Ensemble Classifiers For Malware Classification Using Static And Dynamic Analysis

Malwares or malicious software is an umbrella term pertaining to different categories of destructive and intrusive software. Malware thus are a harmful to the computer system security and a threatening challenge for cyber security and information security personnel. New generation malware is built using sophisticated techniques which helps them to evade any detection strategy. This paper talks about leveraging artificial intelligence models to build an efficient system to detect and prevent malwares from our system. Classical learning models and deep learning models have their own positives and negatives, and ensemble learning gives a balance between the two approaches. This paper implements a stacked ensemble model and compares the performance of the model build using static features and dynamic features independently. The base learners used are SVM, Naive Bayes, KNN, Random Forest, XGBoost, Gradient Boost Classifier, Logistic Regression and AdaBoost Classifier. The Meta Classifier used to build the final classifier is Gradient Boost Classifier. Experiments were conducted on two popular dataset build using static and dynamic analysis and the results are compared.