Xgboost-Based Multi-Steps Cybersecurity Attacks Detection Model

With the rapid growth of software and networks, the rate of cyber-attacks has increased rapidly. As a result, the demand for a dependable and suitable Intrusion Detection System (IDS) solution for safeguarding devices and networks has become essential. Nevertheless, in order to accurately detect the activities of new kinds of crimes, particularly tasking-step incidents, an effective IDS requires an accurate and up-to-date dataset. In this study, MSCAD is used which comes with multi-step attack tasking: the initial attack is for password-cracking type, and the subsequent one is a volume-driven Distributed Denial of Service (DDoS) attack. The dataset used (MSCAD) contains five types of internet attacks including Port Scan Traffic, App-based DDoS, Volume-based DDoS, Web Crawling, and Password Cracking (Brute Force). Nine algorithms including Gaussian Naive Bayes, Bournuli Naive Bayes, Decision Tree, K-Nearest Neighbors, Catboost, XGB, and Random Forest were employed as classifiers. The accuracy rate reached over 99.9% in terms of accuracy and AUC-ROC.