中文摘要：

云存储平台是多租户共享环境，能否实现其中不同租户数据之间的有效安全隔离成为了用户最为关心的问题．以RBAC（Role Based Access Control）策略为基础，结合组织标签和多种安全属性的逻辑组合，提出一种灵活的访问控制策略，它一方面保证云端不同企业之间数据的强隔离性，使某企业用户无法越权访问其他企业的用户数据；另一方面保证云存储企业内部数据的适度隔离，即可以根据公司自身的安全需求灵活定制企业内部策略．同时，引入虚拟组织的概念实现企业之间可能的数据共享；引入利益冲突的概念限制竞争企业之间的共享．给出了该策略在基于HDFS（Hadoop Distributed File System）的云存储架构中的原型实现．实验结果表明，该策略能够有效保障云存储平台多租户数据之间的恰当安全隔离．

英文摘要：

Cloud Storage is a multi-tenancy shared environment, so achieving data separation between different users effectively in the platform has become one of issues most concerned by users. In this paper, we provide business users a flexible access control policy, which is built on top of RBAC （Role-basedAccess Control）, combined with organization label and a variety of security attributes with logical combinations. First of all, it provides strict inter-enterprise data isolation on cloud storage, ensuring that business users could not access data which doesnt belong to their organization. Moreover, it provides proper separation of organization internal data. Business users could customize the policy flexibly according to their own security requirements, isolating data from different sectors or geographical area. Finally, the policy provides a mechanism for corporations to share data on cloud storage by introducing the concept of ＂virtual organization＂, and guarantee companies with the same conflict set of interest could not be allowed to share data through traditional Chinese Wall Policy. This paper presents the design and implementation of a prototype based Hadoop distributed file system { I-IDFS）, including security label, security policy, security decision module, enforcement procedure of security decision and user command interface. Then it analyzes the effectiveness and performance of the security mechanisms with experiments. The result shows that the policy meets the security requirement well and loss of system run-time performance is within an acceptable range.