PADRAIG boosts Android test coverage by 16% in 84% less time
This model-based framework beats state-of-the-art test generators at line coverage and speed.
Android app testing has long relied on random input generation, which is simple but inefficient. Systematic and search-based approaches offer better coverage but demand disproportionate runtime. Model-based tools, which build a representation of the app under test (AUT), promise speed but carry upfront modeling overhead. Now, researchers Jordan Doyle, Thomas Laurent, and Anthony Ventresque from the University of Limerick introduce PADRAIG (Precise AnDRoid Automated Input Generation), a novel framework that constructs a detailed control-flow model of the AUT to generate tests more intelligently.
In a study of 19 randomly selected apps from the F-Droid store, PADRAIG outperformed three state-of-the-art tools—each using a different technique (random, systematic, and search-based). It delivered an average of 16% more line coverage while cutting test generation runtime by 84%. The framework's precision comes from using the control-flow model to focus inputs on yet-unexplored code paths, avoiding wasteful random taps. This dual advantage of higher coverage and lower runtime makes PADRAIG particularly attractive for continuous integration pipelines, where test generation time directly impacts developer productivity. The work was published at the 24th International Conference on Software Quality, Reliability and Security (QRS) and is available on arXiv.
- PADRAIG is a model-based Android test generator using a detailed control-flow model of the app.
- Achieves 16% higher average line coverage than state-of-the-art tools across 19 F-Droid apps.
- Generates tests with 84% less runtime, making it suited for continuous integration environments.
Why It Matters
Faster, higher-coverage Android tests mean more reliable apps with fewer bugs and shorter CI cycles.