Trust-Enabling Middleware (TEM)

点击翻译按钮获取中文摘要

The overall objectives of the project are to enhance the self-x properties of an organic computing middleware (self-configuration, self-optimization, and self-healing) with trust and to protect the trust metrics itself against threats (self-protection). The basic self-x algorithms of a current organic computing middleware are based on the cooperation of autonomic components assuming that the components can "trust" each other. The dynamic nature of an OC system, however, may lead to unpredictable changes of the system configuration by newly added, failed or soon failing components. It is the target of this research project to develop trust metrics to improve the self-x algorithms of the Organic Computing Middleware OCµ that was developed in the DFG priority program 1183 "Organic Computing". Important services should be executed preferably on reliable nodes resulting in a more robust system that is able to handle unreliable components. Trust metrics (direct trust, reputation, and the combination of both) were developed in the first two project phases. The metrics will be applied to self-configuration and self-optimization until the end of the second phase. The increase of robustness will be evaluated and the developed techniques built into the Trust-Enabling Middleware (TEM), which is the middleware used by all other researchers of the research group.The third phase targets trust-improved self-healing as well as self-protection algorithms. Selfhealing will be regarded in two steps: proactive and reactive. In the proactive step an upcoming node failure should be predicted and trigger a reconfiguration based on the developed selfconfiguration algorithms before the failure takes place. In the second step, when a node failure is reported, the concerned services will be automatically restarted on other nodes. In case of restart, the trust metrics will be used for a robust execution of the important services. Because self-x properties are now based on trust, the metrics themselves will be protected against attacks (selfprotection).As additional objective, metrics should be enabled to handle hierarchical systems as well which requires an adaption of the currently used metrics. Moreover the problem of contradictory metrics values will be investigated. Locality effects (e.g. a wall between two nodes) could result in different trust values of two nodes about the same third node and yield conflicting reactions of the self-x mechanisms. The trust enhanced self-x properties will again be implemented in the Trust-Enabling Middleware (TEM) and made available as open source software at the end of the project.