Another Nail in the Coffin of White-Box AES Implementations

The goal of white-box cryptography is to design implementations of common cryptographic algorithm (e.g. AES) that remain secure against an attacker with full control of the implementation and execution environment. This concept was put forward a decade ago by Chow et al. (SAC 2002) who proposed the first white-box implementation of AES. Since then, several works have been dedicated to the design of new implementations and/or the breaking of existing ones. In this paper, we describe a new attack against the original implementation of Chow et al. (SAC 2002), which efficiently recovers the AES secret key as well as the private external encodings in complexity 2. Compared to the previous attack due to Billet et al. (SAC 2004) of complexity 2, our attack is not only more efficient but also simpler to implement. Then, we show that the last candidate white-box AES implementation due to Karroumi (ICISC 2010) can be broken by a direct application of either Billet et al. attack or ours. Specifically, we show that for any given secret key, the overall implementation has the exact same distribution as the implementation of Chow et al. making them both vulnerable to the same attacks. By improving the state of the art of white-box cryptanalysis and putting forward new attack techniques, we believe our work brings new insights on the failure of existing whitebox implementations, which could be useful for the design of future solutions.