Random Forest-Based Intrusion Detection System with Real-Time Visualization

Intrusion Detection Systems (IDS) are important to protect computer networks of the modern era against more complex cyberattacks. Old signature-based IDS do not work well in identifying new and changing threats. In this paper, the Intrusion Detection System will be proposed based on machine learning and a Random Forest classifier trained on the CICIDS2017 dataset. Normalization techniques and Synthetic Minority Oversampling Technique (SMOTE) are used to preprocess the dataset in order to deal with class imbalance. The model suggested organizes the network traffic into the categories of Normal, DoS, DDoS, Probe, R2L and U2R attacks. Moreover, a real-time visualization framework and automated reporting module is also incorporated to make it easier to use. The experimental data reveals that the proposed system has a high detection performance with 97% accuracy, 96% precision, 95% recall, and 95.5% F1-score, and is thus appropriate to be used practically in network security settings.