Research Of Network Intrusion Detection Method Based On Outlier Mining

Intrusion detection methods based on outliers mining are studied in this paper. Firstly, we introduce the various types of outlier detection algorithms and focus on analyzing the advantages and disadvantages of distance-based algorithms and density-based algorithms, then, the idea of gravity-based is introduced and is combined with improved clustering algorithm proposed in this paper, thus the CGIDA (intrusion detection algorithm based on clustering and gravity) is designed. Secondly, the simulation experiment is designed in classical intrusion detection KDD CUP99 data set, experimental results show that detection rate of the algorithm is improved and false rate is reduced, besides, this paper proposes eigenvector method to standardize categorical attributes and analyze the reasonableness of the method. Finally, CGIDA is applied to the Snort system and the rationality and efficiency of the application is proved by simulation experiments.