A Model-based Approach to Certification of Adaptive MILS

In this work, we tackle the problem of certifying Adaptive systems. These are able to automatically perform self-reconfiguration at runtime, which makes classical certification approaches inapplicable. The need for certification approaches for these systems is thus becoming urgent, especially due to their prevalent use in safety- and mission critical settings. Due to the inherent complexity of adaptive systems and the absence of a principled methodology for their construction and assurance, there has been little movement by certification authorities to accept such systems. Among the challenges for certification are a way of generating an adequate assurance case for initial state of the adaptive system and for each step in its incremental adaptation, and generation and management of the evidence upon which the assurance case relies. We contribute in this research by proposing a novel modular approach to the certification of adaptive systems in the context of the Adaptive MILS architecture.
The proposed approach is backed by an Evidential-Tool Bus implementation that allows a continuous on-demand generation of assurance cases.