Optimal Contracts for Outsourced Computation

While expensive cryptographically verifiable computation aims at defeating malicious agents, many civil purposes of outsourced computation tolerate a weaker notion of security, i.e., "lazy-but-honest" contractors. Targeting this type of agents, we develop optimal contracts for outsourcing of computational tasks via appropriate use of rewards, punishments, auditing rate, and "redundancy". Our contracts provably minimize the expense of the outsourcer (principal) while guaranteeing correct computation. Furthermore, we incorporate practical restrictions of the maximum enforceable fine, limited and/or costly auditing, and bounded budget of the outsourcer. By examining the optimal contracts, we provide insights on how resources should be utilized when auditing capacity and enforceability are limited. Finally, we present a light-weight cryptographic implementation of the contracts to mitigate the double moral hazard problem between the principal and the agents.