This paper proposes a study to overcome the limitation of existing models that simply output the result of "fake face" in the field of Face Anti-Spoofing (FAS), which is becoming increasingly important due to the increasing use of face recognition technology in daily life. It points out the problem that although existing FAS models achieve high accuracy, they are unreliable and cause user confusion because they cannot explain the reason. Therefore, this paper defines a new problem called Explainable Face Anti-Spoofing (X-FAS) by integrating Explainable Artificial Intelligence (XAI) into FAS, and proposes SPoof Trace Discovery (SPTD), a X-FAS method that discovers spoofing traces and provides a reliable explanation. Furthermore, we present the X-FAS benchmark, which includes expert-annotated spoofing traces to evaluate the quality of the X-FAS method, and experimentally demonstrate the reliable explanation generation ability of SPTD by analyzing the explanations of SPTD and comparing them quantitatively and qualitatively with existing XAI methods.
