Hybrid Deep Learning-Based Deepfake Detection Using VGG19 and InceptionV3
H. S. Shreyas, Achal Anandmurthy Choudhari, Tanvi Arvind, Rashmi N Ugarakhod · 2025
Deepfake technology has become a significant threat to the authenticity of digital media, cybersecurity and public trust, as it uses AI and machine learning-specifically Generative Adversarial Networks (GANs)-to produce highly realistic fake videos that are increasingly difficult to detect using traditional methods. In response to this growing challenge, this paper proposes a hybrid deep learning model that combines the strengths of two powerful convolutional neural networks: VGG19 and InceptionV3. VGG19 is effective at extracting finegrained texture features, while InceptionV3 excels at capturing complex patterns at multiple scales. The model also incorporates a Multi-Task Cascaded Convolutional Network (MTCNN) for face detection and alignment from video frames. Experimental results on benchmark datasets demonstrate that this hybrid model outperforms the individual models in terms of accuracy, precision, and stability. This research makes valuable contributions to the development of scalable, intelligent systems for multimedia forensics and AI-based digital security.