Prediction of Cyber Security Attacks Using Data Science Techniques: A Comparative Study of Machine Learning Algorithms

computer engineering

In today’s interconnected digital world, cybersecurity threats are becoming more frequent, complex, and adaptive—often outpacing traditional security measures. Static, rule-based systems, such as signature-based intrusion detection, are increasingly unable to detect novel or evolving attack patterns. This research addresses the need for more intelligent, data-driven solutions by investigating how machine learning algorithms can be used to anticipate and classify different forms of cyber attacks. The study centers on four common and impactful attack types: Denial of Service (DoS), Remote to Local (R2L), User to Root (U2R), and Malware. Using a carefully curated dataset of 40,000 network activity records, each with 25 distinct features, we evaluate the predictive performance of four widely used machine learning models—Logistic Regression, Decision Tree, Random Forest, and Support Vector Classifier (SVC). Our methodology includes robust data preprocessing, thoughtful feature engineering, and comprehensive model evaluation using performance metrics such as accuracy, precision, recall, F1-score, sensitivity, and specificity. Among the evaluated models, Random Forest emerged as the most reliable performer overall, although each algorithm demonstrated strengths depending on the specific attack type. These insights can guide the development of more effective and responsive intrusion detection systems, ultimately contributing to a stronger cybersecurity posture across digital infrastructures.

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.