• Version
• Download 7
• File Size 348.16 KB
• File Count 1
• Create Date 19 December 2025
• Last Updated 19 December 2025

Detecting Deepfake Videos Using Hybrid Machine Learning Models

 

 

VISHAL RAJBHAR

Modern Education Society's D. G. Ruparel College of Arts, Science and Commerce Matunga West, Mumbai

vishalmanojrajbhar982@gmail.com

 

 

Abstract: Deepfake videos have emerged as one of the most serious technological threats in the digital era due to their ability to manipulate facial expressions, replace identities, and generate highly convincing forged content. These videos are created using advanced deep learning techniques, especially Generative Adversarial Networks (GANs), which continually improve their realism and become harder to detect by the human eye. As a result, deepfakes pose significant risks in areas such as political misinformation, financial fraud, cybercrime, harassment, and the spread of misleading social media content. This growing concern highlights the urgent need for reliable detection mechanisms that can accurately identify manipulated videos before they cause social or personal harm.

In this research, we propose a hybrid machine learning approach that combines Convolutional Neural Networks (CNNs) for spatial (frame-level) feature extraction and Recurrent Neural Networks (RNNs)/LSTM for temporal (motion-based) analysis. The hybrid model is designed to capture both pixel-level abnormalities and unnatural movement patterns that are commonly present in deepfake videos but may not be noticeable through traditional or single-model methods. By integrating these two complementary models, the system achieves improved detection performance while maintaining interpretability and efficiency.

The results of the study demonstrate that hybrid ML models outperform individual CNN or RNN models in terms of accuracy, reliability, and robustness. The proposed model offers a promising direction for building practical deepfake detection systems suitable for real-time verification on social media, digital security platforms, and media authentication tools. Overall, this paper aims to provide a clear and simplified understanding of hybrid deepfake detection techniques while highlighting their importance in ensuring digital trust and online safety.

Keywords —

Deepfake Detection, Hybrid Machine Learning, CNN, RNN, Face Forgery, Video Forgery, AI Security.

Download