Identity is the most important thing for identification process. Identification processes that are increasingly vulnerable to hacking require secure system input. Biometrics is a solution for input identification systems that are not vulnerable to hacking. one type of biometric input is a fingerprint because it is unique to each individual and does not change over time. Fingerprint identification system can be done by classification. The classification in this study uses the Gabor filter method with four orientation angles that range from 00, 450, 900 and 1350 as fingerprint feature extraction, and Support Vector Machine (SVM) one against all as classifier. With 25 classes and 3 data per fingerprint class obtained the greatest accuracy by radial Basis Function (RBF) kernel for 73% for initial data and 76% for additions augmentation data. The difference in accuracy is due to the possibility of more data position changes in the augmented fingerprint image.