CS Events

Computer Science Department Colloquium

Price of Active Security in Multiparty Computation

 

Download as iCal file

Monday, March 09, 2020, 10:30am

 
3_9_2020_muthuhp.jpg

Speaker:  Muthu Venkitasubramaniam, University of Rochester

Bio

Muthu Venkitasubramaniam is an Associate Professor at the University of Rochester. He received his B.Tech degree in computer science from the Indian Institute of Technology, Madras in 2004 and then his PhD from Cornell in 2011. He spent a year at the Courant Institute as a postdoc researcher under the CIF, and then joined the faculty at Rochester where he is currently an associate professor.  He is a recipient of the GFRA and the ICDE Influential Paper Award.

Location : CoRE A 301

Event Type: Computer Science Department Colloquium

Abstract: Secure multiparty computation (MPC) allows a group of mutually distrusting parties to collaborate and compute a function jointly over their individual data while guaranteeing maximal privacy. The last decade has seen pivotal developments in cryptography that have enabled dozens of built systems demonstrating novel privacy use cases for this technique.The strongest type of MPC protocol guarantees security against malicious participants who may arbitrarily deviate from the protocol. This is often called “active” security. Designing MPC with active security involves two steps: (1) building a protocol with (weak) passive security, i.e. tolerating only honest-but-curious participants, and then, (2) amplifying to obtain active security. While the price of achieving active security has been well understood in theory, all practical approaches incur a significant communication and computation overhead (20-100x) over the best practical MPC protocols with passive security.In this talk, I will survey my works focused on theoretical and practical advancements towards designing MPC with active security. In particular, I will discuss a recent work where I introduce a new paradigm that allows implementing MPC with active security whose communication is roughly twice the cost of the best protocol with passive security. I will further show how this approach can lead to constructions with essentially no communication overhead by relating the efficiency to certain “leakage-resilience” properties of error-correcting codes. As a culmination in this line of work, I will showcase a recent application for generation of RSA moduli in an MPC that is slated to be used by the Ethereum foundation to bootstrap their next consensus protocol. Using our techniques, we have designed and executed a protocol for n=10,000 participants (the largest known MPC execution) and achieves active security even when all-but-one of the parties are malicious.

Contact  Faculty Host: Shubhangi Saraf