Machine Learning-Driven Intrusion Detection System for Enhanced Accuracy and Scalability

Intrusion detection systems (IDS) are critical for safeguarding networks against evolving cyber threats, but traditional solutions often face challenges such as high costs, integration difficulties, and inaccurate detection. This manuscript proposes a novel IDS framework that integrates machine learning techniques, including random forest (RF), support vector machine (SVM), and long short-term memory (LSTM)-based anomaly detection, to address these limitations. The proposed system was evaluated using the NSL-KDD dataset in a simulated enterprise network environment, and its performance was measured using key metrics such as accuracy, precision, recall, F1-score, false positive rate (FPR), false negative rate (FNR), and response time. The results showed that LSTM outperformed RF and SVM with an accuracy of 97.2%, precision of 97.0%, recall of 98.1%, and F1-score of 97.5%. It also demonstrated the lowest FPR (1.9%) and FNR (2.1%) and the fastest response time (124 ms). RF followed closely with 96.2% accuracy and an F1-score of 95.9%, while SVM achieved 94.6% accuracy and 94.0% F1-score. The system's integration with tools like Snort and Suricata, coupled with real-time alert generation and a centralized dashboard, significantly improved the efficiency of incident response. The proposed IDS framework demonstrated a 13–17% increase in detection accuracy and a 40% reduction in false alerts compared to baseline systems, highlighting its effectiveness for real-time network security monitoring.