Generalized probabilistic satisfiability and applications to modelling attackers with side-channel capabilities

We analyze a generalized probabilistic satisfiability problem (GenPSAT) which consists in deciding the satisfiability of linear inequalities involving probabilities of classical propositional formulas. GenPSAT is proved to be NP-complete and we present a polynomial reduction to Mixed-Integer Programming. Capitalizing on this translation, we implement and test a solver for the GenPSAT problem. As previously observed for many other NP-complete problems, we are able to detect a phase transition behaviour for GenPSAT. We also describe GGenPSAT, which generalizes GenPSAT by allowing Boolean combinations of linear inequalities involving probabilities of classical propositional formulas which we use to develop applications in information security. Namely, in the context of cryptographic protocols, we model classes of attackers with side-channel capabilities, and study the problem of deciding whether a formula is perfectly masked in the presence of such attackers. (C) 2019 Elsevier B.V. All rights reserved.