Security Engineering for Large Scale Distributed Applications

The way security mechanisms for large-scale distributed applications are engineered today has a number of serious drawbacks. As a result, secure distributed applications are a) very expensive and error-prone to build, deploy, and integrate, b) complex and error-prone to operate and administer, and still c) far from being adequate to the real-life problems. Drawing on my academic and industrial experiences, I will discuss several recently invented techniques that can improve engineering of security mechanisms for distributed systems. I will specifically talk about improving those mechanisms that are based on the decision-enforcement paradigm, and will use access control as a representative example. I will examine in detail one particular method, Attribute Function, which enables the use of application-specific data in authorization decisions while keeping distributed applications security unaware. The talkl was given at the following organizations: * Departement Computerwetenschappen, Katholieke Universiteit Leuven, on June 19, 2003. * Department of Electrical and Computer Engineering, University of British Columbia, on March 7, 2003. * The Department of Computing and Software, McMaster University, on February 25, 2003. * Faculty of Computer Science, Dalhousie University, on January 28, 2003