Data Mining Based Algorithms for Intrusion Detection Systems

Abstract

With the tremendous increase in usage of network based services. Security has remained one challenging area for networking experts. There are various security technologies that help fight the inevitable network and security attacks; they have been so vulnerable to exploitations from internal threats. This led to the development of Intrusion Detection Systems (IDS) to complement on the existing methods. Responding to and evaluating IDS alerts is labor intensive requiring vast human resource. Data Mining provides invaluable method to analyze large volume of historical computer systems data, identify patterns, trends and evaluate the behavior of threats and potential vulnerabilities and classify traffic as normal or anomalous. There are many data mining algorithms in use such as rule based approaches, Bayesian networks, Support Vector Machines (SVM) and so on. However, the performance of these algorithms is affected when no optimized features are provided. This leads to high systems processing costs and reduced performance. This paper shows a comparative study on the various data mining techniques.