Intrusion Detection and Prevention Approach in Wlan Using Cyber Security

Abstract

Today's world has a significantly increased requirement for networking and data sharing. Network security is required due to the rising internationalisation of information technology. Despite the security that firewalls may offer, they never warn administrators of impending threats. There is a need for a trustworthy detection system to enhance efficiency and accuracy when looking for such aberrant behaviour in network packets. The threat of new types of assaults on the network is there constant in the evolving network environment of today. Therefore, updating the network management system regularly is necessary for upgrading the security level. Intrusion Detection Systems is one of the systems used to monitor network packets (IDS).Numerous studies examined the application of machine learning to improve intrusion detection system effectiveness and automatically identify malicious network activity based on network packet patterns. The suggested model was created using a machine-learning method to identify malicious network packet activity. KDD-99 dataset is utilized for that. The decrease computation complexity, the dataset is first standardized. Then, additional characteristics are removed using the corelation method, Particle Swarm Optimization (PSO), and Genetic Algorithm (GA). Only practical elements can identify harmful conduct according to the reduced features. According to the analysis of the result, corelation works best when choosing less than 15 features, whereas PSO performs best when selecting more than 15 features. Following feature reduction, the k-mean clustering technique is used to cluster the data. The efficiency of classifiers may be improved by significantly reducing the training time by employing clustering to create tiny datasets that accurately replicate the original dataset. The proposed algorithm’s last phase involves classifying the dataset into the five attack categories of DOS, U2R, R2L, Probe, and Normal using multilevel hybrid classifiers based on SVM, ELM, and RF. The suggested approach demonstrates its effectiveness in high accuracy, high detection rate, and low false alarm rate compared to other multilevel classification works (FAR).