Dealing with the security behaviour of large scale systems

This paper describes a risk assessment method suited for large systems. In essence, the method has been introduced in previous works [5, 8, 9, 10, 27], where its properties have been analysed. In this paper, we develop and mathematically justify a variant of it which allows to divide a large system into overlapping subsystems, each one analysed by an expert. We will show that, independently from the division strategy, there is an effective way to combine the experts’ assessments into a global picture, as far as a few and natural hypotheses on the metrics are satisfied.