Multimodal Deepfake Detection using Multi-Scale Transformers – A Detailed Review
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
This paper reviews recent progress in multimodal deepfake detection with an emphasis on multi-scale transformer architectures.It examines the challenges of detecting manipulations across both visual and audio modalities, focusing on cross-modal inconsistencies and synchronization issues.Approaches such as multi-scale attention, hybrid CNN-transformer models, and multimodal fusion are analyzed.Benchmark datasets including Face Forensics++, Celeb-DF, DFDC, and FakeAVCeleb, are discussed for training and evaluation.The study highlights the limitations of CNN-based methods while demonstrating the advantages of transformers in capturing spatial, temporal, and auditory cues.Finally, it outlines future directions for robust and scalable deepfake detection
Fields of Study & Subjects
Digital Media Forensic Detection
Generative Adversarial Networks and Image Synthesis
Biometric Identification and Security
Publication details
- Journal / source
- Advances in intelligent systems research/Advances in Intelligent Systems Research
- Pages
- 518–530
- Publication date
- 1 April 2026
- Language
- EN
- Document type
- Book chapter