Backdoor Mitigation in Object Detection via Adversarial Fine-Tuning

TL;DR

This paper presents a detection-aware adversarial fine-tuning method to mitigate backdoor attacks in object detection models, effectively reducing attack success while maintaining detection performance.

Contribution

It introduces a novel framework with soft-branch minimisation and dual-objective loss for effective backdoor mitigation in object detection without attack objective knowledge.

Findings

More effective reduction of attack success compared to baselines.

Preserves true detection performance on clean data.

Applicable to CNN- and Transformer-based detectors.

Abstract

Backdoor attacks can implant malicious behaviours into deep models while preserving performance on clean data, posing a serious threat to safety-critical vision systems. Although backdoor mitigation has been studied extensively for image classification, defenses for object detection remain comparatively underdeveloped. Adversarial fine-tuning is a common backdoor mitigation approach in classification, but adapting it to detection is nontrivial as classification-oriented adversarial generation does not match the detection attack space, where attacks may cause object misclassification or disappearance, and standard detection losses can dilute the repair signal across many predictions. We address these challenges through a detection-aware adversarial fine-tuning framework for mitigating object-detection backdoors when the defender has access only to a compromised detector and a small clean dataset, without knowing the attack objective. For adversarial generation that does not require knowledge of the attack objective, we introduce soft-branch minimisation, which uses a soft gate to combine objectives aligned with misclassification and disappearance attacks, together with a detection-aware classification-loss maximisation. For targeted repair, we introduce a dual-objective fine-tuning loss applied to target-matched predictions, concentrating the defensive update on predictions most relevant to the backdoor behaviour. Experiments across CNN- and Transformer-based detectors show that our approach more effectively reduces attack success while preserving true detections, compared with classification-oriented baselines, and maintains competitive clean detection performance.