DepthPulse+: A Depth and Vital Sign Based Method for Face Presentation Attack Detection

Nafiz Sadman, Furkan Alaca, Farhana Zulkernine.

IEEE International Conference on Communications 2026, May 24-28, Glasgow, Scotland, UK.

Face Presentation Attacks (FPA) are a widespread threat against face authentication systems. FPA are performed by presenting face representations of the authorized user to the camera for authentication, e.g., by displaying printed photos (photo attacks), displaying images or videos on a screen (video attacks), or wearing a face mask (mask attacks). In this paper, we propose an ensemble method that combines depth-based and vital-sign based (using remote photoplethysmography) FPA detection (FPAD). DepthPulse+ achieves state-of-the-art attack classification error rate (ACER) on all four evaluation protocols of the widely-used OULU-NPU benchmark dataset (achieving reductions of 35%, 44%, 84%, and 10% in Protocols 1 to 4, respectively, compared to the benchmark method), highlighting improved generalizability to conditions not encountered in training.