Generation of Deepfake Video Dataset

Designed and initiated the creation of a custom dataset that includes both authentic and manipulated videos generated using FaceSwap, GAN, and diffusion-based techniques. Applied varied video compression methods to mimic real-world conditions and make detection more challenging. Benchmarked the dataset against state-of-the-art Deepfake detection models to identify strengths, weaknesses, and opportunities for improvement. The project aims to provide a robust benchmark resource to support cybersecurity research and safeguard online media integrity.

Development and analysis of a challenging Deepfake dataset for advancing detection research.