Research And Implementation Of Intrusion Detection System Based On Ensemble Learning

With the rapid development of Internet technology, more concerns have been put on network security. As an active technology of security defense, IDS becomes an important method of assuring the security of network.It provides the real-time detection of inner or outer attack, but towards to more complex ways of attack, traditional IDS systems show the shortcomings of itself. Machine Learning focus on how to improve the performance with experience accumulation automatically, so taking advantage of Machine Learning to help the IDS system is now a consensus and it explores a brand new way to make the IDS system better.Firstly, the history and theory of IDS system was learned, including basic concept, principles and classification. Different kinds of IDS technology were compared, and the disadvantage of current IDS system and its future was indicated too.Secondly, the Feature Selection and Ensemble Learning technology of Machine Learning was introduced. Since the combination of IDS and Machine Learning improves the intrusion detection,machine learning was used in IDS. Because Feature Selection can achieve the data dimension reduction of the real-time network data package and Ensemble Learning classifies the network data better, the combination of the two technologies in data analysis for IDS can obviously improve the effect of IDS.At last, the ensemble learning algorithm ReFeatEn based on feature selection was studied. The validity of this algorithm was verified by an experiment after depicting its process.Then a model based on theories above was build ,and the data collection part,pre-process part,detection part of this model was designed and implemented.