Vid-Freeze: Protecting Images from Malicious Image-to-Video Generation via Temporal Freezing

2025-09-27

AI Summary

To mitigate the security risk of image-to-video (I2V) generative models being maliciously exploited to synthesize deceptive videos, this paper proposes “Temporal Freezing”—a defense mechanism that injects imperceptible, image-specific adversarial perturbations into static input images to selectively suppress spatiotemporal attention mechanisms within I2V models. This targeted suppression disrupts motion modeling capability, forcing the model to generate entirely static or near-static video outputs. Unlike prior defenses, our approach pioneers attention inhibition as the core pathway for blocking motion synthesis, preserving the semantic fidelity of the original image while enabling precise, input-aware intervention in the video generation process. Extensive experiments across multiple state-of-the-art I2V models demonstrate that Temporal Freezing significantly outperforms existing methods, achieving over 92% success rate in generating static videos. The method establishes a deployable paradigm for protecting static image copyright and preventing I2V misuse.

Technology Category

Computer Vision: Adversarial Attacks & RobustnessMachine Learning: Adversarial Learning & RobustnessNatural Language Processing: Generation

Application Category

User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systemsGraph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphsSecurity and Privacy: Large-scale security measurements

Abstract

The rapid progress of image-to-video (I2V) generation models has introduced significant risks, enabling video synthesis from static images and facilitating deceptive or malicious content creation. While prior defenses such as I2VGuard attempt to immunize images, effective and principled protection to block motion remains underexplored. In this work, we introduce Vid-Freeze - a novel attention-suppressing adversarial attack that adds carefully crafted adversarial perturbations to images. Our method explicitly targets the attention mechanism of I2V models, completely disrupting motion synthesis while preserving semantic fidelity of the input image. The resulting immunized images generate stand-still or near-static videos, effectively blocking malicious content creation. Our experiments demonstrate the impressive protection provided by the proposed approach, highlighting the importance of attention attacks as a promising direction for robust and proactive defenses against misuse of I2V generation models.

Problem

• Preventing malicious video generation from static images
• Disrupting motion synthesis while preserving image semantics
• Protecting images via attention-suppressing adversarial attacks

Innovation

• Adversarial perturbations disrupt motion synthesis
• Targets attention mechanism in I2V models
• Generates static videos while preserving image semantics