Density Based Outlier Mining Algorithm with Application to Intrusion Detection

Presently, outlier mining is used for many areas such as telecommunication, finance and intrusion detection. However, finding outliers needs amounts of computation with most traditional algorithms. Thus, we propose a modified density based outlier mining algorithm in this paper. For every object in dataset, our algorithm need not judge whether there are core objects within the epsiv-neighborhood of it. In addition, the module information of data object is introduced in our algorithm and it can avoid large numbers of unnecessary computation to finding all outliers. The algorithm is applied on the intrusion dataset and experimental results show it obtains efficient performance for outlier mining while maintaining stable detection rates.