Detection of Micro-Texture Anomalies in Synthetic Fingerprint Forgery Using Generative Adversarial Networks and Diffusion Models in Biometric Authentication Systems

The uncontrolled advancement of generative artificial intelligence and Generative Adversarial Networks (GANs) has enabled the creation of high-resolution synthetic fingerprints that are indistinguishable from real ones, posing a serious threat to biometric security infrastructures. Traditional fingerprint authentication systems, which primarily focus on macro ridge structures, are insufficient to detect these sophisticated forgery techniques. In this study, we propose a novel anti-spoofing model based on the Generative Artifact Matrix (GAM), introduced for the first time in the literature, to capture anomalies in the microscopic surface structures of AI-generated fingerprints. A hybrid dataset comprising 2,310 fingerprint images was constructed, including 810 real samples from the FVC2004 dataset and 1,500 synthetic fingerprints generated using GAN and diffusion models. The images were transformed into 1,000-dimensional mathematical GAM matrices using the InceptionV3 architecture and evaluated with five-fold cross-validation. Experimental results demonstrate that Random Forest and Support Vector Machine (SVM) algorithms achieved 100% accuracy (AUC, CA, and F1 = 1.000) in detecting synthetic fingerprints, while Artificial Neural Networks reached 99.9% accuracy. The consistency of the 0.500 p- value across model comparisons confirms the absolute discriminative power of the feature space. This study presents a low-cost, high-performance, and explainable software defense framework for cybersecurity infrastructures.