In a zero-knowledge proof market, we have two sides. On one side, bidders with proofs of different sizes and some private value to have this proof computed. On the other side, we have distributors (also called sellers) which have computational power available to process the proofs by the bidders, and these distributors have a certain private cost to process these proofs (dependent on the size). More broadly, this setting applies to any online resource allocation where we have bidders who desire a certain amount of a resource and distributors that can provide this resource. In this work, we study how to devise double auctions for this setting which are truthful for users, weak group strategy proof, weak budget balanced, computationally efficient, and achieve a good approximation of the maximum welfare possible by the set of bids. We denote such auctions as robust.