The paper presents a new metric, t-EER, for the joint evaluation of Presentation Attack Detection (PAD) solutions and biometric verification. The t-EER is parameter-free and provides an equal false alarm and miss rates, regardless of the prevalence of attacks. This makes it a strong candidate metric for evaluating biometric systems under attack. The authors demonstrate its application using both simulated and real scores for a voice biometrics application.
Publication date: 22 Sep 2023
Project Page: https://github.com/TakHemlata/T-EER
Paper: https://arxiv.org/pdf/2309.12237
Leave a comment