Scalable and Efficient Data Streaming Algorithms for Detecting Common Content in Internet Traffic

Abstract

Recent research on data streaming algorithms has provided powerful tools to efficiently monitor various characteristics of traffic passing through a single network link or node. However, it is often desirable to perform data streaming analysis on the traffic aggregated over hundreds or even thousands of links/nodes, which will provide network operators with a holistic view of the network operation. Shipping raw traffic data to a centralized location (i.e., "raw aggregation") for streaming analysis is clearly not a feasible approach for a large network. In this paper, we propose a set of novel Distributed Collaborative Streaming (DCS) algorithms that allow scalable and efficient monitoring of aggregated traffic without the need for raw aggregation. Our algorithms target the specific network monitoring problem of finding common content in the Internet traffic traversing several nodes/links, which has applications in network-wide intrusion detection, early warning for fast propagating worms, and detection of hot objects and spam traffic. We evaluate our algorithms through extensive simulations and experiments on traffic traces collected from a tier-1 ISP. The experimental results demonstrate that our algorithms can effectively detect common content in the traffic traversing across a large network.