Automated Driver Generation for Analysis of Web Applications

Abstract

With web applications in high demand, one cannot underestimate the importance of their quality assurance process. Web applications are open event-driven systems that take sequences of user events and produce changes in the user interface or the underlying application. Web applications are difficult to test because the set of possible sequences of user inputs allowed by the interface of a web application can be very large. Software model checking techniques can be effective for validating such applications but they only work for closed systems. In this paper, we present an approach for closing web applications with a driver that contains two parts: (1) the application-specific Page Transition Graph (PTG), which encodes the application's possible pages, user and server events, their corresponding event-handlers, and user data and (2) the application-independent PTG-based driver, which generates test sequences that can be executed with analysis tools such as Java PathFinder (JPF). The first part can be automatically extracted from the implementation of a web application and the second part is written once and reused across multiple web applications belonging to the same framework. We implemented our approach in a driver generator that automatically extracts PTG models from implementation of JSP-based web applications, checks the extracted PTGs for navigation inconsistencies, and enables JPF analysis. We evaluated our approach on ten open-source and industrial web applications and present the detected errors.