Deepfake Face Image Detection: Benchmarking CNN Models' Performance

The Deepfake Face Image Detection System uses Convolutional Neural Networks (CNNs) to address the growing threat of deepfakes, which manipulate images to falsely depict events or actions, threatening privacy, and security. The development began with data collection to understand deepfakes and explore deep learning solutions. A diverse dataset of real and fake faces was used for training, with the system built in Python using VS Code and a user-friendly interface. Various CNN models, including basic CNN, ResNet-50, VGG-19, and XceptionNet, were evaluated, achieving 92 % accuracy in classifying real and fake images. The system's unique dataset, combining high- and poor-quality images, enhances its ability to detect both clear and blurry deepfakes. The Graphical User Interface (GUI) demonstrated reliable predictions. Future improvements could include real-time detection in audio and video. This system highlights the effectiveness of CNNs in combating deep-fake threats, contributing to privacy and security awareness.