Radio Frequency Identification (RFID) technology raises significant privacy issues because it enables tracking of items and people possibly without their knowledge or consent. One of the biggest challenges for RFID technology is to provide privacy protection without raising tag production and management cost. We introduce a new architecture that uses trusted computing primitives to solve this problem. Our design splits the RFID reader into three software modules: a Reader Core with basic functionality, a Policy Engine that controls the use of RFID-derived data, and a Consumer Agent that performs privacy audits on the RFID reader and exports audit results to third party auditors. Readers use remote attestation to prove they are running a specific Reader Core, Policy Engine, and Consumer Agent. As a result, remote attestation allows concerned individuals to verify that RFID readers comply with privacy regulations, while also allowing the reader owner to verify that the reader has not...