Undercover Deepfakes: Detecting Fake Segments in Videos

• DOI: 10.48550/arxiv.2305.06564
• Publication Date: 2023-01-01
• Authors: Sanjay Saha; Rashindrie Perera; Sachith Seneviratne; Tamasha Malepathi...; Sanka Rasnayaka; Deshani Geethika; Terence Sim; Saman Halgamuge
• Keywords: Benchmark (surveying); Generative model

The recent renaissance in generative models, driven primarily by the advent of diffusion models and iterative improvement GAN methods, has enabled many creative applications. However, each advancement is also accompanied a rise potential for misuse. In arena deepfake generation, this key societal issue. particular, ability to modify segments videos using such techniques creates new paradigm deepfakes which are mostly real altered slightly distort truth. This been under-explored current detection methods academic literature. paper, we present method that can address issue performing prediction at frame video levels. To facilitate testing our method, prepared benchmark dataset where have both fake sequences with very subtle transitions. We provide on proposed utilizes Vision Transformer based Scaling Shifting learn spatial features, Timeseries temporal features help interpretation possible deepfakes. Extensive experiments variety generation show excellent results segmentation classical video-level predictions as well. will form powerful tool moderation deepfakes, human oversight be better targeted parts suspected being All reproduced at: github.com/rgb91/temporal-deepfake-segmentation.