Author Description

Verifying the security posture as a system evolves is indispensable for building deployable software systems. Traditional security testing lacks flexibility in (1) providing early feedback to the architect on the ability of the software to predict security threats so that changes are made before the system is built, (2) responding to changes in user and behavior requirements that could affect the security of software, and (3) offering real design fixes that do not merely hide the symptoms of the problem (i.e., patching). We motivate the need for an architecture-level testing for security grounded on incremental and continuous refinements to support agile principles. We use architecture as an artifact for initiating the testing process for security through subsequent and iterative refinements. We extend the use of implied scenario to reveal undesirable behavior caused by ambiguities in users’ requirements and we analyze detection their security implications. This approach demonstrates how architecture-centric evaluation and analysis can assist in securing systems developed using an agile development cycle. We apply this approach to a case study to evaluate the security of identity management architectures. We reflect on the effectiveness of this approach in detecting vulnerable behaviors and the cost-effectiveness of refining the architecture before vulnerabilities are built into the system.