Alex
Norton,
Master’s
candidate
David
R.
Cheriton
School
of
Computer
Science
Secure multiparty computation allows two or more parties to compute a function without leaking unnecessary information about their inputs to other parties. In traditional secure multiparty computation protocols, the function is represented as a circuit and each gate of the circuit is securely computed. The mixed mode model is a generalization, where only some gates are computed securely, and other gates are computed in a local, unsecured manner. There are computations where mixed mode protocols are known to be just as secure and much more efficient, and so it is natural to ask if it is possible to automatically construct optimized mixed mode secure protocols for a given function.
Previous results describe powerful compilation techniques to transform circuits into efficient mixed mode protocols, but the results are only secure against very restricted (passive) adversaries. These passively secure protocols can be secured against active adversaries using extensions of classic secure multiparty computation compilation techniques. However, this comes with a significant loss of concrete efficiency which negates the mixed mode efficiency advantages.
I will present novel techniques that can efficiently compile mixed mode two party protocols from passive to active security. The techniques exploit structural properties of the underlying circuits to reduce the overhead of compilation without compromising the security. The gain in efficiency varies based on the circuit that is being compiled, and although for some circuits the techniques will yield no gains, for others the resulting secure protocols have exponentially lower computation and communication cost.