Detection of Misconfigurations in Access Control Systems

Time: April 23, 2009 - 10:00 AM - 11:00 AM

Location: INI Lower Level Conference Room


Student project presentation presented by Sudeep Modi

Project Advisor: Dr. Lujo Bauer
Reader: Dr. Anupam Data

Drafting access control policies in a system involving a lot of shared resources and users accessing different subsets of those resources is difficult. The complexity of the system further increases as more users and resources are added to the system and some others are removed. Erroneous or overlooked policies often cause legitimate accesses to be denied that can not only be annoying to users but can also have severe consequences if timely access is critical. Therefore, there is a need to detect misconfigurations in access control policies before the accesses occur in an automated manner.

The objective of this thesis is to describe one technique to detect such misconfigurations that is generic enough to be used for different access control systems. We make use of Bayesian inference to predict new accesses that are currently not covered by the policy, by examining the history of all accesses made in the system. We examine various parameters that can be tuned to improve the performance of the predictor and adapt it to the system. Using data collected from our testbed, we show that this technique can correct a significant fraction of misconfigurations before they impact a legitimate access, and that we can detect most of the discrepancies between the policy that an administrator implemented and the policy that was intended.

We thus show that the technique of Bayesian inference presented in this thesis is a suitable one to detect misconfigurations in real world access control systems.