AN ADAPTIVE INSTRUCTION DETECTION SYSTEM USING K-NEAREST NEIGHBORS CLASSIFIER WITH KMEANS AND KMEDIODS CLUSTERING ALGORITHMS

In recent years, the number of attacks has increased and intrusion detection has become the standard for information assurance. The goal of any intrusion detection system is to help computer systems prepare for and respond to attacks. The ultimate goal is to detect these attacks before access or during the access process. Firewalls do not offer complete protection and must nevertheless be supplemented by an intrusion detection system.Various tools and methods are used to monitor user and system events to detect illegal and abnormal activities in networks and systems. All these tools and methods are called intrusion detection systems (IDS). These systems have been implemented keeping in mind the use of artificial intelligence (AI) approaches. , such as Genetic Algorithm (GA), Decision Tree (DT), Expert System (ES), Neural Network (NN).In our research, we built an intrusion detection system that uses K-Nearest Neighbors (KNN) classifier with the help ofKMeans and KMediods clustering algorithms. The system was trained and evaluated on the KDD 99 dataset. The results gave a good indication about the ability of the system to detect type of inserted data(attack or normal) especially forNormal, denial of service(DoS) and Prob types also they displayed the superiority of KMediods over KMeans due to outlier problem in KMeans.