Under review · Targeting: Human Behavior and Emerging Technologies (Wiley)

FakeSmash: A Multimodal Deep Learning Framework for Real-Time Detection of Video, Audio, and Image-Based Deepfakes in Social Media Applications

A. A. Olagunju, et al. (7 co-authors)

Abstract

A multimodal deep-learning framework built with a Design Science Research method that fuses frequency-domain signals across video, audio, and image modalities to detect deepfakes in real-time social-media contexts. Trained and evaluated on FakeAVCeleb and FaceForensics++.

Keywords

deepfake detectionmultimodal deep learningfrequency-domain featuresDSR

PDF / DOI to be added once the record is finalised.

UniOsun PhD thesis →

← All publications

FakeSmash: A Multimodal Deep Learning Framework for Real-Time Detection of Video, Audio, and Image-Based Deepfakes in Social Media Applications

Abstract

Keywords

Related