Improving information flow control design with security contexts

Abstract

This thesis outlines a new language feature for Ruby: the security context, which enables complex information flow control schemes to be written in Ruby without modification to the virtual machine. Security contexts are Ruby objects that act as transparent proxies and can be attached to other objects, allowing them to seamlessly modify parameters and return values into and out of those objects' methods. Security contexts are demonstrated to be simple and effective in creating two flow control applications that would otherwise pose significant challenge to build: taint tracking as a Ruby library, and data flow assertions for Ruby on Rails applications. The performance of these systems was benchmarked while running as a part of a Rails application, and reached acceptable performance: taint tracking had no impact on performance, and data flow assertions saw a 50% throughput decrease, while providing considerable protection against privacy leaks and security vulnerabilities.