Security principles for public-resource modeling research

Large-scale distributed computing projects have many security concerns due to their public and often "open" nature. Climateprediction.net (CPDN) is taking the concept of public-resource, high-throughput grid computing a stage further, by using it for a major piece of modeling research. The aim is to harness the spare CPU cycles of potentially millions of individual users' PCs to run a massive ensemble of climate simulations. In doing so it has been faced with a range of security and integrity issues beyond those encountered by previous projects but likely also to confront similar initiatives in the future. This paper introduces the project and its software architecture, and outlines a threat model for such situations, including threats to participants, threats to the experiment and threats to other stakeholders. It goes on to discuss how these threats have been addressed, with the procedures presented expected to form a valuable foundation for an increasing number of similar projects.