Univerself Deliverable D4.16 Assessment Results of Trust in Autonomics - Release 2

In the 1st release of this report we have formulated the key requirements for operator trust in autonomics (TiA) and described the approaches to meet them; those are summarised in the current release and enhanced further. Stability, robustness and security issues arising from future Self-Organising Networks (SONs) must be understood today, and involved in their design, standardisation and certification. We address the issue of Operator trust in autonomic features of their networks (e.g. LTE SON) through the following five requirements and our approaches to meet them: 1. Trust must be measurable (we consider the three facets of operator trust - reliable operation, trustworthy interworking and seamless deployment and suggest a composite metric for SON stability), 2. Trust must be SON-specific (we define a KPI-based envelope of dependable adaptations), 3. Trust must be model-driven (we demonstrate how to construct such models based on predicates), 4. Trust must be propagated end-to-end (we show that trust networks emerge from predicate-enabled behaviours), 5. Trust must be certified (we outline the certification process). Trust predicates that are defined at the design phase as abstract behaviours, and verified at run-time as fully qualified ones, prove to have the power of policies - check them once and re-use many times; rewrite them to cater for new behaviours. The five outlined approaches were integrated in the emerging TiA methodology, which is exemplified by the creation of raw certificates for a couple of sample Network Empowerment Mechanisms also developed in the project. The methodology requires further work; we report here some fundamental issues and consider their discovery and identification as one of the main achievements of the TiA task in the project. Yet we were able to demonstrate that our approaches are viable when the problem of trust is defined with sufficient precision. First, we limit the understanding of trust only to Operator Trust in Automation, where automation is applied to certain management functions traditionally performed/controlled by humans. Second, we demonstrated that our five approaches to trust facets make trust being operationally defined in Deming's sense This definition precision did allow us to succesfully create a relevant Work Item within ETSI, where further work on certification proceeds. The second major enhancement as compared to the 1st releas is the inclusion of business aspect of certification. We report here our findings on possible certification scope, discuss the assumptions and requirements, and describe from a business impact point of view a number of certification scenarios such as offline and online certification, self-certification and 3rd party certification.