Distributed Access Control for Accountable Systems

分布式系统的保障是基于不断加强保护边界，防止和检测对计算机系统的未经授权的访问。问责制，即确保校长对自己的行为负责，是一个同样重要的问题。该项目研究了一种在分布式系统中跟踪责任的编程语言方法，特别是通过扩展Java语言来维护用于责任目的的细粒度审计跟踪。所提出的方法是基于将保密性（访问控制）和完整性信息与通过网络检索和复制的数据相关联。分布式访问控制不仅在访问数据的网络中的某一点上实施，而且可以在访问数据副本的任何地方实施。分布式信息流类型强制执行访问权限，并强制执行一种形式的因果关系，以支持对主体责任的跟踪。由于认识到通信可能在不安全的网络上进行，因此假定敏感数据是“虚拟”签名或加密的，以分别保证完整性和保密性。通过在可能的情况下使用加密类型进行静态检查，避免了无处不在的加密操作应用程序的计算负担。

Assurance for distributed systems is based on constantly strengthening the protective boundaries that prevent and detect unauthorized access to computer systems. Accountability, ensuring that principals are liable for their actions, is an equally important issue. This project investigates a programming language approach to tracking accountability in distributed systems, specifically by extending the Java language to maintain a fine-grained audit trail for accountability purposes.The proposed approach is based on associating secrecy (access control) and integrity information with data as it is retrieved and copied over the network. Distributed access control is not just enforced at one point in the network where the data is accessed, but may be enforced wherever a copy of the data is accessed. Distributed information flow types enforce access rights and enforce a form of causality that supports tracking of tracking of accountability for principals. In recognition of the fact that communication may be over an insecure network, sensitive data is assumed to be "virtually" signed or encrypted, for integrity and secrecy guarantees, respectively. The computational burden of the ubiquitous application of cryptographic operations is avoided through the use of cryptographic types for static checking where possible.