AI/ML Based Malware (.exe) Detection Software using Ensemble Learning

The threat in cyberspaces is becoming more pronounced in the Windows environment and this carries significant risks for confidentiality. The variety of space like AI/ML technologies has also played a role in advancing such attacks. It has also been observed that several efforts have been made in eliminating the problem by using only the signature-based method but such problems can be resolved through AI/ML approaches efficiently. These studies derive a malware detector exploiting the Feature selection using extra-trees and other ensemble learning methods for example Decision Tree, Random Forest, Adaboost, Gradient Boosting, GAUSSIAN NB and Linear regression. Return on extra tree from random forest is highest at 99.6% compared to other techniques such as 98.5% for Adaboost, 98.8% for Gradient Boosting, 70.1% for Gaussian NB and 57.7% for Linear Regression which is the least. The voting on the other hand yields an accuracy of 90.1% that is low and not competitive with random forest. The final model thus achieved is made to work concurrently with a malware detection algorithm designed to detect malware through the PE attributes of the exe that is being assessed. The results of the research suggest that the tree-based ensemble method is effective and efficient in malware classification.