Section: New Results

Static Analysis of Programs with Imprecise Probabilities

Participant : Jean Goubault-Larrecq [correspondant] .

Static analyses allows one to obtain guarantees about the behavior of programs, without running them. Programs that handle numerical data such as feedback control loops pose a challenge in this area. This gets even harder when one considers programs that read numerical data from sensors, and write to actuators, as these data are imprecise, and are governed by probability distributions that may themselves be unknown, and only know to fall into some interval of distributions.

As part of the ANR projet blanc CPP, an efficient static analysis framework that deals with this kind of programs was proposed in 2011 by J. Goubault-Larrecq, O. Bouissou, E. Goubault, Sylvie Putot, based on P-boxes and Dempster-Shafer structures to handle imprecise probabilities.

The semantic foundations were made clearer, a new, improved algorithm was proposed, and new applications were examined in:

• A. Adjé, O. Bouissou, J. Goubault-Larrecq, E. Goubault and S. Putot. Static Analysis of Programs with Imprecise Probabilistic Inputs. In VSTTE'13, LNCS. Springer, 2013.