Siddaiah, Darshan (2024) DeepFakeCNN: Deep Fake Image and Video Detection using Convolutional Neural Networks. Masters thesis, Dublin, National College of Ireland.
Preview |
PDF (Master of Science)
Download (1MB) | Preview |
Preview |
PDF (Configuration Manual)
Download (419kB) | Preview |
Abstract
Organisations have significant challenges in dealing with social cybercrimes and safeguarding against the spread of manipulated media due to the emergence of deepfake technology. This study presents an advanced deep learning system, "DeepFakeCNN", built for the purpose of detecting and alerting users about the presence of deepfake images and videos. The DeepFakeCNN model uses a convolutional neural network to accurately distinguish between genuine and falsely created pictures. Additionally, it provides monitoring and analysis capabilities to security personnel. The methodology may be readily employed by several social media platforms to detect deepfake videos, pictures, reels, and other similar content. This includes popular communication services such as Microsoft Teams, Google Meet, and Meta's WhatsApp/Messenger. It immediately provides instant notifications when suspicious deepfakes are found. In this EfficientNetB7 a convolutional neural network demonstrated superior performance in the evaluation, achieving an impressive accuracy of 93.99%. This model demonstrated balanced performance by correctly distinguishing between genuine and fake images and videos, achieving a recall rate of 75.33%, a precision rate of 74.34% and F1 score of 74.83%.
Actions (login required)
![]() |
View Item |