We observe that being a "pure capability" protocol, the plain vanilla OSD protocol is incapable of enforcing confinement. We show, however, that given a trustworthy infrastructure for authentication and secure channels, the protocol can be used in a manner that achieves the desired property (and does not require any change in the message format). Thus we demonstrate that object stores can in principle be used in a standard fashion in applications that require protection against leakage of secret data.
Having identified a problem and proposed a solution, we proceed to prove formally that the proposed protocol indeed meets all its security goals. In the process we refine common cryptographic models in order to be able to reason about confinement, and then devise a precise model for a distributed capability-based access-control mechanism. To our knowledge, this is the first time such a model for access-control is defined in a cryptographic setting, and defining it highlights what can and cannot be achieved by such mechanisms.Category / Keywords: applications / foundations, access-control, adaptive corruptions Date: received 7 Jun 2005, last revised 8 Jun 2005 Contact author: shaih at alum mit edu Available formats: Postscript (PS) | Compressed Postscript (PS.GZ) | PDF | BibTeX Citation Version: 20050608:130656 (All versions of this report) Discussion forum: Show discussion | Start new discussion