UCL’s Sapienz wins ISSTA 2026 award for powering Meta and WhatsApp app reliability at scale
The automated testing engine, originally developed at University College London, has been recognised for a decade of improving app reliability at scale.
Key points
- Ke Mao, Mark Harman, and Yue Jia received the ISSTA 2026 Most Influential Paper Award for their 2016 research on Sapienz.
- The tool runs millions of tests daily at Meta, surfacing thousands of unique crashes each week to aid engineering teams.
- Sapienz now powers dynamic taint analysis at WhatsApp and supports the development of autonomous testing agents.
Meta’s internal engineering team has announced that Sapienz, its multi-objective automated testing framework for Android applications, has received the ACM SIGSOFT International Symposium on Software Testing and Analysis (ISSTA) 2026 Most Influential Paper Award. This recognition highlights the enduring operational value of academic research when integrated into large-scale software development pipelines, particularly for organisations managing complex mobile ecosystems.
Root cause
The award recognises the 2016 paper “Sapienz: multi-objective automated testing for Android applications” by Ke Mao, Mark Harman, and Yue Jia. The original research aimed to solve three specific testing challenges: exploring more of an application’s state space, identifying a higher volume of crashes, and reducing the number of steps required to reproduce those failures. This approach makes debugging significantly more efficient for engineering teams.
Attack path
Sapienz does not exploit vulnerabilities but rather simulates user interactions to uncover stability issues. In its initial evaluation across 1,000 popular Google Play apps, the system identified 558 previously unknown crashes. By automating the exploration of app states, it reveals faults that manual testing might miss, allowing developers to address reliability gaps before release.
Affected versions
The tool is not a product with versioned releases for end-users but an internal infrastructure component at Meta. It was deployed at scale within 17 months of the original paper’s publication. Today, it operates across Meta’s platforms, supporting hundreds of app releases monthly and functioning as the exploration engine for WhatsApp’s dynamic taint analysis.
Mitigation
- Integrate automated multi-objective testing into continuous integration pipelines to catch crashes early in the development lifecycle.
- Use synthetic data and test accounts to safely explore application state spaces without exposing real user information.
- Prioritise reducing reproduction steps for identified bugs to accelerate engineering response times and improve overall product quality.
What it means for organisations
Organisations should assess whether their current testing strategies adequately cover complex application states and edge cases. Engineering leaders need to evaluate if automated tools are effectively reducing the time between bug discovery and resolution. Teams should consider how academic research or external tools can be adapted to improve internal reliability metrics. It is crucial to ensure that testing infrastructure supports both functional correctness and security analysis, such as taint tracking.
What to do and how to stay safe: Meta
- Review your current automated testing coverage to ensure it includes multi-objective exploration strategies rather than simple regression checks.
- Implement tools that can automatically generate minimal reproduction steps for discovered crashes to speed up developer debugging.
- Evaluate the integration of dynamic analysis techniques, such as taint analysis, within your automated testing workflows.
- Monitor industry awards and research publications for emerging testing methodologies that could enhance your organisation’s software quality processes.
General security guidance from the TechSec Observer newsroom. It is not confirmed advice from the organisations named in this story.
Frequently asked questions
Who developed the Sapienz testing tool?
Sapienz was developed by Ke Mao, Mark Harman, and Yue Jia, with research originating at University College London before being adopted by Meta.
What is the primary function of Sapienz?
Sapienz automatically tests Android applications to find crashes and helps engineers reproduce them in fewer steps, making failures easier to investigate.
Has Sapienz been used outside of crash detection?
Yes, it serves as the app-exploration engine for dynamic taint analysis at WhatsApp and is being extended to support large language model agents for functional correctness.
