Physiological trait-based Fraud Prevention System in ATM Transactions

Knowledge-based and token-based automatic personal identificationapproaches have been the two traditionaltechniques widely used. Because knowledge-based and token-based approaches are unable to differentiate between an authorized person and an impostor who fraudulently acquires the token or knowledge of the authorized person. This paper investigate the use of Physiological traits in online transactions using particle warm optimization. In this experiment 120 tongue images and fingerprints of different individuals were acquired using digital camera and webcam. The traits were preprocessed using segmentation scheme and particle swarm optimizationwas used to select salient features. Three testing scenarios were created; (i) testing with all images in database and (ii) testing with another selected pose of images in database and (iii) poses of images not in databasebut captured in a bad illumination. The decision to recognize or classify the images was determined by threshold at 0.50. Scenario 1 results showed that the experiment produced sensitivity of 94.6%, specificity of 95.7% and accuracy rate of 94.6% for tongues and sensitivity of 87.2%, specificity of 81.4%, and accuracy of 90.1% for fingerprints. Scenario 2 results showed that the experiment produced sensitivity of 78.6%, specificity of 74.7% and accuracy of 93.3% for tongues while fingerprints has sensitivity of 65.2%, specificity of 69.4%, and accuracy of 75.5%. while scenario 3 results produced sensitivity of 0.6%, specificity of 0.77% and accuracy of 0.0% for fingerprints and sensitivity of 22.6%, specificity of 18.4%, and accuracy of 20.7% for fingerprints. However, using tongue along with PSO will enhance better performance accuracy for this automatic identification and authentication in ATM system.