Mask-based Invisible Backdoor Attacks on Object Detection

TL;DR

This paper introduces a novel invisible backdoor attack on object detection models using a mask-based method, demonstrating its effectiveness across multiple attack scenarios and evaluating potential defenses.

Contribution

It presents the first mask-based invisible backdoor attack tailored for object detection, expanding backdoor attack research beyond image classification.

Findings

Effective attack success across scenarios

Partial success of defense methods

Code availability for reproducibility

Abstract

Deep learning models have achieved unprecedented performance in the domain of object detection, resulting in breakthroughs in areas such as autonomous driving and security. However, deep learning models are vulnerable to backdoor attacks. These attacks prompt models to behave similarly to standard models without a trigger; however, they act maliciously upon detecting a predefined trigger. Despite extensive research on backdoor attacks in image classification, their application to object detection remains relatively underexplored. Given the widespread application of object detection in critical real-world scenarios, the sensitivity and potential impact of these vulnerabilities cannot be overstated. In this study, we propose an effective invisible backdoor attack on object detection utilizing a mask-based approach. Three distinct attack scenarios were explored for object detection: object disappearance, object misclassification, and object generation attack. Through extensive experiments, we comprehensively examined the effectiveness of these attacks and tested certain defense methods to determine effective countermeasures. Code will be available at https://github.com/jeongjin0/invisible-backdoor-object-detection