DeepFakesON-Phys: DeepFakes Detection based on Heart Rate Estimation

TL;DR

This paper presents DeepFakesON-Phys, a novel DeepFake detection method leveraging heart rate signals from remote photoplethysmography, achieving over 98% AUC and outperforming existing techniques on public datasets.

Contribution

The work introduces a new DeepFake detection framework based on physiological signals, specifically heart rate, combined with a Convolutional Attention Network for improved accuracy.

Findings

Achieved over 98% AUC on Celeb-DF and DFDC datasets.

Outperformed state-of-the-art DeepFake detection methods.

Validated the effectiveness of physiological signals for DeepFake detection.

Abstract

This work introduces a novel DeepFake detection framework based on physiological measurement. In particular, we consider information related to the heart rate using remote photoplethysmography (rPPG). rPPG methods analyze video sequences looking for subtle color changes in the human skin, revealing the presence of human blood under the tissues. In this work we investigate to what extent rPPG is useful for the detection of DeepFake videos. The proposed fake detector named DeepFakesON-Phys uses a Convolutional Attention Network (CAN), which extracts spatial and temporal information from video frames, analyzing and combining both sources to better detect fake videos. This detection approach has been experimentally evaluated using the latest public databases in the field: Celeb-DF and DFDC. The results achieved, above 98% AUC (Area Under the Curve) on both databases, outperform the state of the art and prove the success of fake detectors based on physiological measurement to detect the latest DeepFake videos.