Enhanced Network Detection System using NSL-KDD

Intrusion detection systems must be able to adjust to various attack patterns in various network environments due to the increasing sophistication of cyber threats. When used in real-world situations, traditional systems that were trained on a single dataset frequently have trouble generalizing. This study suggests a hybrid method that uses ensemble techniques to integrate various machine learning models that have been trained on various datasets. We developed two distinct models that captured a variety of attack signatures and common behavior patterns using various network traffic datasets. Improved detection capabilities were obtained by combining these models with majority voting procedures, stacking classifiers, and weighted averaging. While keeping false positive rates low, our experimental implementation showed improved accuracy in identifying known and unknown intrusions. Python was used in the Google Collab environment to develop the system, enabling scalable computation and simple reproducibility. The hybrid approach outperforms individual models in terms of detection accuracy, precision, recall, and F1-score, according to performance evaluation across a number of metrics. A useful framework for creating more robust network intrusion detection systems is provided by this work.