Deterministic encryption with AES SIV and where it’s indispensable
When it comes to symmetric block ciphers, there are several modes of operation. These differ in how the ciphertext that exceed the block size of the cipher is formed. With AES, for example, the block size is 128 bits. Without going into too much detail, it's important to know that you can't simply concatenate the ciphertexts produced by a block cipher without leaking plaintext data.
Most modes of operation rely on a so-called initialization vector or IV, which is a fixed-size byte input that (as opposed to the encryption key) doesn't need to be secret but is required to decrypt a ciphertext. There are different ways to generate such an IV. The most common are simply random IVs. However, randomness means that encrypting the same cleartext twice will result in two different ciphertexts. For many applications this is a desirable property, as an attacker wouldn't be able to determine whether two ciphertexts correspond to the same cleartext. However, for certain scenarios we strictly want a given cleartext to always encrypt to the same ciphertext. In this case we speak of deterministic encryption.
One mode of operation that allows such deterministic encryption is the SIV mode, specified in RFC 5297. SIV stands for synthetic initialization vector and works by deterministically deriving an IV from the input during encryption. The IV is then prepended to the ciphertext. During decryption it can then be used to validate the cleartext and therefore notice whether the ciphertext has been tampered with. This property makes it a so-called authenticated encryption algorithm.
One field of application of this algorithm is deterministic filename encryption, as used in the Cryptomator cloud encryption utility. Since Cryptomator is an open source project which has its encryption code published as separate libraries, other software can easily integrate it. Connect2id employed the AES SIV implementation in the OpenID Connect SDK to generate deterministic encrypted (pairwise) user identifiers, for increasing privacy in single sign-on.
Example user ID encryption with AES SIV:
user ID: alice encrypted: Osy9fFgs2PRJAk9JE0-4d3kIvxMFOLpoyX_yVZZmkG4r
The open source nature of the Cryptomator AES SIV library allows for security audits to be performed. One was undertaken by Tim McLean on behalf of Connect2id and its finds were published on the project page. All found issues have been successfully resolved in close collaboration with the Cryptomator team.
The Cryptomator AES SIV library for Java can be found on GitHub: