Distributed Denial of Service Attacks: Analysis of Defenses

Distributed Denial of Service (DDoS) attacks are attempts to overwhelm a computer system in order to deny access by legitimate users. They are generally unstoppable, but there is a good deal of on-going research on methods to reduce their negative effects. This paper will deal with the design of a model that simulates such an attack. The simulation model is then used to study possible ways to defend against these attacks. Three experiments are run: 1) using a priority queue to sort messages from clients based on how many connections they have open on the server; 2) limiting the number of connections each client can create; and 3) having the server forcefully delete the oldest established connection, whenever its connection table becomes full. Results show that method 1 is totally ineffective while method 2 somewhat improves the overall performance of the system. However, method 3, combined with method 2, produces significantly improved performance against a DDoS attack.

Eric Chan-Tin, echantin@alumni.macalester.edu. Ph.D. Computer Science (Expected 2011), Master of Science, Univ. of Minnesota. Research interests in network security, peer-to-peer (P2P) network, anonymity/privacy, botnets, and network coordinate systems; B.A. Computer Science and Mathematics, Macalester College. For a full CV, contact the author.