Reusable Formal Models for Secure Software Architectures

Formal modelling techniques are often disregarded as their semantics are too distant from the mainstream practice of software architecture design, which is dominated by the use of component based modelling and patterns. This paper advocates the need for formal modelling techniques for humans, i.e., software architects who need to precisely ascertain the security properties of their design models. We contribute a technique that enables architects to more easily construct verified, secure architecture designs by assembling already verified security pattern models. Our approach is illustrated with a pattern language for accountability. It is validated by an observational study that shows that the approach produces reusable results, and is able to uncover relevant architectural security flaws.