Study of Anomaly Detection Method using Hurst Space.

本研究提出了一种新定义的流量特征的表达式来量化互联网流量时间序列的非平稳性，并使用该特性的周期性时间序列的异常检测方法。实验结果表明，该方法能够有效地检测出像长期端口扫描这样周期性的低速率攻击。此外，从支持网络管理者的角度出发，提出了利用Hurst空间增强特征变化的可视性的可视化方法。

This study proposes a newly defined expression of traffic characteristic to quantify a non-stationarity of the Internet traffic time series, and an anomaly detection method for periodic time series using that characteristic as well. The proposed method showed that it was effective for the detection of low-rate attacks having such a cycle as long-term port scan.Furthermore, from the viewpoint of supporting a network manager, we proposed helpful visualization methods to the manager by using the Hurst space to step up the visibility of the characteristic changes.