2506.19533

Total: 1

#1 Identifying Physically Realizable Triggers for Backdoored Face Recognition Networks [PDF1] [Copy] [Kimi] [REL]

Authors: Ankita Raj, Ambar Pal, Chetan Arora

Backdoor attacks embed a hidden functionality into deep neural networks, causing the network to display anomalous behavior when activated by a predetermined pattern in the input Trigger, while behaving well otherwise on public test data. Recent works have shown that backdoored face recognition (FR) systems can respond to natural-looking triggers like a particular pair of sunglasses. Such attacks pose a serious threat to the applicability of FR systems in high-security applications. We propose a novel technique to (1) detect whether an FR network is compromised with a natural, physically realizable trigger, and (2) identify such triggers given a compromised network. We demonstrate the effectiveness of our methods with a compromised FR network, where we are able to identify the trigger (e.g., green sunglasses or red hat) with a top-5 accuracy of 74%, whereas a naive brute force baseline achieves 56% accuracy.

Subjects: Computer Vision and Pattern Recognition , Cryptography and Security , Machine Learning

Publish: 2025-06-24 11:36:27 UTC