Malicious Malware Detection Using Machine Learning Perspectives

Abstract

The opportunity for potential attackers to use more advanced techniques to exploit more people who are online is growing. These methods include getting visitors to click on dangerous URLs that could expose them to spam and ads, financial fraud, defacement of their website, and malware.  In this study, we tested different machine learning algorithms against a set of harmful URLs to see how well they worked overall and how well they found malware, spam, defacement, or phishing. The ISXC-URL-2016 dataset from the University of New Brunswick was used to make the dataset. The data was evaluated in Weka using the Random Forest, Decision Tree, Naïve Bayes, and Support Vector Machine algorithms. Each evaluation had a split of 80% of the data and a 5-fold, 10-fold, or 15-fold cross-validation. It was found that the 10-fold Random Forest algorithm correctly categorized 98.8% of the dataset's cases with the most accuracy.  The results of this experiment showed that machine learning can be a useful tool for companies that want to improve their security. Despite different limitations encountered in the completion of this research, This study is the most comprehensive available on the use of practices relevant to Malware detection.

Keywords:Machine Learning, URLs, Random Forest, Naive Bayes, Decision Tree, Support Vector Machine

DOI: 10.7176/JIEA/12-2-02

Publication date: November 30^th 2022