Abstract: Anomaly detection is a key challenge in data mining which has a wide range of applications in the field of the Internet, including network security, image recognition and intelligent operation. In particular, intelligent operation has made great progress in recent years. Existing anomaly detection algorithms have many problems, such as low accuracy and inability to update automatically. The problem of anomaly detection in the context of intelligent operation and a practical need for high-accuracy, online and universal anomaly detection algorithms is studied. Based on the existing algorithms, an online distance-based anomaly detection algorithm is identified. Through the experiments on Yahoo Web-scope S5 dataset it is shown that the algorithm can detect anomalies successfully. A comparative study of several anomaly detectors verifies the effectiveness of the proposed algorithm.

Key words: anomaly detection, time series, online algorithm, euclidean distance, intelligent operation