Deepfake Detection in Digital Images Using Data Augmentation and Layer Unfreezing Across Various Deep Learning Models

A rise of deepfake technology presents a substantial risk to the integrity of digital media, requiring the adoption of robust detection techniques. This research presents a thorough method for identifying deepfake content in digital images. It involves combining data augmentation approaches and layer unfreezing across a range of deep learning (DL) models. The process comprises augmenting the training dataset to enhance the generalization of the model. Additionally, it involves selectively unfreezing layers in the deep learning models to permit fine-tuning and adaptability. The effectiveness of the proposed approach is assessed through extensive testing using various deep learning models. The results indicate significant improvements in the accuracy of detecting deepfakes and the ability to withstand adversarial attacks. Combining data augmentation and layer unfreezing strategies yielded the best performance among these models, highlighting their significance. This research contributes to advancing the state-of-the-art in deepfake detection and provides valuable insights for enhancing the security and reliability of digital media content.