Export to
Thursday, March 27, 2014 at 10:12am.
Galois tech talk: Practical Challenges to Secure Computation
Website
Description
Presented by John Launchbury.
In secure computation, one or more parties collaborate to compute a result while keeping all the inputs private. That is, no-one can gain knowledge about the inputs from the other parties, except what can be determined from the output of the computation. Methods of secure computation include fully homomorphic encryption (where one party owns the input data and the other party performs the whole computation), and secure multiparty computation (where multiple parties collaborate in the computation itself). The underlying methods are still exceedingly costly in time, space, and communication requirements, but there are also many other practical problems to be solved before secure computation can be usable. For programmers, the algorithm construction is often nonintuitive; for compiler writers, the machine assumptions are very different from usual; and for application designers, the application information flow has to match the security architecture. In this talk we will highlight these challenges, and indicate promising research directions.