D2.2.2 - Advances in Secure Multiparty Protocols

The report contains an overview of the results of the second year of UaESMC, pertaining to secure
multiparty computation techniques. The studies of these techniques have been directed by the example problems selected during the rst year, as well as by the desire to have a comprehensive framework of privacy-preserving computation techniques by the end of the project.

In this deliverable, we report of the following findings and advances:

  • We provide improved privacy preserving algorithms for giving an overview of the data. We also give privacy preserving versions of algorithms for several most common statistical tests. As result, we were able to conduct a full-scale experimental statistical study so that confidential data were always processed using SMC. The strengths of our solution are generality, precision and practicality. We show that secure multi-party computation is flexible enough for implementing complex applications.
  • We have found that the class of techniques currently used for problem transformation based solving of linear programming tasks cannot be privacy-preserving. This leaves the implementations of standard LP-solving algorithms on top of generic protocol sets for privacy-preserving arithmetic as the only general method for privacy-preserving LP, unless some radically new ideas for transforming LP problems are proposed.
  • We provide efficient algorithms for privacy-preserving finite automata execution, that achieve online efficiency through oine precomputations.
  • We provide a protocol set for actively-secure two-party computation that also acheives efficiency through offline precomputations.
  • We provide a protocol transformation that turns any passively secure multiparty computation protocol with honest majority to a protocol where any misbehaviour is detected after the execution.

D2.2.2 - Advances in Secure Multiparty Protocols