|Abstract:||We have built and deployed a system called Folklore that implements a new property called data untraceability through a proactive migratory replication protocol in a large-scale distributed setting. A data untraceable object, i.e., an object whose replicas cannot be located, is probabilistically difficult to delete from the network. Folklore is a proof-of-concept system that allows data untraceability to be provided hand-in-hand with data availability, updatability, integrity, and scalability. Folklore consists of two subsystems - a pure "Proactive Folklore" system, and a "Reactive Folklore" system (R-Folklore). R-Folklore is a distributed file storage system that makes files untraceable on-demand at run-time, e.g., when a file is sensed to be under attack through intrusions. Intuitively, R-Folklore stealthily turns vulnerable files into an untraceable form. We have deployed Folklore on the CSIL cluster at UIUC (our undergraduate laboratory). We present experimental results from the deployment, as well as simulation experiments that match deployment results but extend to larger system sizes.