MalNet: Detection of Malwares Using Ensemble Learning Techniques

Threats to users' security and safety are multiplying constantly in this era of the Internet. Malicious software, sometimes known as malware (ransomware, viruses, Trojans, etc.), is one of these dangers. Important information (such as bank account details, etc.) may be lost or maliciously replaced as a result of this vulnerability. Traditional malware detection techniques have been circumvented by malware developers because of time-consuming and unreliable software that is not yet known. This highlights the necessity for sophisticated methods of malware detection, particularly for newly discovered or unresearched malware. The detection of malware can be done intelligently using machine learning, which consists of the two stages of feature extraction and classification. Due to the metamorphic and polymorphic nature of malware, a single classifier malware detection methodology is impossible to produce meaningful findings. An ensemble learning-based approach for malware detection is suggested by this work. In comparison to its individual classifiers, the model suggested in this research performs relatively better overall. The foundation classifiers for this ensemble model are Gaussian naive Bayes, decision trees, XGBoost, logistic regression, bagging, random forest, and stochastic gradient descent. Using four datasets—namely, Malimg, BIG 2015, MaleVis, and Malicia—the efficiency of the anticipated model and the separate classifiers utilized to generate the collaborative model is trained and evaluated. Both detection and classification are evaluated in terms of performance. The recommended approach uses Correlation based Feature Selection, a feature selection method, to select only the most pertinent features.