Academics from the University of California (Berkeley), University of Washington, University of California (San Diego) and Carnegie Mellon University disclosed a new pixel‑stealing side‑channel attack on Android devices they have codenamed Pixnapping, according to a paper published by the group.
The researchers said the technique can be used to covertly capture sensitive content displayed on a device, including two‑factor authentication codes and Google Maps timelines, by recovering pixels from a target app one by one. The report describes the vulnerability as a form of side‑channel attack, the researchers said.
According to the paper, Pixnapping exploits Android APIs to push another app’s displayed pixels into the rendering pipeline and then measures color‑dependent side effects of graphical operations to recover those pixels. The authors likened the approach to earlier browser timing attacks (see the archived Paul Stone‑style attacks) and said a malicious app can drive victim pixels into rendering via Android intents and stacks of semi‑transparent activities, then apply graphical operations such as blur to leak rendering data.
The study tested five Google and Samsung devices running Android 13 to 16, and the authors said the underlying methodology is present across the platform. Google is tracking the issue as CVE‑2025‑48561 (CVSS 5.5) and issued patches in its September 2025 Android Security Bulletin, the company said in its advisory and related source commits https://source.android.com/docs/security/bulletin/2025-09-01 and https://android.googlesource.com/platform/frameworks/native/+/41eafc6fe601f972dc86f6edf10ad7efbd3f59bd, though the researchers said a workaround exists that can re‑enable the method and Google is working on an additional fix.
The paper also reported that the behavior can be abused to determine whether arbitrary apps are installed on a device, bypassing Android’s post‑Android‑11 restrictions on querying installed applications. Google has marked that app‑list bypass as “won’t fix,” the article said, and the researchers urged measures such as allowing sensitive apps to opt out and limiting attacker measurement capabilities.