Towards Optimization of Malware Detection using Extra-Tree and Random Forest Feature Selections on Ensemble Classifiers

The proliferation of Malware on computer communication systems posed great security challenges to confidential data stored and other valuable substances across the globe. There have been several attempts in curbing menace using a signature-based approach recent times, machine learning techniques extensively explored. This paper proposes framework combining exploit both feature selections based extra tree random forest eight ensemble five base learners- KNN, Naive Bayes, SVM, Decision Trees, Logistic Regression. K-Nearest Neighbors returns highest accuracy 96.48%, 96.40%, 87.89% extra-tree, forest, without selection (WFS) respectively. Random Feature Selections are with 98.50% 98.16% extra-tree Extreme Gradient Boosting Classifier is next random-forest FS an 98.37% while Voting least detection 95.80%. On FS, Bagging 98.09% 95.54%. Forest has all seven evaluative measures techniques. study results uncover tree-based model proficient successful for malware classification.