From Pixels to Provenance: Harnessing Source Camera Fingerprints to Detect AI-Created Images

The rapid advancement of generative artificial intelligence has led to the widespread creation of highly realistic synthetic images, posing significant challenges to image authenticity, digital trust, and misinformation detection. Traditional visual inspection and content-based forensic techniques are increasingly inadequate in distinguishing AI-generated images from those captured by real-world cameras. This study presents a robust forensic framework that leverages source camera fingerprints to detect AIcreated images by analyzing intrinsic device-level artifacts embedded during the image formation process. By extracting and modeling sensor-specific patterns such as Photo-Response Non-Uniformity (PRNU) and noise residuals, the proposed approach differentiates between images originating from physical imaging sensors and those synthesized by generative models. The framework integrates signal processing techniques with machine learning classifiers to evaluate provenance authenticity at the pixel level. Experimental results demonstrate that camera fingerprint–based analysis significantly improves detection accuracy, even for visually convincing AI-generated images and post-processed content. The findings highlight the effectiveness of provenance-aware forensics in combating deepfake imagery and reinforcing trust in digital media ecosystems. This work contributes toward the development of reliable image authentication systems essential for digital forensics, journalism, legal evidence verification, and secure multimedia communication.