Machine Learning Approach for Predicting Cyberattacks Using a Bayesian Network Model

Abstract

The increasing complexity and sophistication of cyber threats requires advanced, proactive strategies for information security. Traditional defensive measures are no longer sufficient, prompting organizations to adopt Artificial Intelligence (AI) and machine learning techniques for threat detection and risk management. In this context, Bayesian Networks (BNs) have emerged as a powerful tool to model probabilistic relationships among key behavioral and environmental factors influencing cybersecurity. This study presents a Bayesian Network-based approach to assess individual cybersecurity awareness and predict potential vulnerabilities based on user behavior. Using survey data collected from 102 participants, the model evaluates user responses to risk-related behaviors—such as the use of multifactor authentication, application sources, update practices, and email handling—to generate a predictive risk score. A weighted scoring system that incorporates interaction terms was employed to account for compound risk scenarios. The model achieved a strong performance with an overall accuracy of 92.6%, recall of 88.9%, precision of 57.1%, and an F1-Score of 0.696. These results demonstrate the proposed model’s capability to effectively identify users with high-risk behavior patterns, thereby providing valuable support for organizational cybersecurity awareness programs and decision-making processes.