Anomaly Detection Based on Symmetric Neighborhood Relationship
Zhiyi Qu, Wenxiu Zheng · 2007
Some particular attacks can be detected when applying outlier mining to anomaly detection. Besides classical outlier analysis algorithms, recent studies have focused on mining local outliers, for example, Wen Jin et al. proposed a measure which mines outliers based on symmetric neighborhood relationship [1]. In network intrusion detection, the processing precision and efficiency of the existing anomaly detection measures are not satisfactory. To avoid this problem, we introduce an outlier mining measure based on a symmetric neighborhood relationship and its algorithm, and describe the use of this approach to detect anomalies. Primary experiments suggest that this method be feasible and much more effective and efficient.