimeraBreak

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Multimodal Large Language Models (MLLMs) are increasingly used for content moderation, yet their robustness in short-form video contexts remains underexplored. Current safety evaluations often rely on unimodal attacks, failing to address combined attack vulnerabilities. In this paper, we introduce a comprehensive framework for evaluating the tri-modal safety of MLLMs. First, we present the Short-Video Multimodal Adversarial (SVMA) dataset, comprising diverse short-form videos with human-guided synthetic adversarial attacks. Second, we propose ChimeraBreak, a novel tri-modal attack strategy that simultaneously challenges visual, auditory, and semantic reasoning pathways. * Equal Contributions Extensive experiments on state-of-the-art MLLMs reveal significant vulnerabilities with high Attack Success Rates (ASR). Our findings uncover distinct failure modes, showing model biases toward misclassifying benign or policyviolating content. We assess results using LLM-as-a-judge, demonstrating attack reasoning efficacy. Our dataset and findings provide crucial insights for developing more robust and safe MLLMs. Warning: This paper contains content that may be offensive to readers.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2507.11968")
get_code_for_paper("2507.11968")
have("2507.11968")

Connect an agent — have() is free.