Machine Learning for Malware Detection on Balanced and Imbalanced Datasets
Manish Kumar Goyal, Raman Kumar · 2020 International Conference on Decision Aid Sciences and Application (DASA) · 2020
There is a tremendous growth of malware with each passing day. It has become difficult to cope up with such an increasing number of malware, especially with new and unseen malware. It has posed a serious threat to software and the internet. Malware and machine learning is like a pair made in heaven. The malware contains various similar patterns due to the reuse of code while machine learning is used to detect those similarities. In this paper, two experiments are performed for balanced and imbalanced data on a previously build a dataset of malware detection on API calls using various machine learning classifiers like k-Nearest Neighbors, Gaussian Naive Bayes, Multi Naive Bayes, Decision Tree, and Random Forest. In both experiments, Random Forest provides the best results with an accuracy of 90.38% on a balanced dataset and 98.94% on an imbalanced dataset.