Cyber Security Fraud Detection Using Machine Learning Approach

It is notoriously harder to identify fraudsters for its fluid nature in terms of recognizable trends. Latest technological breakthroughs were used by scammers. Individuals overcome the protection, incurring millions of dollars of lost revenue. Data mining approaches could be employed to crunch numbers and discover out-of-the-ordinary behaviors, allowing it to be pursued to its cause, in this case a fraudulent payment, activities. In this study, we really would like to evaluate and contrast various popular ml algorithms, namely k-nearest peer (KNN), randomized forest (RF), and support vector (SVM), along with popular deep neural networks, including auto - encoder, Classifiers, Multiple solutions, and multi - layer perceptron (DBN). The European Union (EU), Canada, and Netherlands information will be used. The measures utilized for evaluations are the Region That under Fitted Model (AUC), the Matthew R Squared (MCC), and the cost of failing.

I. Introduction

For as long as there have been transactions and contactless transactions, con men have explored new methods to take advantage of the weaknesses individuals by taking their payment details and executing suspicious transactions. As a consequently, many phony transactions are made nearly every day. Large banks and online marketplaces are taking measures to identify and prevent these types of fraudulent transactions. By using Computer Vision and Machine Learning approaches, they hope to identify the criminals just before payment is finalized and validated.