A Hybrid Supervised Machine Learning Model for the Prediction of Insider Threats

Abstract

The quest and sensitivity of organizational resources has permeated need for information confidentiality while ensuring availability and integrity are met, if organizations are to thrive and survive. To fend off malicious insider, organizations have implemented strategies, policies and techniques to manage malicious insider attacks. Machine Learning (ML) algorithms are implemented as learning paradigms, having the ability to learn from prior instances. ML present intelligent implicitly designed models having the capability of predicting possible outcomes from machine learning dataset based on perceived features which might be computationally explored. Hybrid Supervised Machine Learning Model for the Prediction of Insider Threats (HSMLM-IT) has been designed, simulated and validated utilizing Support Vector Machine (SVM) for label classification and Adaptive Neuro Fuzzy Inference System (ANFSI) for predictive learning. The SVM blocks provides a classification accuracy of 92% and precision of 93% while the ANFIS training blocks provides an ANFIS accuracy of 91% and ANFIS error of 9%.