Deepfake detection improves with explicit forensic video features
Interpretable Deepfake Detection in Videos via Explicit Forensic Features and Temporal Modeling
Computer Vision and Pattern Recognition
Summary
Detecting videos altered by deepfake technology is hard because fake faces can look real in single frames but show small inconsistencies over time. The authors present a method that watches how facial features change across frames using clear and understandable clues, like lighting, texture, shape, and video compression artifacts. Their approach helps detect deepfakes better across different video collections and explains why the detection works. It uses a special type of neural network to spot subtle timing problems that reveal if the video is fake.
What this means in practice
- •For video security teams: Integrate interpretable forensic video features to detect manipulated videos with strong accuracy across diverse sources.
- •For law enforcement agencies: Use explicit spatial and temporal facial features to identify deepfake videos for digital evidence verification.
Authors
Chahira Benhama, Mohand Saïd Allili, Assia Hamadene
Abstract
Deepfake detection in videos remains challenging, as manipulated content may appear visually consistent at the frame level while exhibiting subtle temporal inconsistencies. This paper introduces an interpretable deepfake detection framework that models spatially and temporally coherent facial features in video sequences. Unlike end-to-end deep models relying on implicit representations, the proposed approach explicitly encodes physically grounded forensic cues, enabling transparent analysis and improved multi-dataset generalization. The pipeline transforms videos into identity-consistent facial trajectories, segments them into fixed-length temporal windows, and represents each frame using 68 structured descriptors spanning four complementary domains: photometric, textural, geometric, and compression-based features. These descriptors provide a compact multi-domain representation of manipulation artifacts and are processed by a Long Short-Term Memory (LSTM) network to capture temporal dependencies and subtle irregularities. Evaluation on four benchmark datasets, FaceForensics++, Celeb-DF v2, a curated subset of the DeepFake Detection Challenge (DFDC), and DeeperForensics, yields strong and consistent F1-scores of 98.0%, 91.0%, 97.6%, and 96.2%, respectively. The approach also demonstrated a good cross-dataset generalization, providing a robust and interpretable solution for video deepfake detection.