Abstract

Existing mechanisms for authorizing and auditing the flow of information in networked computer systems are insufficient to meet the security requirements of high-assurance software systems. Current best practices typically rely on operating-system provided file permissions for authorization and an ad-hoc combination of OS and network-level (e.g. firewall-level) logging to generate audit trails.

This talk will describe work on a security-oriented programming language called Aura that attempts to address this problem of auditable information flows in a more principled way. Aura supports a built-in notion of principal and its type system incorporates ideas from authorization logic and information-flow constraints. These features, together with the Aura run-time system, enforce strong information-flow policies while generating good audit trails. These audit trails record access-control decisions (such as uses of downgrading or declassification) that influence how information flows through the system. Aura’s programming model is intended to smoothly integrate information-flow and access control constraints with the cryptographic enforcement mechanisms necessary in a distributed computing environment.

This is joint work with Jeff Vaughan, Limin Jia, Karl Mazurak, Jianzhou Zhou, Joseph Schorr, and Luke Zarko.