A Comprehensive Review of Deepfake Detection: Advancements with ResNeXt-101 and Transformer-Based Temporal Modeling

View at publisher Summarise with VIRAT
Authors: Pramila S, Snigdha Smita Choubey, R. Roja, Usha Rani N, T. Sasikala, J. Joshua Daniel Raj
Cited by: 0
References: 15

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

Deepfakes have become a major concern due to their ability to create highly realistic manipulated videos that can mislead viewers and cause serious social and security issues. As these manipulations improve, detecting them reliably has become increasingly difficult. This paper reviews recent deepfake detection approaches with a specific focus on models that combine ResNeXt-101 for spatial feature extraction and Transformer-based networks for temporal analysis. The review highlights how these models can be improved through techniques such as data augmentation, feature fusion, regularization, and adversarial-resistant training. It also explains how accuracy, precision, and reliability can be enhanced through cross dataset testing, calibration methods, and strong validation practices. In addition, the paper summarizes common visual inconsistencies found in manipulated videos and discusses the limitations of older CNN-based methods that rely only on single-frame information. Overall, this work provides a clear overview of current progress, challenges, and practical strategies for building more accurate and dependable deepfake detection systems.

Fields of Study & Subjects

Digital Media Forensic Detection Generative Adversarial Networks and Image Synthesis Misinformation and Its Impacts

Publication details

Journal / source
International Conference on Communication, Computing and Electronics Systems (ICCCES)
Pages
1223–1228
Publication date
21 January 2026
Language
EN