Paul PERROTIN thesis defense

Nowadays, the construction of systems is mostly done via the interconnection of already existing systems in which the human being holds such a place that she/he is  considered as an already existing system too. This is called a system of socio-technical systems (SoSTS). In engineering, one of the main current problems is the secure design of this type of systems. Indeed, nowadays the human is the main vector used by cybernetic attackers. In this thesis, we propose a methodology allowing the detection of the human vulnerability within an architecture of a SoSTS. More specifically, we have specified a metamodel to capture human factors that have a potential impact on human vulnerability within a SoSTS. For that aim, we propose a language, called HoSML, in order to model an architecture of human operators composing a functional chain of a SoSTS. Based on the models produced with that language, we have defined a method for estimating human vulnerability and and its impacts on the SoSTS. This method is based on a probabilistic approach, nemelly Bayesian network. Finally, we have implemented this approach, and the associated language, as a software called HoS-ML Editor. We have validated our approach using industrial case studies and we have submitted the results to industrial experts.This allowed us to test, discuss and estimate the limits of our approach.