Due to the anonymity and low pay of workers in crowdsourcing platforms, there may be concerns regarding reliability and privacy-preservation when using such platforms to deliver services. This paper describes a technique for jointly providing privacy and reliability through stochastic perturbation of microtask definitions and fusion rules to combine the work of several workers. A mathematical model of a crowdsourcing system using this technique is proposed and precise threshold conditions on loss of privacy when workers collude are provided. Tradeoffs between privacy, reliability, and cost are determined. © 2012 IEEE.