Understanding Complex Malware

Abstract

With the surge of cybercrime and contribution of malware (malicious software) attacks in cybercrime, there is need to design a smart and deep engine-based Intrusion Detection Systems (IDS). The malware could be virous, worm, trojan, etc. and their behaviors are dynamic and static in nature. IDS may monitor events and activity of malware and classify them in order that prediction of potential attacks can be made for users’ sensitive data and associated computational resources. Moreover, the false positive rate alarming of IDS systems is high. Therefore, there is need to reconsider the design of IDS systems, increase its detection accuracy, and elevate prediction of vulnerable attacks. Further, new IDS must capable to deal with nonlinear behavior of malware datasets and model must have better fitting ability. Therefore, in order to protect and avoid vulnerable attacks, we would like to contribute an implementation of a deep learning algorithm - Extreme Gradient Boosting (XGBoost) and Recurrent Neural Network (RNN) - on Microsoft Malware Classification Dataset (BIG 2015) datasets. Also, BIG 2015 raw datasets will be preprocessed and resized to make the data compatible to these algorithms. Moreover, the performance of these algorithms will be measures and compared using these parameters - Accuracy, Precision, Recall, F1 score, Loss, True Positives, True Negatives, False Positives, and False Negatives, and receiver operating characteristic (ROC) curve by calculating the AUC (the area under the ROC curve).