Aspects with Program Analysis for Security Policies

Enforcing security policies to IT systems, especially for a mobile distributed system, is challenging. As society becomes more IT-savvy, our expectations about security and privacy evolve. This is usually followed by changes in regulation in the form of standards and legislation. In many cases, small modification of the security requirement might lead to substantial changes in a number of modules within a large mobile distributed system.

Indeed, security is a crosscutting concern which can spread to many business modules within a system, and is difficult to be integrated in a modular way. This dissertation explores the principles of adding challenging security policies to existing systems with great flexibility and modularity. The policies concerned cover both classical access control and explicit information flow policies.

We built our solution by combining aspect-oriented programming techniques with static program analysis techniques. The former technique can separate security concerns out of the main logic, and thus improves system modularity. The latter can analyze the system behavior, and thus helps detect software bugs or potential malicious code.

We present AspectKE, an aspect-oriented extension of the process calculus KLAIM that excels at modeling mobile, distributed systems. A novel feature of our approach is that advices are able to analyze the future use of data, which is achieved by using program analysis techniques. We also present AspectK to propose other possible aspect-oriented extensions based on KLAIM, followed by a discussion of open joinpoints that commonly exist in coordination languages such as KLAIM.

Based on the idea of AspectKE, we design and implement a proof-of-concept programming language AspectKE*, which enables programmers to easily specify analysis-based security policies with the help of high-level program analysis predicates and functions. The prototype is efficiently realized by a two-stage implementation strategy and a static-dynamic dual value evaluation mechanism.

We have performed two case studies to evaluate our programming model and language design. One application is based on a electronic health care workflow system. The other is a distributed chat system. We considered a number of security policies for both primary and secondary use of data, classical access control and predictive access control - control access based on the future behavior of a program.

Some of the above mentioned policies can only be enforced by analysis of process continuations.