首页 > AI前沿 > Pulse or Provenance? What rPPG-Based Deepfake Detectors Learn

Pulse or Provenance? What rPPG-Based Deepfake Detectors Learn

arXiv机器学习 2026-11-10 20:36 17 阅读 查看原文

Remote photoplethysmography (rPPG) has been proposed as a physiological cue for deepfake detection, based on the premise that synthesized faces lack a genuine cardiac signal.

We examine what rPPG-style detectors learn on Celeb-DF++ (22 forgery methods) and FaceForensics++ (FF++), using one identically sampled window per real and fake video.

Chrominance traces and discriminative signal

On talking-face forgeries, simple chrominance traces reach an AUC of 0.998, exceeding a dedicated rPPG extractor (0.87).

Controlled experiments indicate that this signal is not physiological in origin.

Fully synthesized forgeries and signal localization

On fully synthesized forgeries, it is equally discriminative when extracted from background regions or from full frames without face detection, and it is concentrated above the 0.7 to 4 Hz heart-rate band.

Frame order, normalization, and reproducibility

It is largely removed by shuffling frame order but persists after normalizing resolution, frame rate and encoding, and an independent reimplementation reproduces these findings.

Manipulating high-frequency content

Removing high-frequency content makes most synthesized videos appear real, adding it makes real videos appear fake, and injecting a synthetic pulse has little effect.

Self-reenactment experiments

In controlled self-reenactment experiments with two generators, animating a face from a single frame introduces the same high-frequency signal, whereas reconstructing each frame from its own source does not.

Cross-dataset behavior and source encoding

Face-based chrominance detectors trained on Celeb-DF++ classify over 99% of real FF++ videos as fake, indicating that they largely encode video source rather than manipulation, and a CLIP-based detector exhibits the same behavior.

CLIP-based detection and feature fusion

A frozen CLIP backbone matches or exceeds fusions of CLIP with rPPG and frequency features.

Data splits and evaluation protocols

Finally, we show that the official Celeb-DF and Celeb-DF++ splits share identities and source-video content between training and test, and that custom protocols reusing test videos and identities in training inflate weaker detectors by up to 0.30 AUC.