multimedia-forensics

Structure matters: analyzing videos via graph neural networks for social media platform attribution

Official implementation of the paper ``DeepFeatureX-SN: Generalization of deepfake detection via contrastive learning'', accepted for publication at the Springer journal 'Multimedia Tools and Application'.

Official code for CapST (ACM TOMM 2025): A Capsule Network and Temporal Attention-based framework for Deepfake Model Attribution. Achieves efficient and accurate source identification across DFDM and GAN-generated datasets.

A unified multi-layer verification engine that detects real human presence and validates physical, biological, and temporal consistency in digital media to produce an explainable trust score

Closed-Set SDI Using log-Mel Spectrograms from Videos (IEEE Access)

Contribution: This repository extends the original CLIP implementation by adding training code for an SVM classifier using CLIP-extracted features, with detailed setup instructions in README_TRAIN.md.