Research On Intrusion Detection Based On Feature Selection

Internet has brought convenience to people and also poses threats, and its explosive growth makes people face more and more network threats, and network security is becoming increasingly prominent.?Intrusion detection as a proactive and dynamic defense means of security is a strong complement to traditional security technologies.?However, with the enlargement of network data volume, information overload makes intrusion detection technology faces enormous challenges. How to effectively reduce resources'consumption in the process of intrusion detection and how to improve the detection performance becomes the most direct problem of intrusion detection technology.Feature selection which is developed with the development of the large-scale machine learning, aims to find the optimal subset from the data's original features. Feature selection can identify the most relevant features through processes of subset searching and evaluation, then remove redundant features and reduce the data dimension so as to extract the most useful information. Thereby, it can reduce the computational complexity of classification algorithms and improve the efficiency of classification algorithms.PSO is an evolutionary computing technology based on swarm intelligence, and it is characterized by fast convergence, simple calculation and so on, in the process of dynamic optimization. Using it in the process of subset searching in feature selection enable us to effectively shorten the search process and improve the efficiency of feature selection.?Discrete binary particle swarm optimization takes the same advantages with particle swarm optimization and when used in feature selection, it does not require users to pre-determine the size of optimal feature subset, but the disadvantage is that it may get trapped in local optimum.Through careful analysis of the feature selection process and advantages and disadvantages of the binary particle swarm optimization, this dissertation makes some corresponding improvement in the binary particle cluster algorithm. On this basis, we apply a feature selection method with the application of the binary particle cluster algorithm and least squares support vector machine and finally applied in intrusion detection. The experiments show that the method can improve the efficiency of intrusion detection.