Design and implementation of secure ATM system using machine learning and crypto–stego methodology

Abstract

The crucial prerequisite in these days is to get rid of various forms of attacks. Nowadays, for financial transaction, automated teller machines (ATMs) are the mostly used gadgets in which personal identification numbers (PINs) are generally used for transaction. But personal identification numbers (PINs) are not secured from many types of threats (spoofing, eavesdropping, man-in-the middle attack etc.), which can affect the security of the confidential and private information. Due to this reason, different biometric systems gain popularity worldwide for their behavioral and physiological features. However, the current biometric systems, for example, iris, palm, faces fingerprints or voice are extremely complex to use and have different disadvantages. In order to overcome these disadvantages a new concept has been introduced in this paper, for authentication in ATM a finger vein authentication method and for information (finger vein image) transfer a combined approach of steganography and cryptography scheme is used. Finger vein authentication system is implemented by the combination of machine learning and image processing procedure. For the purpose of information (finger vein image) transfer, a combined approach of light-weight cryptography and steganography (variable MSB–LSB algorithm) has been proposed. For finger-vein authentication system, the experiment shows that in proposed classification procedure (one-versus-one and one-versus-all) the average recognition accuracy is 98.75% and 97.92% and the execution time is 0.168 s and 0.187 s respectively. For transferring finger vein image, light-weight cryptography and variable MSB–LSB algorithm are proposed, because light weight cryptography is superior to traditional cryptography in terms of time consumption and variable MSB–LSB algorithm is superior to simple LSB in different aspects such as randomness, security and amount of space consumed.