Reduce false intrusion alerts by using PSO feature selection in NSL-KDD dataset
Aditya Kumar Shukla, Ashish Kumar Sharma · IET conference proceedings. · 2023
Due to the enormous computing power, the network traffic in the intrusion classification system exhibits unexpected behaviour. As the system becomes more complicated, it becomes necessary to analyze its vast array of characteristics. However, the performance of the IDS (Intrusion Detection System) can be significantly impacted by the characteristics of unsuitable or noisy data. In order to eliminate pointless qualities, we used a random forest technique to conduct intrusion detection along with PSO feature selection (FS) in this research. It improves the effectiveness and efficiency of the classification job of intrusion detection. Multiple IDS metrics are measured using a variety of classifiers, including k-NN (k nearest neighbour), SVM, LR (logistic regression), DT (decision tree), and NB. In comparison to the aforementioned state-of-the-art classifiers, the RF + PSO (particle swarm optimization) characteristics selection method was applied on the NSL-KDD datasets, which decreased the noise from the alarm rate and increased the classification rate and performance of the IDS. Performance parameters for the IDSs in this research include AC (accuracy), PR (precision), FPR (false-positive rate), and classification rate. The findings demonstrate better accuracy (99.26%), and PSO helped reduce the dimension of the data.