Equivocable oblivious transfer

Abstract

We analyze and enhance Oblivious Transfer (OT) protocols to accommodate security against adaptive attacks. Previous analysis has been static in nature, treating the security of Alice and the security of Bob as separate cases, determined in advance. It remains unclear whether existing protocols are provably secure against adaptive attacks, but we provide enhancements to make them provably secure against attacks by adaptive 1-adversaries, who can choose at any time whether to corrupt Alice or Bob. We determine circumstances under which OT can be ex- ecuted “in the open,” without encrypting the messages, thereby giving simple alternatives to encrypting an entire interaction. We isolate equivocation properties that provide enough flexibility for a simulator to handle adaptive attacks. These properties also provide a means for classifying OT protocols and understanding the subtle demands of security against adaptive adversaries, as well as designing protocols that can be proven secure against adaptive attacks.