Proxy re-encryption (PRE) allows a proxy to convert a ciphertext encrypted under one key into an encryption of the same message under another key. The main idea is to place as little trust and reveal as little information to the proxy as necessary to allow it to perform its translations. At the very least, the proxy should not be able to learn the keys of the participants or the content of the messages it re-encrypts. However, in all prior PRE schemes, it is easy for the proxy to determine between which participants a re-encryption key can transform ciphertexts. This can be a problem in practice. For example, in a secure distributed file system, content owners may want to use the proxy to help re-encrypt sensitive information without revealing to the proxy the identity of the recipients. In this work, we propose key-private (or anonymous) re-encryption keys as an additional useful property of PRE schemes. We formulate a definition of what it means for a PRE scheme to be secure and k...