Towards Explainable Deepfake Detection: A Survey

Authors:

T S R Sriraga, Surekha T S, Sadhana M, Varshitha

Department of Computer Science and Engineering, K.S. Institute of Technology #14, Raghuvanahalli, Kanakapura Main Road, Bengaluru - 560109

Guide: Ms. Maddela Bhargavi

Assistant Professor, Department of Computer Science and Engineering, K.S. Institute of Technology #14, Raghuvanahalli, Kanakapura Main Road, Bengaluru - 560109

ABSTRACT:

The rapid growth of artificial intelligence has enabled the creation of highly realistic manipulated media known as deepfakes. These synthetic images and videos are generated using deep learning techniques and are increasingly difficult to distinguish from real content. Although deepfake technology has useful applications in entertainment, education, and virtual content creation, its misuse can lead to misinformation, identity theft, financial fraud, and social manipulation. Therefore, the development of reliable deepfake detection systems has become an important research area.

This survey presents a comprehensive review of deep learning-based deepfake detection methods, including Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), hybrid architectures, and transformer- based approaches. The study also discusses Explainable Artificial Intelligence (XAI) techniques such as Grad-CAM, which improve the interpretability and transparency of detection models. In addition, commonly used datasets, evaluation metrics, advantages, limitations, and current research challenges are analyzed. Based on the comparative analysis, hybrid and transformer-based models show improved performance in detecting manipulated media, while XAI techniques enhance trust and understanding of model predictions. The survey concludes by highlighting future directions for building accurate, efficient, and explainable deepfake detection systems suitable for real-world applications.

Key Words: Deepfake Detection, Explainable AI, CNN, RNN, Transformer, Grad-CAM, Deep Learning