Adversarial Video Promotion Against Text-to-Video Retrieval

2025-08-09

AI Summary

Existing text-to-video retrieval (T2VR) systems lack rigorous evaluation of adversarial robustness, particularly against “malicious video ranking promotion” attacks. This work introduces the first adversarial video promotion attack tailored to T2VR, generating highly transferable and perceptually imperceptible perturbations under white-box, gray-box, and black-box settings. Our method leverages fine-grained cross-modal interaction enhancement and optimized alignment in the joint embedding space. We further propose Modality-Refined (MoRe) prompting—a novel strategy enabling synchronized multi-query attacks. Extensive experiments across three state-of-the-art T2VR models and benchmark datasets demonstrate average rank improvements of 30% (white-box), 10% (gray-box), and 4% (black-box) for targeted videos—substantially outperforming existing baselines. These results expose a critical security vulnerability in real-world cross-modal retrieval systems, highlighting urgent needs for robustness-aware design and evaluation.

Technology Category

• Computer Vision: Adversarial Attacks & Robustness
• Machine Learning: Adversarial Learning & Robustness
• Multiagent Systems: Adversarial Agents

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Security and Privacy: Large-scale security measurements

Abstract

Thanks to the development of cross-modal models, text-to-video retrieval (T2VR) is advancing rapidly, but its robustness remains largely unexamined. Existing attacks against T2VR are designed to push videos away from queries, i.e., suppressing the ranks of videos, while the attacks that pull videos towards selected queries, i.e., promoting the ranks of videos, remain largely unexplored. These attacks can be more impactful as attackers may gain more views/clicks for financial benefits and widespread (mis)information. To this end, we pioneer the first attack against T2VR to promote videos adversarially, dubbed the Video Promotion attack (ViPro). We further propose Modal Refinement (MoRe) to capture the finer-grained, intricate interaction between visual and textual modalities to enhance black-box transferability. Comprehensive experiments cover 2 existing baselines, 3 leading T2VR models, 3 prevailing datasets with over 10k videos, evaluated under 3 scenarios. All experiments are conducted in a multi-target setting to reflect realistic scenarios where attackers seek to promote the video regarding multiple queries simultaneously. We also evaluated our attacks for defences and imperceptibility. Overall, ViPro surpasses other baselines by over $30/10/4%$ for white/grey/black-box settings on average. Our work highlights an overlooked vulnerability, provides a qualitative analysis on the upper/lower bound of our attacks, and offers insights into potential counterplays. Code will be publicly available at https://github.com/michaeltian108/ViPro.

Problem

• Adversarial attacks promoting videos in text-to-video retrieval
• Enhancing black-box transferability via modal refinement
• Evaluating attacks across multiple models and datasets

Innovation

• Adversarial video promotion attack (ViPro)
• Modal Refinement (MoRe) for transferability
• Multi-target promotion across diverse queries