International Journal of Science, Engineering and TechnologyAuthors: Mr.S.P.Gunjal, Yash Kohakade, Adarsh Gurekar, Sakshi Gadkar
Abstract: Deepfake technology has emerged as a major threat to digital media authenticity in recent years. Deepfakes are manipulated images or videos generated using deep learning models that can closely resemble real human faces. These fake media contents are often misused for spreading misinformation, identity theft, and reputation damage. Although many deep learning models have been developed for deepfake detection, most of them work as black-box systems and do not provide any explanation for their predictions. This paper presents an Explainable Deepfake Detection System for facial images using XceptionNet, Grad-CAM, and MediaPipe Face Mesh. The proposed system is implemented as a Python Flask web application with SQLite as the backend database. XceptionNet is used as the classification model to detect whether an image is real or fake. Grad-CAM is applied to generate heatmaps highlighting suspicious facial regions, and MediaPipe Face Mesh maps these regions onto facial landmarks to provide meaningful visual and textual explanations. The system offers real-time detection with interpretable results, making it suitable for forensic analysis, journalism, and educational purposes. Experimental results show that the proposed system achieves high detection accuracy while maintaining transparency and usability.
