Ensemble of Soft Computing Techniques for Inline Intrusion Detection System
An intrusion detection system automates the supervising activities in a computer network and computer system. It is used to analyses activities in network or computer. Basically, intrusion detection system is used to identify abuse or incomplete threats of abuse of computer security policies. It detects intruders, malicious actions, malicious code, and unwanted communications over the Internet. Despite the advancements and substantial research efforts, the general intrusion detection system gives high false positive rate, low classification accuracy and slow speed. For overcoming these limitations, many researchers are trying to design and implement intrusion detection systems that are easy to use and easy to install. There are many methods and techniques of intrusion detection system. Soft computing techniques are gradually being used for intrusion detection system. In this chapter, we present the ensemble approach of different soft computing techniques for designing and implementing inline intrusion detection system. In this work, three base classifiers are implemented using different artificial neural networks. Initially, Neuro-fuzzy neural network, Multilayer Perceptron and Radial Basis Function neural network have been constructed. These three networks have been combined using voting methods of machine learning. Three base classifiers are separately trained and evaluated in term of classification accuracy, false positive rate, false negative rate, sensitivity, specificity and precision. The voting combination ensemble method of machine learning has used to combine these three trained models. The performance ensemble classifier is evaluated and compared with the performances of base classifiers. In our study, we found that final ensemble classifier using Neuro-fuzzy, Multilayer Perceptron and Radial Basis Function neural network is superior to the individual base classifier in detection of intruder in network. The performance of ensemble classifier is measured in terms of classification accuracy and sensitivity. It is also found that ensemble based classifier for intrusion detection system has reasonable classification accuracy, the best sensitivity and false negative rate with very low false positive rate on test data set. The experimental results show that the base classifiers take very less time to build models and the proposed ensemble classifier for intrusion detection system takes very less time to test data set. These advantages can help to deploy the intrusion detection system to easily capture and detect online packets.
For more information contact author
D. P. Gaikwad
Department of Computer Engineering, All India Shri Shivaji Memorial Society’s College of Engineering, Pune, India.
E-mail: [email protected]