A Comprehensive Approach to Digital Image Forgery Detection using Multi-Stage Manipulation Analysis

The proliferation of powerful image editing software has made digital image forgery a significant threat to the integrity of visual information in journalism, forensics, and social media. Current detection methods often focus on single, specific artifacts, making them vulnerable to sophisticated forgeries that employ multiple manipulation stages. This paper proposes a comprehensive, multi-stage analysis framework for detecting image forgeries by examining traces left across the entire manipulation pipeline. Our method systematically analyses artifacts from pre-processing (e.g., resampling, compression), primary manipulation (e.g., splicing, copy-move, removal), and post-processing (e.g., blending, re-compression, noise addition). By fusing features from multiple forensic domains—including pixel-level consistency, camera-based sensor noise, and format-level compression artifacts—we create a robust classifier resilient to anti-forensic techniques. Experimental results on standard datasets (CASIA, NC2016, WildWeb) demonstrate that our holistic approach outperforms state-of-the-art single-method detectors, especially in complex, multi-operated forgeries.