THREATS AND POTENTIAL RISK DETECTION OF NETWORK INBOUND DATA USING MACHINE LEARNING

Abstract

 In this research, an attempt is made to detect threads and potential risks caused by incoming data through the network. Our research inspects three different types of risks the environmental, the operation risk, and the technical risk.  Two Machine Learning methods have been used for the sake of research comparison and the rigidity of the results. Naive-Bayes and K-Nearest Neighbor algorithms have been applied to a structured data set with 15 input features representing the incoming risk data set to the network and a target prediction column of three different risk categories, namely environmental, operation, and technical risk. Results from Naive-Bays obtained an accuracy of 84% in risk detection, while K-Nearest Neighbor with 5 neighbors produced a 75% accuracy in risk detection.