An Intelligent Real-Time Intrusion Detection System Using Optimized Random Forest on NSL-KDD Dataset

Network security is getting tougher day by day due to the massive increase in IoT devices and cloud usage. The main problem with old IDSs is that they often remain too slow and trigger too many false alarms. We built, in this project, a smart Network IDS using an optimized Random Forest classifier. We also applied Feature Selection to make the system faster by removing useless data. We trained this model on the NSL-KDD dataset, which is a clean version of the older KDD'99 dataset. Surprisingly, this model achieved an accuracy of 99.90 % that outperformed many recent studies. Unlike other research that just shows graphs, we actually created a Real-Time Web Dashboard using Python Flask. This allows anyone to check network traffic instantly. This paper proves that using the right feature selection with Random Forest creates a highly accurate and practical security tool.