Sampling-Based Stream Mining for Network Risk Management

Network security is an important issue in maintaining the Internet as an important social infrastructure. Especially, finding excessive consumption of network bandwidth caused by P2P mass flow, finding internet viruses, and finding DDoS attacks are important security issues. Although stream mining techniques seem to be promising techniques for network security, extensive network flow prevents the simple application of such techniques. Since conventional methods require non-realistic memory resources, a mining technique which works well using limited memory is required. This paper proposes a sampling-based mining method to achieve network security. By analyzing the characteristics of the proposed method with real Internet backbone flow data, we show the advantages of the proposed method, i.e. less memory consumption.