| Intrusion detection system as an important means of network security, there has been more and more attention. In the data mining algorithm is applied to the intrusion detection technology is an important step in intelligent intrusion detection system. Intrusion detection system at work need to be monitored network for data collection, which is bound to generate vast amounts of network data, how effective pre-treatment of these data is a key step in to ensure that the system works correctly. Data preprocessing work is generally accomplished by the feature extraction algorithm, which contained principal component analysis (PCA) and independent component analysis (ICA) are a feature extraction more extensive use two algorithms.The article first introduces the intrusion detection system development status and future trends, followed by a detailed description of the KDD Cup99data sets used to illustrate the importance of data preprocessing in the intrusion detection system by working principle of PCA and ICA analysis and calculation methods, given the KDD Cup99data feature extraction method of calculating combined in the experimental section three classification algorithms:Bayesian classifier, K-nearest neighbor and artificial neural network, verify the data making the classification accuracy of PCA and ICA processing. Through the detailed data of the experiment are given these three algorithms in intrusion detection evaluation, draw the Bayesian classifier satisfaction in the KDD Cup99dataset. In the end of the experiment proposed a weighted PCA and ICA Intrusion Detection System Model, and the experimental data, integrated after the system has the characteristics of classification accuracy is high.Finally, the conclusion and Prospect of research work of this thesis, points out that ICA is not suitable for single use in intrusion detection system, as a necessary auxiliary measures might be better. Clearly the next step to be based on the existing research results research. |
PDF Full Text Request