Malware Attack Predictive Analytics in a Cyber Supply Chain Context Using Machine Learning

Abstract

Due to the invincibility nature of cyber attacks onthe cyber supply chain (CSC), and the cascading effects ofmalware infections, we use machine learning to predictattacks. As organizations have become more reliant on CSCsystems for business continuity, so are the increase invulnerabilities and the threat landscapes. Some traditionalapproach to detecting and defending malware attack haslargely been antimalware or antivirus software such asspam filters, firewall, and IDS/IPS. These tools largelysucceed, however, as threat actors get more intelligent, theyare able to circumvent and affect nodes on systems whichthen propagates. In our previous work, we characterizedthreat actor activities, including presumed intent andhistorically observed behaviour, for the purpose ofascertaining the current threats that could be exploited. Inthis paper, we use ML techniques to learn dataset andpredict which CSC nodes have detection or no detection. The purpose is to predict which modes are venerable tocyberattacks and for predicting future trends. Todemonstrate the applicability of our approach, we used adataset from Microsoft Malware Prediction website. Further, an ensemble is used to link Logistic Regression, and Decision Tree and SVM algorithms in Majority Votingand run on the training data and then use 10-fold crossvalidation to test the parameter estimation, accurate resultsand predictions. The results show that ML algorithms inDecision Trees methods can be used in cyber supply chainpredict analytics to detect and predict future cyber attacktrends.