This paper introduces lronMask, a new versatile verification tool for masking security. lronMask is the first to offer the verification of standard simulation-based security notions in the probing model as well as recent composition and expandability notions in the random probing model. It supports any masking gadgets with linear randomness (e.g. addition, copy and refresh gadgets) as well as quadratic gadgets (e.g. multiplication gadgets) that might include non-linear randomness (e.g. by refreshing their inputs), while providing complete verification results for both types of gadgets. We achieve this complete verifiability by introducing a new algebraic characterization for such quadratic gadgets and exhibiting a complete method to determine the sets of input shares which are necessary and sufficient to perform a perfect simulation of any set of probes. We report various benchmarks which show that lronMask is competitive with state-of-the-art verification tools in the probing model (maskVerif, scVerif, SILVEH, matverif). lronMask is also several orders of magnitude faster than VHAPS -the only previous tool verifying random probing composability and expandability- as well as SILVEH -the only previous tool providing complete verification for quadratic gadgets with nonlinear randomness. Thanks to this completeness and increased performance, we obtain better bounds for the tolerated leakage probability of state-of-the-art random probing secure compilers.

In this paper, we propose a constant-time implementation of the BLISS lattice-based signature scheme. BLISS is possibly the most efficient lattice-based signature scheme proposed so far, with a level of performance on par with widely used pre-quantum primitives like ECDSA. It is only one of the few postquantum signatures to have seen real-world deployment, as part of the strongSwan VPN software suite. The outstanding performance of the BLISS signature scheme stems in large part from its reliance on discrete Gaussian distributions, which allow for better parameters and security reductions. However, that advantage has also proved to be its Achilles' heel, as discrete Gaussians pose serious challenges in terms of secure implementations. Implementations of BLISS so far have included secret-dependent branches and memory accesses, both as part of the discrete Gaussian sampling and of the essential rejection sampling step in signature generation. These defects have led to multiple devastating timing attacks, and were a key reason why BLISS was not submitted to the NIST postquantum standardization effort. In fact, almost all of the actual candidates chose to stay away from Gaussians despite their efficiency advantage, due to the serious concerns surrounding implementation security. Moreover, naive countermeasures will often not cut it: we show that a reasonable-looking countermeasure suggested in previous work to protect the BLISS rejection sampling can again be defeated using novel timing attacks, in which the timing information is fed to phase retrieval machine learning algorithm in order to achieve a full key recovery. Fortunately, we also present careful implementation techniques that allow us to describe an implementation of BLISS with complete timing attack protection, achieving the same level of efficiency as the original unprotected code, without resorting on floating point arithmetic or platform-specific optimizations like AVX intrinsics. These techniques, including a new approach to the polynomial approximation of transcendental function, can also be applied to the masking of the BLISS signature scheme, and will hopefully make more efficient and secure implementations of lattice-based cryptography possible going forward.

Differential power analysis (DPA) is a side-channel attack in which an adversary retrieves cryptographic material by measuring and analyzing the power consumption of the device on which the cryptographic algorithm under attack executes. An effective countermeasure against DPA is to mask secrets by probabilistically encoding them over a set of shares, and to run masked algorithms that compute on these encodings. Masked algorithms are often expected to provide, at least, a certain level of probing security. Leveraging the deep connections between probabilistic information flow and probing security, we develop a precise, scalable, and fully automated methodology to verify the probing security of masked algorithms, and generate them from unprotected descriptions of the algorithm. Our methodology relies on several contributions of independent interest, including a stronger notion of probing security that supports compositional reasoning, and a type system for enforcing an expressive class of probing policies. Finally, we validate our methodology on examples that go significantly beyond the state-of-the-art.