AI-Based Deepfake Analysis for Multiple Sectors

This paper presents an AI-based deepfake detection system capable of analyzing text, image, and video content within a unified framework. The proposed approach integrates transformer-based Natural Language Processing models such as RoBERTa for text analysis, Vision Transformer architectures such as SigLIP for image-based detection, and a frame-based video analysis approach where frames are extracted using OpenCV and analyzed using the SigLIP model. The system performs preprocessing, feature extraction, and classification to identify manipulated media with high accuracy. Experimental results demonstrate that the system can effectively distinguish between authentic and deepfake content across multiple media formats. This research contributes to improving digital trust and supports applications in cybersecurity, media verification, and digital forensics.