40 likes | 229 Views
Privacy-Preserving Multi-Domain Data Aggregation. Prof. Dr. Bernhard Plattner (ETH Zurich) Joint work with Martin Burkhart, Xenofontas Dimitropoulos (ETH Zurich), Douglas Dykeman, Thomas Locher (IBM Research Zurich) SWINOG Meeting, Bern, November 11, 2010. Total UDP flow count per 5’.
E N D
Privacy-Preserving Multi-Domain Data Aggregation Prof. Dr. Bernhard Plattner (ETH Zurich) Joint work with Martin Burkhart, Xenofontas Dimitropoulos (ETH Zurich), Douglas Dykeman, Thomas Locher (IBM Research Zurich) SWINOG Meeting, Bern, November 11, 2010
Total UDP flow count per 5’ The Problem • Useful multi-domain statistics, e.g., • Bytes/packets/flows per second and port • Port histograms • Dynamics of common heavy-hitters (e.g., 20 IP addresses/ports with most traffic) • Common IDS alerts • Benefits for participants • Interpret local traffic knowing global conditions • Detect anomalies earlier, with better accuracy • Easier root cause analysis (distributed/local? How much are others affected?) • However: Privacy concerns prevent multi-domain network monitoring • Data protection legislation • Company business/security policies Aggregate view Skype Outage (August 2007) Privacy-Preserving Multi-Domain Flow Analyzer
The Solution • SEPIA: Uses secure multiparty computation (MPC) • Semi-honest adversary model • Majority of sites honest input data confidential Privacy-Preserving Multi-Domain Flow Analyzer
The Offer • What you get • Free TNPFA appliance for your network • Multi-domain aggregrate traffic statistics • A chance to bring in your specific needs and suggestions • What we get • Evaluation in real-world scenarios • Prototype testing • Your feedback http://www.sepia.ee.ethz.ch Privacy-Preserving Multi-Domain Flow Analyzer