Making secure computation practical – challenges and applications

17.09.2015, 15:00 – 16:30

2015/09/17 15:00-16:30

Speaker: Daniel Demmler | Location: Mornewegstraße 32 (S4|14), Room 3.1.01, Darmstadt

Organizer: Giorgia Azzurra Marson / Nina Bindel

Abstract

Secure computation allows multiple parties to securely evaluate functions on their private inputs, while revealing nothing about those inputs, except what can be inferred from the function's output. Since the first theoretical results in the 1980s, secure computation has moved towards becoming practical due to many algorithmic optimizations and efficient implementations. One stepping stone on the way to truly practical secure computation is our open source ABY framework (NDSS'15), that allows to implement and evaluate secure computation protocols with different secure computation schemes and conversions between them. ABY abstracts from the underlying protocol to ease development and is designed to use the most efficient building blocks available.

In order to further improve usability and performance of secure computation we follow the approach of TinyGable (S&P'15) and generate Boolean circuits for secure computation from hardware description languages. We interface hardware synthesis tools with the ABY framework and compare the performance of the GMW protocol with Yao's garbled circuits for several example functionalities. We also extend the ABY framework with secure floating point operations, thus allowing novel applications that require arithmetic operations with arbitrary precision (to appear at CCS’15).

An interesting example of applications of secure computation is secure inter-domain routing with BGP. With the use of secure computation we are able to overcome several deficiencies of the BGP protocol while at the same time offering privacy for the domain's inputs. When evaluated on real world BGP data (>50k autonomous systems), our protocol computes routes from all domains to a given destination in 3 seconds (plus 9 seconds of independent precomputation).

Short Bio

Daniel graduated with a M.Sc. in Informationssystemtechnik from TU Darmstadt in July 2013. Since October 2013 he is working as a PhD candidate in the Encrypto group lead by Dr. Thomas Schneider. His field of research focuses on the design, implementation and evaluation of secure computation and practical private-preserving protocols.