An Improvement in Secure Two-party Computation Secure Against Malicious Adversaries

We propose an improvement in the area of malicious-secure two-party computation based on garbled circuit (GC). We build on the updated protocol of Mohassel and Franklin. As the same input must be provided to all s GCs used in the computation, 1/2 s^2 commitments per each input bit are generated and sent to catch the possibly cheating GC constructor. In this work, we reduce the number of commitments by the factor of 2, an important improvement for a mature technology such as GC. This improves the highest-order term in both communication and computation of secure function evaluation (SFE). We validate our constant-factor improvement by giving exact bounds on the adversarial advantage and comparing it to previous work. Finally, we give a full simulation-based proof of security, which also applies to the Mohassel-Franklin protocol.