TEAR: Temporal-aware Automated Red-teaming for Text-to-Video Models

TL;DR

TEAR is a novel framework that systematically uncovers safety risks in text-to-video models by exploiting their temporal dynamics, significantly improving attack success rates over previous methods.

Contribution

The paper introduces TEAR, the first temporal-aware automated red-teaming framework specifically designed for evaluating safety in text-to-video models.

Findings

Achieves over 80% attack success rate on various T2V systems.

Outperforms prior methods with a 23% improvement in attack success rate.

Demonstrates effectiveness on both open-source and commercial T2V models.

Abstract

Text-to-Video (T2V) models are capable of synthesizing high-quality, temporally coherent dynamic video content, but the diverse generation also inherently introduces critical safety challenges. Existing safety evaluation methods,which focus on static image and text generation, are insufficient to capture the complex temporal dynamics in video generation. To address this, we propose a TEmporal-aware Automated Red-teaming framework, named TEAR, an automated framework designed to uncover safety risks specifically linked to the dynamic temporal sequencing of T2V models. TEAR employs a temporal-aware test generator optimized via a two-stage approach: initial generator training and temporal-aware online preference learning, to craft textually innocuous prompts that exploit temporal dynamics to elicit policy-violating video output. And a refine model is adopted to improve the prompt stealthiness and adversarial effectiveness cyclically. Extensive experimental evaluation demonstrates the effectiveness of TEAR across open-source and commercial T2V systems with over 80% attack success rate, a significant boost from prior best result of 57%.