Temporal-Distributed Backdoor Attack Against Video Based Action Recognition

TL;DR

This paper introduces a novel, imperceptible, temporally distributed backdoor attack on video action recognition models that remains effective against current defenses, demonstrated through extensive experiments on multiple benchmarks.

Contribution

It proposes a new backdoor attack method for video data using transformed domain perturbations, which is more resilient and less detectable than existing approaches.

Findings

The attack is effective on various models and benchmarks.

It remains resilient against existing defensive strategies.

The study reveals an effect called 'collateral damage' impacting attack dynamics.

Abstract

Deep neural networks (DNNs) have achieved tremendous success in various applications including video action recognition, yet remain vulnerable to backdoor attacks (Trojans). The backdoor-compromised model will mis-classify to the target class chosen by the attacker when a test instance (from a non-target class) is embedded with a specific trigger, while maintaining high accuracy on attack-free instances. Although there are extensive studies on backdoor attacks against image data, the susceptibility of video-based systems under backdoor attacks remains largely unexplored. Current studies are direct extensions of approaches proposed for image data, e.g., the triggers are independently embedded within the frames, which tend to be detectable by existing defenses. In this paper, we introduce a simple yet effective backdoor attack against video data. Our proposed attack, adding perturbations in a transformed domain, plants an imperceptible, temporally distributed trigger across the video frames, and is shown to be resilient to existing defensive strategies. The effectiveness of the proposed attack is demonstrated by extensive experiments with various well-known models on two video recognition benchmarks, UCF101 and HMDB51, and a sign language recognition benchmark, Greek Sign Language (GSL) dataset. We delve into the impact of several influential factors on our proposed attack and identify an intriguing effect termed "collateral damage" through extensive studies.