Intelligent and Soft Computing Systems for Green Energy

16Detecting Fraudulent Data Using Stacked Auto-Encoding: A Three-Layer Approach

P. Saravanan¹, V. Indragandhi² and V. Subramaniyaswamy¹*

¹School of Computing, SASTRA Deemed University, Thanjavur, India ²School of Electrical Engineering, Vellore Institute of Technology, Vellore, India

Abstract

With the growing number of online transactions due to ease of use, the probability of fraud being involved in smart card transactions is increased. Online users are most vulnerable with a credit card, and banks and vendors employ debit card facilities for online banking systems. There are multiple middle-man transaction processing websites, and the user may not know whether he is being subjected to fraud or not. Auto-encoder is a deep learning technology used for many real-world applications, especially in fault detection and fault diagnosis. In this paper, a deep learning-based, stacked auto-encoder system has been designed for predicting the accuracy of normal and fraud clauses in credit card fraudulent datasets. The system is built using TensorFlow with a three-layer stacked auto-encoder and tested using a real-world credit card transaction set. The experimental results show that the proposed system outperforms in terms of accuracy of prediction.

Keywords: Smart card security, fraudulent data, classification, prediction, deep learning, stacked auto-encoder

16.1 Introduction

Banking systems have generally been built using expert systems running on outdated rules to identify fraud, ...