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.

Official code for FAME: A lightweight spatio-temporal framework for Deepfake Model Attribution using VGG-based feature extraction and temporal attention. Outperforms prior methods across DFDM, FF++, and FakeAVCeleb datasets.

This is the public version of my master's research work.