Unsupervised Anomaly Detection Using HDG-Clustering Algorithm

摘要: As intrusion posing a serious security threat in network environments, many detection schemes have been proposed recent years. Most such methods employ signature-based or data-mining based techniques that rely on labeled training data, but cannot detect new types of attacks. Anomaly can be adopted to solve this problem with purely normal data. However, extracting these data is very costly task. Unlike the approaches unsupervised anomaly discover "unseen" attacks by unlabeled This investigation presents mixed clustering algorithm named HDG-Clustering for detection. The evaluated using 1999 KDD Cup set. Experimental results indicate approach outperforms several existing techniques.