Analysis of online suspicious behavior patterns

Abstract

The online users’ behavior is mainly categorized into two types: normal and suspicious behaviors. When any user is involved in group of activities that are deviated from normal behavior which is observed as misuse, damaging, intrusive, illegal money transfer, and online credit card fraud, then such users are become harmful. These users should be detected prior to the harmful activities. In this work, different online suspicious behavior patterns are studied and analyzed such as terrorists’ activities, Web spam, e-mail spam, OSNs spam, fake reviews, malicious crawlers, banking frauds, hacking. Online credit card fraud is a suspicious activity and is taken into consideration for study. It is necessary to generate decision support system to predict the online credit card fraud as it will help to protect the illegal money transaction. The work represented in this research paper intends to find the best model to predict the online credit card fraud by using machine learning methods. Various machine learning methods such as logistic regression, SVM, kNN, Naïve Bayes, kernel SVM, decision tree, and random forest are implemented in order to predict the credit card fraud. The performance of the system is tested on the dataset which comes from Kaggle Web site. The performance of the algorithms is studied by considering different evaluation measures. Random forest algorithm turns out to be the best model in the prediction task. It has achieved accuracy of 99% and F-measure of 86%.