Section: New Results

Management of large distributed systems

Non-interference in partial order models

Participant : Loïc Hélouët.

We obtained new results on security issues such as non-interference [41]. Noninterference (NI) is a property of systems stating that confidential actions should not cause effects observable by unauthorized users. Several variants of NI have been studied for many types of models but rarely for true concurrency or unbounded models. In  [45], we had already demonstrated the discriminating power of partial orders, and investigated NI for High-level Message Sequence Charts (HMSCs), a partial order language for the description of distributed systems. We had proposed a general definition of security properties in terms of equivalence among observations of behaviors, and showed that equivalence, inclusion, and NI properties were undecidable for HMSCs. We defined a new formalism called partial order automata, that captures natural observations of distributed systems, and in particular observations of HMSCs. It generalizes HMSCs and permits assembling partial orders. We have then considered subclasses of partial order automata and HMSCs for which Non-Interference is decidable. This allowed us to exhibit more classes of HMSCs for which NI is decidable. Finally, we have defined weaker local Non-interfernce properties, describing situations where a system is attacked by a single agent, and shown that local NI is decidable. We have then refined local NI to a finer notion of causal NI that emphasizes causal dependencies between confidential actions and observations and extended it to causal NI with (selective) declassification of confidential events, which allows to consider that confidential actions need can be kept secret for a limited duration and can then be declassified. Checking whether a system satisfies local and causal NI and their declassified variants are PSPACE-complete problems.

Simulations for stochastic abstractions of large systems

Participants : Éric Fabre, Blaise Genest, Matthieu Pichené.

In [32], we developped a new simulation strategy to accurately simulate DBNs (Dynamic Bayesian Networks) obtained as stochastic abstractions of large systems. The DBN abstractions are given under the form of probability tables, describing the probability for a variable to take a given value given the values of some variables at the previous time point. To be able to handle large systems with many variables, there is a table for each variable (coupling between variable is not explicitely represented). This creates discrepencies when simulating variables independently. Our new algorithm simulates tuples of variables together by looking ahead for such discrepencies in order to avoid them. Such simulations are still efficient, and match more faithfully the original systems.