Ransomware Detection Framework using Soft Voting-based Ensemble Learning

In recent years, many ransomware malware attacks on critical infrastructures have been reported to steal sensitive information and to damage and disrupt the services of the targeted system. Fast and efficient detection of ransomware malware attack is a fundamental research issue. Most of the existing approaches experience low accuracy and a high false positive rate and are unable to detect zero-day ransomware. Thus, this paper proposes a novel fast and efficient ransomware malware detection framework using soft voting-based ensemble model. In the proposed soft-voting based ensemble model, five base learners are used such as random forest, Adaboost, ExtraTree, XGBoost, and decision Tree. Performance analysis is evaluated using CIC-AndMa12017 ransomware dataset in terms of F1-scores, recall, precision, and accuracy. Comparative study of the proposed model with three other models such as k-NN, Random Forest, and AdaBoost are presented. It is opined through experiments that the proposed model accomplished 98.42% accuracy and performs well compared to the other models.