An approach for cross-domain access control policy generation from reputation evaluation to trust management

In view of existing access control mechanisms having coarse granularity and being lack of dynamic adaptability in cross-domain open environment with a lot of strange entities, an access control policy generation approach from reputation evaluation up to trust management is proposed. This approach can generate finegrained authorization policies automatically and dynamically according to entities' behaviour features and context attributes, which is achieved by integrating reputation evaluation and trust management in an original data mining based way. Such integration makes the approach surpass the limitations of existing integration models, in which reputation evaluation is only subjectively taken as an extension to trust management without overcoming the pre-set policies' limited ability to adapt to dynamic environment. The reputation evaluation combines cloud model and Bayesian networks, which can represent and evaluate the uncertainty of trust more accurately and efficiently. Then the association relationships between entities' attributes and their reputation are extracted from reputation evaluation results. Finally, these relationships will be transformed into attribute based access control policies. Simulation results show that the entities' behaviour features can be automatically mapped into access control policies which can better adapt to cross-domain dynamic environment.