Data Preprocessing Technology Research Of Dendritic Cells Algorithm

At present, with the prevalence of Internet, network attacks are becoming increasingly violent and fierce, The Intrusion Detection been more and more concerned about the attack because it provides a proactive defense technology. In the recent years, the study of the human immune system, applied to the computer field, has brought new inspiration to the intrusion detection system. But most of the traditional artificial immune systems are built on the basis of "self and non-self", which had the defects of oversize autologous set and the difficulty in distinguishing self from non-self. The introduction of danger theory is good enough to make up for such defects. As the danger theory based on the dendritic cells (DCA) algorithm does not require a large number of training samples, simple and convenient, easy to implement, And what’s more, it possesses its unique biological features, such as self adaptability, robustness, tolerance and so on, which are just what intrusion detection system needs. So dendritic cells algorithm is applied to the intrusion detection system can better protect the computer security.Because of the data from the network with characteristic high-dimensional, complex, large amounts of redundancy, the property is missing, incomplete, etc.. However, mature algorithm has certain requirements for input data.For instance, the input data should be clean, neat and have low dimension.Therefore, the data preprocessing needs. Currently, DCA-based intrusion detection system preprocesses data mostly use relatively simple processing techniques. This way is not only time-consuming and laborious, but also difficult to achieve. Thus it becomes especially important to realize automatic data preprocessing of DCA.This article is mainly to do the following work:1. The idea of combining the principal component analysis with the dendritic cell algorithm. To achieve the automatic Preprocessing of the input data.Which involves adopts the method of dimensionality reduction to reduce high-dimensional data.Uses principal component analysis to extracted Eigen values.2. This paper studies the mathematical derivation of the principal component analysis and its algorithms.and do-depth study in the biological mechanism of dendritic cell algorithm and its working principle and its algorithm, and set the input and output signals of the weight conversion function and parameters.3. This paper uses the part of the property of a subset of the KDD CUP99as experimental data set.We can conclude that compare with detection rate and false alarm rate of intrusion detection system which through the use of information gain method and correlation coefficient method to extract the characteristics, that the application of PCA to DCA to realize automatic data preprocessing can make the intrusion detection system have higher detection performance and lower false alarm rate.