GitHub’s AI agent found 24 Android app vulnerabilities
GitHub Security Lab's AI taskflows found 24 Android app vulnerabilities, including location tracking in OsmAnd and session-cookie theft in Wikipedia's app.
GitHub Security Lab researcher Kevin Stubbings built AI-driven audit workflows (taskflows) on the open-source Taskflow Agent and used them to report more than 20 Android app vulnerabilities. In OsmAnd, with over 10 million downloads, an exported MapActivity accepted restricted intent extras, letting any app silently swap the map tile source to an attacker server and log victims' coordinates. The Wikipedia Android app's deeplink handler used endsWith() for hostname checks, allowing attacker JavaScript in a trusted WebView and theft of session cookies valid across all Wikimedia projects. The AI over-flagged low-severity issues and misjudged real-world impact, so every finding still requires an expert human reviewer.
- AI taskflows found 24 Android vulnerabilities including OsmAnd (10M+ downloads) and Wikipedia apps.
- OsmAnd intent-extras flaw enabled silent settings import and route reconstruction via malicious tile server.
- Wikipedia endsWith() hostname check allowed lookalike-domain JavaScript and cross-project session-cookie theft.
- Open-source taskflows require a Copilot license and consume many premium model requests.
- Model over-flagged low-severity bugs; human mobile-security review remains mandatory.
- Organizations
- GitHubOsmAndWikipediaKevin Stubbings
From the original article
2 min read at helpnetsecurity.com
GitHub Security Lab researcher Kevin Stubbings built custom AI-driven audit workflows, called taskflows, on top of the lab’s open source Taskflow Agent, and used them to find and report more than 20 vulnerabilities in Android apps. Two of the disclosed bugs show what’s at stake.
By Anamarija Pogorelec · helpnetsecurity.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.