Predictive Security in eGovernment Systems Using Machine Learning Algorithms

The increasing reliance on E-Government systems has transformed public service delivery, making it faster and more accessible. However, this digital shift also exposes government platforms to escalating cybersecurity threats such as ransomware, phishing, and DDoS attacks. These threats can disrupt essential services and compromise sensitive public data, making proactive security a critical need. This study aims to develop a predictive security model for E-Government systems using machine learning techniques to detect potential cyber threats before they cause damage. The used of CICIDS2017 dataset–recognized for its diversity and realism–to train and evaluate various algorithms, including Decision Trees, Random Forests, Support Vector Machines (SVM), and Neural Networks. Data preprocessing steps involved normalization, feature selection, and class balancing using SMOTE. Our approach was benchmarked against traditional intrusion detection systems (IDS), namely Snort and Suricata. Results show that machine learning models, especially Random Forest, significantly outperform conventional IDS tools across all performance metrics, achieving 90% accuracy and an AUC score of 0.92. These findings underscore the value of integrating machine learning into public-sector cybersecurity frameworks. Future work will focus on real-time implementation, expanding dataset diversity, and improving model transparency through explainable AI to ensure both robust security and public accountability.