| （英） |
In recent growth of ICT, especially in Universities, the enhancement of education by ICT such as remote lectures and the increase of BYODs (Bring Your Own Devices) of students, leads the importance of campus network as infrastructure of educational and/or research activities. On the other hand, the spread of BYODs, which are not suitable to be contralled centrally, may degrade safety and stability of network bahavior due to anomally traffic from such devices. It is therefore important to realize a detection of anomalies at the very early stage. However, to follow the behavior of campus network accurately, it is necessary to collect huge log data from network switches, so a light-weight method to detect the event of anomalies is promising for real-time detection. In this study we consider a method to detect anomalies based on traffic patterns of network switches using a learning-based approach, and perform a feasilibity through numerical evaluations. |