Critical infrastructures such as energy supply, medical systems or traffic control, which are described in modern cyber-physical systems (CPS), are inevitably becoming increasingly complex. Traditional methods to ensure the operational safety of these complex systems through calculations, modelling and analytical procedures in advance have to fail due to the increasing complexity of the systems themselves and the multitude of unpredictable inputs from outside.
PYRATE develops an intelligent, learning system for the analysis of CPS. A system of software agents is used, which adapts itself fully automatically to the CPS only by a description of the existing sensors and actuators, which is represented in the investigation by a so-called digital twin. The PYRATE software technology independently develops a model of the system; the name of the multiform polymorphic agents derives from this ability to adapt using the interfaces. These agents coordinate to find a weak point, where the subdomains of the whole system work within nominal parameters, but the whole system is destabilized by emergent effects in the interaction of the domains.
In particular, so-called attackers, i.e. market players who use "loopholes" in regulations, are the aim of the analysis strategy. PYRATE enables experts to close these loopholes, which would not have been noticeable in the traditional view of a CPS. The attackers are also confronted with AI defenders who are supposed to keep the system reliable. They learn their strategy for maintaining operational security directly from the attackers.