The re-keying scheme is a variant of the symmetric encryption scheme where a sender (respectively, receiver) encrypts (respectively, decrypts) plaintext with a temporal session key derived from a master secret key and publicly-shared randomness. It is one of the system-level countermeasures against the side channel attacks (SCAs), which make attackers unable to collect enough power consumption traces for their analyses by updating the randomness (i.e., session key) frequently. In 2015, Dobraunig et al. proposed two kinds of re-keying schemes. The first one is a scheme without the beyond birthday security, which fixes the security vulnerability of the previous re-keying scheme of Medwed et al. Their second scheme is an abstract scheme with the beyond birthday security, which, as a black-box, consists of two functions; a re-keying function to generate a session key and a tweakable block cipher to encrypt plaintext. They assumed that the tweakable block cipher was ideal (namely, secure against the related key, chosen plaintext, and chosen ciphertext attacks) and proved the security of their scheme as a secure tweakable block cipher. In this paper, we revisit the re-keying scheme. The previous works did not discuss security in considering the SCA well. They just considered that the re-keying scheme was SCA resistant when the temporal session key was always refreshed with randomness. In this paper, we point out that such a discussion is insufficient by showing a concrete attack. We then introduce the definition of an SCA-resistant re-keying scheme, which captures the security against such an attack. We also give concrete schemes and discuss their security and applications.
Mathematical Backdoors in Symmetric Encryption Systems - Proposal for a Backdoored AES-like Block Cipher