DESIGN AND DEVELOPMENT OF A MACHINE LEARNING-BASED IMAGE METADATA FORENSIC SYSTEM FOR AI-GENERATED IMAGE DETECTION

This paper presents the design and development of a Machine Learning-Based Image Metadata Forensic System for AI-generated image detection. The proposed system is designed to support digital image analysis through the extraction and examination of metadata contained in image files. The system consists of image upload, metadata extraction, data preprocessing, feature extraction, machine learning classification, and forensic report generation components. It focuses on metadata fields including camera make, camera model, software tag, date and time information, compression type, image resolution, color space, and GPS presence. The proposed approach is informed by Astillero's (2025) study on forensic analysis of image metadata for distinguishing AI-generated images from human-captured images. Unlike the reference study, which primarily evaluates metadata-based classification, this paper focuses on the design of an integrated forensic workflow for automated metadata analysis and classification. The proposed system is intended as a forensic decision-support tool rather than a definitive authentication mechanism. Actual system performance and classification effectiveness require future implementation and experimental evaluation.