Citation

Singh, Pallavi and Sasikumar, S. and Geddam, Ratna Sekhar and Tharun, K and Kavitha, M. and Bhuvaneswari, Thangavel (2025) Deepfake Forensics Leveraging MobileNetV2 for High Precision Detection. In: 2025 Multimedia University Engineering Conference, MECON 2025, 21 July 2025 - 23 July 2025, Cyberjaya, Malaysia.

Text 50.pdf - Published Version Restricted to Repository staff only Download (1MB)

Abstract

This Deepfake technology has emerged as a powerful tool for generating synthetic media, leveraging advancements in artificial intelligence and deep learning to create highly realistic fake images and videos. While deepfakes offer potential benefits in fields such as entertainment, media production, and virtual reality, they also present serious threats, including misinformation, identity fraud, and privacy violations. The ability to detect deepfake content accurately and efficiently is critical in safeguarding digital authenticity and preventing misuse. This research focuses on the development of deepfake detection models using Convolutional Neural Networks (CNNs) and MobileNetV2. Three primary detection methods— handcrafted feature analysis, machine learning-based classification, and artifact detection—were implemented and evaluated. A dataset consisting of 20,000 images, equally split between real and fake, was used for model training and testing, and optimized using dropout regularization, learning rate tuning, and early stopping. The CNN model achieved an accuracy of 90%, while an optimized MobileNetV2 Subtraction model improved detection accuracy to 97% by reducing overfitting and refining hyperparameters. The study underscores the importance of robust deepfake detection frameworks, particularly in an era where manipulated content is becoming increasingly difficult to differentiate from real media. Future enhancements will explore real-time detection solutions and adversarial training techniques to further improve detection accuracy and resilience.

Item Type:	Conference or Workshop Item (Paper)
Uncontrolled Keywords:	Deepfake detection, Convolutional Neural Networks (CNN)
Subjects:	T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK5101-6720 Telecommunication. Including telegraphy, telephone, radio, radar, television
Divisions:	Faculty of Engineering and Technology (FET)
Depositing User:	Ms Rosnani Abd Wahab
Date Deposited:	18 Mar 2026 04:34
Last Modified:	18 Mar 2026 05:34
URII:	http://shdl.mmu.edu.my/id/eprint/15540

Downloads

Edit (login required)