Survivability of Backdoor Attacks on Unconstrained Face Recognition Systems
This addresses security concerns for stakeholders deploying face recognition technology, providing a comprehensive system-level analysis and countermeasures.
The paper tackles the vulnerability of real-world face recognition systems to backdoor attacks, demonstrating that a single backdoor can compromise entire systems across various configurations and scenarios.
The widespread deployment of Deep Learning-based Face Recognition Systems raises multiple security concerns. While prior research has identified backdoor vulnerabilities on isolated components, Backdoor Attacks on real-world, unconstrained pipelines remain underexplored. This paper presents the first comprehensive system-level analysis of Backdoor Attacks targeting Face Recognition Systems and provides three contributions. We first show that face feature extractors trained with large margin metric learning losses are susceptible to Backdoor Attacks. By analyzing 20 pipeline configurations and 15 attack scenarios, we then reveal that a single backdoor can compromise an entire Face Recognition System. Finally, we propose effective best practices and countermeasures for stakeholders.