Intrusion Detection on CSE-CIC-IDS2018 Dataset Using Machine Learning Methods

Over the past few decades, the significance of computer and information security has grown exponentially, driven by the escalating frequency and sophistication of cyber threats. Despite the rapid advancements in both intrusion techniques and security technologies, many organizations continue to rely on outdated cybersecurity strategies, leaving them vulnerable to increasingly complex cyberattacks. Conventional defenses, such as basic firewalls and signature-based detection systems, are often insufficient against modern attackers who use advanced methods, including zero-day exploits and polymorphic malware, to evade detection. Government web servers, which house vast amounts of sensitive citizen data, are especially attractive targets for malicious actors. In response to these evolving threats, the deployment of an Intrusion Detection System (IDS) has become a critical component in securing network infrastructures, providing an essential layer of defense against unauthorized access and data breaches.This study explores the efficacy of six distinct machine learning-based classification methods; Random Forest, Gradient Boosting, XGBoost, CatBoost, Logistic Regression, and LightGBM each selected for its particular strengths in handling complex, high-dimensional data. These algorithms were applied to a comprehensive dataset to detect malicious activities, with a focus on achieving high accuracy and robustness in classification performance. Remarkably, all six models demonstrated substantial effectiveness, achieving accuracy rates as high as 0.98 and AUC values reaching 1.00, underscoring their potential in enhancing IDS capabilities. The results highlight the importance of leveraging advanced machine learning techniques in bolstering cybersecurity defenses, particularly in critical domains like government data protection, where precision and reliability are paramount.