aiqbench analysis
This research addresses a critical gap in deepfake detection by evaluating lightweight visual cues—specifically rPPG-derived waveforms and lip-region DCT coefficients—across seven synthetic video generators. The study reveals that while these signals offer complementary strengths for specific forgery methods, their transferability varies significantly depending on the generator architecture, challenging the assumption of universal efficacy for static fusion strategies. For practitioners building robust detection pipelines, the key takeaway is that relying on a single modality or naive concatenation may fail to generalize when facing novel deepfake generation techniques.
Complementary rPPG-Derived and Lip-Region Frequency Cues for Talking-Face Deepfake Detection
arXiv:2609.22284v1 Announce Type: cross Abstract: Talking-face (TF) deepfakes are detected unevenly by rPPG-based methods across generators. We study two lightweight visual-only cues, rPPG-derived waveforms extracted by RhythmFormer and lip-region discrete cosine transform (DCT) coefficients, on the seven TF methods of Celeb-DF++ under a subject-independent protocol. In-domain, lip-region DCT matches or exceeds the rPPG-derived 1D ResNet on every method except SadTalker, and Concat fusion reaches AUC 0.891 against 0.824 and 0.827 for the unimodal baselines. Under leave-one-generator-out evaluation the cues split: each transfers clearly better to three held-out methods, and IP-LAP is near chance for both. Concat averages 0.798 but falls below rPPG alone where DCT transfers poorly, so static fusion only partly exploits this complementarity. Lip-region DCT outperforms full-face DCT on six of seven methods. We treat the rPPG-derived signal as an empirical cue and do not claim it is cardiac in origin.
Abstract:Talking-face (TF) deepfakes are detected unevenly by rPPG-based methods across generators. We study two lightweight visual-only cues, rPPG-derived waveforms extracted by RhythmFormer and lip-region discrete cosine transform (DCT) coefficients, on the seven TF methods of Celeb-DF++ under a subject-independent protocol. In-domain, lip-region DCT matches or exceeds the rPPG-derived 1D ResNet on every method except SadTalker, and Concat fusion reaches AUC 0.891 against 0.824 and 0.827 for the unimodal baselines. Under leave-one-generator-out evaluation the cues split: each transfers clearly better to three held-out methods, and IP-LAP is near chance for both. Concat averages 0.798 but falls below rPPG alone where DCT transfers poorly, so static fusion only partly exploits this complementarity. Lip-region DCT outperforms full-face DCT on six of seven methods. We treat the rPPG-derived signal as an empirical cue and do not claim it is cardiac in origin.
Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR); Image and Video Processing (eess.IV)
arXiv:2609.22284 [cs.CV]
(or arXiv:2609.22284v1 [cs.CV] for this version)
https://doi.org/10.48550/arXiv.2609.22284 arXiv-issued DOI via DataCite (pending registration)
Full text under the publisher's open licence · 171 words
Related coverage
[AINews] Opus 5.5 is good at explainer videos
Latent Space · Sep 29, 2026
[AINews] The Future of Latent Space
Latent Space · Sep 25, 2026
[AINews] not much happened today
Latent Space · Sep 18, 2026
[AINews] Meta Muse Wonder & Wonder Full: MSL’s Claude 5.5-tier Frontier Model
Latent Space · Oct 13, 2026