Challenges in Making Access Control Sensitive to the "Right" Contexts

Abstract

Access control is a fundamental security mechanism that both protects processes from attacks and confines compromised processes that may try to propagate an attack. Nonetheless, we still see an ever increasing number of software vulnerabilities. Researchers have long proposed that improvements in access control could prevent many vulnerabilities, many of which capture contextual information to more accurately detect obviously unsafe operations. However, developers are often hesitant to extend their access control mechanisms to use more sensitive access control policies. My experience leads me to propose that it is imperative that an access control systems be able to extract context accurately and efficiently and be capable of inferring any non-trivial policies. In this talk, I will discuss some recent research that enforces context-sensitive policies by either extracting process context, integrating code to extract context from programs, or extracting user context. We find that context-sensitive mechanisms can prevent some obviously unsafe operations from being authorized efficiently and discuss our experiences in inferring access control policies. Based on this research, we are encouraged that future research may enable context-sensitive access control policies to be produced and enforced to prevent vulnerabilities.