Work

Deepfake Detection Using ResNeXt-50 and LSTM Neural Networks

Year: 2026
Type: article
Abstract: Recent advances in deep learning methods have made it possible to create extremely convincing synthetic media, known by the general public as deepfake videos, that carry enormous dangers such as spreading misinformation, political manipulation, financial scams, and individual blackmail. In this research, we present a system based on deep learning to automate the detection of deepfake videos. In the proposed methodology, we adopt the usage of the ResNeXt-50 architecture to learn the effective frame- level features of the video, which pass through a Long Short-Term Memory (LSTM) network that recognizes the relationships over the whole video. To enhance the generalizability of the model, we adopt a sequential setup with ReLU activation and dropout regularization. To evaluate the model’s effectiveness, we utilize benchmark datasets of deepfakes, and our results show that the model of ResNeXt-50 + LSTM is highly effective in distinguishing between real and manipulated videos. The results of this research provide a significant demonstration of the benefit of fusing multi- dimensional data (i.e., spatial and temporal data) to construct a reliable model to identify deepfakes.
Language: English
Open Access status: hybrid
OpenAlex ID: W7218314860
FWCI:
Cites:
Cited by:
Related to: 10