Employing a CNN Detector to Identify AI-Generated Images and Against Attacks on AI Systems

The advancement of artificial intelligence (AI) technology has made Generative AI a significant concern for many. This technology generates fake images through highly complex algorithms. It involves a detailed analysis of the original image, extracting features like color, context, and feature, which are then used to build a neural network. This network is combined with computer graphics to produce an image that resembles the original. Attackers often use these fake images to attack AI systems, making the detection and prevention of such images a pressing issue. In this study, we propose a method to detect fake images using a detector built on CNN models. Our experimental results demonstrate that the proposed detectors achieve over 95% average accuracy on the test datasets, indicating their potential applicability to real-world problems.