The Detection of AI-Generated Video Sequences
No Thumbnail Available
Date
2025
Journal Title
Journal ISSN
Volume Title
Publisher
Faculty of Science
Abstract
In recent years, the advancement of artificial intelligence has enabled the creation of highly realistic digital content. Among the most notable developments is the rise of deepfake videos — synthetic videos generated by AI models that can closely imitate human faces, voices, and movements. While such technologies offer creative potential, they also raise serious concerns about misinformation, identity fraud, and digital manipulation.
This thesis addresses the challenge of detecting AI-generated videos by proposing a deep learning-based system capable of distinguishing real from synthetic content. The work combines theoretical research on generative models and detection techniques with a practical implementation of a video classification model.
The proposed system uses a hybrid architecture that captures both visual and temporal features in video sequences. The project includes data preparation, model design, training, and evaluation. Results show the model's ability to detect synthetic videos with promising performance, contributing to ongoing efforts in media forensics and digital content verification.
This research reflects the growing need for tools that ensure trust and authenticity in an increasingly digital world.