DroidGraph framework maps Android apps, boosting test coverage 18%

As mobile app releases accelerate, automated test generation has become critical—but Android's event-driven architecture makes modeling applications difficult. Traditional static analysis struggles with the platform's dynamic, event-based nature, while random dynamic test generators are slow and lack application awareness. To address this, researchers Jordan Doyle, Thomas Laurent, and Anthony Ventresque introduced DroidGraph, a framework that fuses static analysis with efficient systematic exploratory testing to build a comprehensive control flow model of Android apps. This model spans everything from low-level method statements to high-level UI structures, providing a solid foundation for automated test generation.

In experiments across 19 diverse apps, DroidGraph's exploratory tests outperformed random exploration significantly—interacting with 18% more of the app on average while requiring 345 fewer interactions. The dynamic analysis results also complemented the static analysis, uncovering an average of 51 additional components and 49% more UI callback links in the application code. For mobile development teams, this means more thorough test coverage in less time, leading to fewer bugs and faster release cycles. DroidGraph's hybrid approach represents a practical step forward in making automated testing for Android both more intelligent and more efficient.