BS-3-11 A Study on Applying Outlier Detection for Network Intrusion Detection(BS-3. Management and Control Technologies for Innovative Networks)

概要

論文の詳細を見る
Network intrusion detection is an important task to prevent malicious activities such as denial of service attacks, port scans, illegal accesses, and so on. Outlier detection is considered as one of suitable technologies because it can discover any abnormal attack behaviors behind large volume of normal data. This paper discusses some general approaches with clustering methods, and also tries to apply our proposed RPGS (Rim Projected Grid Statistic) method to be compared with an ever-proposed and widely-used algorithm, which is LOF (Local Outlier Factor). The comparative experiments are conducted on the KDD Cup 1999 dataset. The experiments results indicate that no perfect outlier detection scheme can fit all the attack type data, however, our proposed outlier detection method performs better than LOF method.
2012-03-06