This project leverages current advances in network hardware and sophisticated data streaming algorithms to solve challenging internet problems involving network management and network attacks. Content addressable memories are commodity hardware that is ubiquitous in network routing hardware due to their efficient, constant time searching abilities. This project exploits these properties of content addressable memories to solve diverse on-line data stream management problems with challenging analysis and security applications. In particular, the results of this project will include: (1) content addressable memory conscious algorithms for the on-line analysis of data streams and the detection of different network attacks, including distributed denial of service attacks and fraud in advertisement networks; and (2) a functioning implementation of these algorithms on a network processor with content addressable memories, thus demonstrating the feasibility and efficiency of the approach.
The research results will have significant impact in resolving many of the problems facing many internet applications: on-line statistics can provide the basis for efficient network management, as well as for detecting denial of service attacks and fraud in advertisement networks, which disrupts the everyday usage of many internet applications. Funds from this project support the training of one Ph.D. student. A new graduate research seminar covering databases, networks, hardware and security will be introduced into the curriculum. Publications, technical reports and experimental data from this research will be disseminated via the project website (www.cs.ucsb.edu/~dsl/hardware.html).
