A CLUSTERING ALGORITHM FOR DETECTING DDoS ATTACKS IN NETWORKS

Abstract

Intrusion detection systems are crucial to maintaining network security as the number of connected machines increases. Data mining techniques have recently become more significant in solving network security problems, such as network intrusion detection, which is a difficult job in network security. The goal of intrusion detection systems is to find assaults with a low false alarm rate and a high detection rate. Signature-based detection is used in the most popular and commercially accessible intrusion detection techniques. However, their techniques are very expensive and time-consuming to get the signatures, and they are unable to identify unexpected incursions that do not match the signatures. Numerous academics have put forward a variety of methods based on unsupervised learning approaches to address the issues. In order to identify intrusions and increase the detection rate while keeping a low false positive rate, we describe in this work a unique clustering-based intrusion detection technique called unsupervised anomaly detection, which trains on unlabeled data. The 2000 DARPA Intrusion Detection Scenario Specific Data Set was used to assess our approach.