Immune2V: Image Immunization Against Dual-Stream Image-to-Video Generation

TL;DR

This paper introduces Immune2V, a defense framework against dual-stream image-to-video generation deepfakes, by preventing adversarial noise dilution and overriding of immunization through novel encoding and alignment strategies.

Contribution

The paper presents a novel defense method, Immune2V, that enhances robustness of image-to-video models against adversarial attacks by maintaining signal integrity and counteracting text-guided overrides.

Findings

Immune2V significantly outperforms baseline defenses in degrading deepfake quality.

The framework maintains persistent adversarial effects under imperceptibility constraints.

Extensive experiments validate the effectiveness of Immune2V against state-of-the-art models.

Abstract

Image-to-video (I2V) generation has the potential for societal harm because it enables the unauthorized animation of static images to create realistic deepfakes. While existing defenses effectively protect against static image manipulation, extending these to I2V generation remains underexplored and non-trivial. In this paper, we systematically analyze why modern I2V models are highly robust against naive image-level adversarial attacks (i.e., immunization). We observe that the video encoding process rapidly dilutes the adversarial noise across future frames, and the continuous text-conditioned guidance actively overrides the intended disruptive effect of the immunization. Building on these findings, we propose the Immune2V framework which enforces temporally balanced latent divergence at the encoder level to prevent signal dilution, and aligns intermediate generative representations with a precomputed collapse-inducing trajectory to counteract the text-guidance override. Extensive experiments demonstrate that Immune2V produces substantially stronger and more persistent degradation than adapted image-level baselines under the same imperceptibility budget.