ALEX: Mixed-Mode Learning of Web Applications at Ease

摘要

In this paper, we present ALEX, a web application that enables non-programmers to fully automatically infer models of web applications via active automata learning. It guides the user in setting up dedicated learning scenarios, and invites her to experiment with the available options in order to infer models at adequate levels of abstraction. In the course of this process, characteristics that go beyond a mere "site map" can be revealed, such as hidden states that are often either specifically designed or indicate errors in the application logic. Characteristic for ALEX is its support for mixed-mode learning: REST and web services can be executed simultaneously in one learning experiment, which is ideal when trying to compare back-end and front-end functionality of a web application. ALEX has been evaluated in a comparative study with 140 undergraduate students, which impressively highlighted its potential to make formal methods like active automata learning more accessible to a non-expert crowd.