Information Flow Control Model In Distributed Systems

Distributed systems are widely used in the realms of finance, health care, etc., in which confidentiality and integrity of data is crucially important. And the demand of privacy policies differs between different realms. Nodes in distributed systems mutually distrust each other, and managing authority and privilege centrally and universally leads to the loss of flexibility and efficiency of the system. As a result, research on privacy and security in distributed systems is always an important subject.The proposition of decentralized label model and information flow control brought a new direction of research to solve the security problem in distributed systems. Taking MapReduce as the environment, this paper studies on the privacy and security in distributed systems, and proposes a novel model based on decentralized information flow control to enforce security of distributed systems. And we try to implement the proposed model to verify its feasibility.First of all, we study the exist techniques to enforce privacy security. The disadvantage of putting those techniques directly into distributed systems is why we proposed information flow control model for distributed systems.Secondly, we study the structure of distributed systems, and raised the privacy and security problems in the current distributed systems. We also point out inadequacy of existing security models.Thirdly, analyzing most of existing information flow control models, we propose a novel model to enforce information flow control in distributed systems. By adding information flow control for network communication to the model, we improved the privacy and security of distributed system.Finally, we try to implement our model based on open source distributed systems and information flow control models. And it has a certain degree of practical significance for further search on privacy and security in distributed systems.