DETECTING AI GENERATED DEEP FAKE IMAGES USING SOURCE FINGERPRINTING

Abstract

The rapid advancement of artificial intelligence has enabled the creation of highly realistic deep fake images that are difficult to distinguish from authentic photographs. These manipulated images pose serious risks to digital media integrity, cybersecurity, and public trust. This research proposes a source fingerprinting-based framework to detect AI-generated deep fake images by analyzing intrinsic artifacts left by generative models. The system extracts unique fingerprints from images using frequency-domain analysis and machine learning techniques. Feature extraction methods such as noise pattern analysis, texture descriptors, and convolutional neural networks are used to identify generation artifacts produced by different AI models. The extracted features are then classified using supervised learning algorithms including Support Vector Machine, Random Forest, and Convolutional Neural Networks. Experimental evaluation demonstrates that the proposed method effectively distinguishes authentic images from AI-generated deep fakes with high accuracy and reliability. The results indicate that source fingerprinting combined with machine learning provides a robust solution for deep fake detection and digital image authentication