Sheila Becker, Securing Application-Level Topology Estimation Networks: Facing the Frog-Boiling Attack

Peer-to-peer real-time communication and media streaming applicationsoptimize their performance by using application-level topology estimationservices such as virtual coordinate systems. Virtual coordinate systems allownodes in a peer-to-peer network to accurately predict latency between arbitrarynodes without the need of performing extensive measurements. However, systemsthat leverage virtual coordinates as supporting building blocks, are prone toattacks conducted by compromised nodes that aim at disrupting, eavesdropping,or mangling with the underlying communications.Recent research proposed techniques to mitigate basic attacks (inflation, deflation,oscillation) considering a single attack strategy model where attackers performonly one type of attack. In this work we explore supervised machine learningtechniques to mitigate more subtle yet highly effective attacks (frog-boiling,network-partition) that are able to bypass existing defenses. We evaluate ourtechniques on the Vivaldi system against a more complex attack strategy model,where attackers perform sequences of all known attacks against virtual coordinatesystems, using both simulations and Internet deployments. About the speaker: Sheila is a Ph.D. candidate at the University of Luxembourg. She is a visiting scholar at Purdue University in the Dependable and Secure Distributed Systems Lab for this academic year with the support of a Fulbright grant. In 2008, she obtained the M. Sc. degree from the University Henri Poincare in Nancy, France. She received an industrial engineering degree in applied informatics from University of Luxembourg in 2007. Her interests lie in Machine Learning, Network Security and Fuzz testing.