Implementation of machine learning for human aspect in information security awareness

Abstract

This research discussed our experience in implementing machine learning algorithms on the human aspect of information security awareness. The implementation of the classification and clustering approach have been conducted by creating a questionnaire, creating dataset, importing data, handling incompleted and imbalanced data, compiling datasets, feature scaling, building models, and subsequently evaluating machine learning models. Datasets are generated based on the collection of questionnaire result of the distributed questionnaire related to the Human Aspects of Information Security Questionnaire (HAIS-Q) to the stakeholder of an Indonesian institution. Models as results of algorithms implementation through the classification approach has been evaluated by several methods, such as: k-fold Cross Validation analysis, Confusion Matrix, Receiver Operating Characteristics, and score calculation for each model. A model of the Support Vector implementation in the classification has an accuracy of 99.7% and an error rate of 0.3%. Models of clustering implementation are used to determine the number of clusters that can optimally divide the dataset. The model of the DBSCAN algorithm on the clustering approach has an adjusted rand index value of always close to 0.