Fingerprint authentication is one of the methods used to prevent the loss of personal data, but the minutiae and angle variance of finger patterns make the process difficult. To overcome the issue, an authentication system for fingerprint orientation using an improved Generative Adversarial Network (GAN) with Support Vector Machine (SVM) is being developed. The process begins with image enhancement using Contrastive Limited Adaptive Histogram Equalization (CLAHE) techniques to increase image contrast. Next, GAN with SVM is employed to generate synthetic data through data augmentation, and decision-making is carried out using SVM. Both real and synthetic samples are binarized using the global thresholding technique, and edge-based thinning methods are applied to enhance the fingerprint patterns. Finally, features such as minutiae points of fingerprints are extracted using the Scale Invariant Feature Transform (SIFT) algorithm, serving as input for the SVM classifier. The implemented GAN-SVM model demonstrates superior performance, achieving a False Acceptance Rate (FAR) of 0.12%, a False Rejection Rate (FRR) of 1.3%, an Equal Error Rate (EER) of 0.29%, and an accuracy of 95.87%. When compared to previous models like Multi-layer Perceptron Neural Network (MLP), Fuzzy Commitment (FC), and Genetic Encryption Algorithm (GEA).