English

Probabilistic Opacity in Refinement-Based Modeling

Cryptography and Security 2015-10-16 v1

Abstract

Given a probabilistic transition system (PTS) A\cal A partially observed by an attacker, and an ω\omega-regular predicate φ\varphiover the traces of A\cal A, measuring the disclosure of the secret φ\varphi in A\cal A means computing the probability that an attacker who observes a run of A\cal A can ascertain that its trace belongs to φ\varphi. In the context of refinement, we consider specifications given as Interval-valued Discrete Time Markov Chains (IDTMCs), which are underspecified Markov chains where probabilities on edges are only required to belong to intervals. Scheduling an IDTMC S\cal S produces a concrete implementation as a PTS and we define the worst case disclosure of secret φ\varphi in S{\cal S} as the maximal disclosure of φ\varphi over all PTSs thus produced. We compute this value for a subclass of IDTMCs and we prove that refinement can only improve the opacity of implementations.

Keywords

Cite

@article{arxiv.1510.04316,
  title  = {Probabilistic Opacity in Refinement-Based Modeling},
  author = {Béatrice Bérard and Olga Kouchnarenko and John Mullins and Mathieu Sassolas},
  journal= {arXiv preprint arXiv:1510.04316},
  year   = {2015}
}
R2 v1 2026-06-22T11:20:41.053Z