Abstract
We propose an inference prevention agent as a tool that enables each of the databases in a distributed system to keep track of probabilistic dependencies with other databases and then use that information to help preserve the confidentiality of sensitive data. This is accomplished with minimal sacrifice of the performance and survivability gains that are associated with distributed database systems.