Cold boot attacks target memory remanence effects in hardware to secret key material. Such attacks were first explored in the scientific literature by Halderman et al. (USENIX Security Symposium 2008) and, since then, different attacks have been developed against a range of asymmetric key and symmetric key algorithms. Such attacks in general receive as input a noisy version of the secret key as stored in memory, and use redundancy in the key (and possibly knowledge of a public key) to recover the secret key. The challenge is to recover the key as efficiently as possible in the face of increasing levels of noise. For the first time, we explore the vulnerability of lattice-based cryptosystems to this form of analysis, focussing in particular on NTRU, a well-established and at- tractive public-key encryption scheme that seems likely to be a strong candidate for standardisation in NIST’s post-quantum process. We look at two distinct NTRU implementations, showing how the attacks that can be developed depend critically on the in-memory representation of the secret key. We develop, efficient, dedicated key-recovery algorithms for the two implementations and provide the results of an empirical eval- uation of our algorithms.
展开▼
