Wrapper-based feature selection on ransomware detection using machine learning
Rushikesh A. Pujari, Pravin S Revankar · 2023
Nowadays, in the world of digitalisation, everyone has an android device. As digitalisation is on the rise to become pervasive, the numbers of cybercrimes are also growing. Malware is one of the biggest threats in new era of smartphones connected to the Internet. Ransomware is a kind of program that threatens to harm you by preventing you from accessing your data. Then the attacker demands a ransom to get data access back. A recent study shows that ransomware can target any kind of device connected to internet. So it is important to detect ransomware attacks. In this study, different machine learning (ML) techniques are used to detect ransomware attacks based on network-traffic features. For this experiment all different ransomware families are combined and relabelled as ransomware, making it a 2-class classification problem, and trained the model. According to results, compared to other supervised ML classifiers, the Random Forest classifier with wrapper-based feature selection technique has achieved the highest weighted accuracy of 87.54%. The suggested strategy is tested and validated using the RandomForest classifier on the CICAndMal2017 and CIC-InvesAndMal2019 datasets, respectively. On these datasets, the proposed system outperforms on seven features from the original independent feature set.