Pixnapping: a new threat for Android users

Attack can dismantle 2FA in seconds, exposes limits of platform assurances that one app cannot access another’s data

A group of researchers have revealed a new Android attack that they call Pixnapping. Attackers employing this can covertly read sensitive information displayed on a device’s screen, including one-time passcodes. The attack requires a user to simply install and open a malicious, no-permissions application.

"Pixel-stealing attacks enable malicious websites to leak sensitive content displayed in victim websites," American univesity professors write in their research paper, Pixnapping: Bringing Pixel Stealing out of the Stone Age.

Anything visible when the target app is opened can be stolen. Two-factor authentication (2FA) codes, chat messages and email messages are all vulnerable.

While existing browser mitigations block pixel stealing on most websites today, their framework "bypasses all these browser-based mitigations," and can leak pixel values from non-browser apps, which they say has not been done before, say the researchers Alan Wang, Christopher W Fletcher, and Yingchen Wang of University of California, Berkeley (UCB), Hovav Shacham of University of California, San Diego, Pranav Gopalkrishnan and David Kohlbrenner of University of Washington, and Riccardo Paccagnella of Carnegie Mellon University. 

One pixel at a time

Once the user opens the malicious app, the attack begins by forcing a target app, such as Google Authenticator, to launch its exported activities. In this instance, it waits until the 2FA code refreshes, then starts retrieving that code one pixel at a time.

Read: Pakistan's VPN conundrum may impact digital economy 

The attack runs graphical operations on individual pixels of interest to the attacker. Once it isolates a single pixel, it waits until the pixel begins rendering, and by measuring how long each frame takes to render, the malicious app can tell whether specific pixels are, for example, white or non-white, allowing characters and symbols to be reconstructed.

The researchers found that, “It is as if the malicious app was taking a screenshot of screen contents it should not have access to”.

Not a first

The approach is similar to the GPU.zip attacks in 2023, which abused side channel attacks: security exploits that use data inadvertently leaked by a system to gain access to sensitive information. These vulnerabilities were patched by websites blocking the attacks in browsers.

In tests on Google devices from Pixels 6 to 9, the researchers recovered full six-digit codes from Google Authenticator in under 30 seconds, with success rates varying by model (code recovery was faster on older models).

Read More: Amnesty calls out Pakistan over mass telecom surveilance 

On Samsung’s Galaxy S25, the team did not meet the 30-second window because of noise in the measurements, but they found that on average, an attack took between 14-26 seconds.

While small regions such as passcode digits can be reconstructed quickly, the leak rate is modest, "around 0.6 to 2.1" pixels per second, meaning that recovering an entire screen, such as a Gmail inbox, can take 10–25 hours.

The attack is designed to remain hidden by using overlays with less than 1% transparency, the researchers observed.

Protecting your pixels

Answering frequently asked questions on their website, the researchers advised users to "install the latest Android patches as soon as they become available" to ensure their pixels are protected.

For developers, they recommended either hiding sensitive visual content from appearing on screen or restricting transparent layering over sensitive activities by checking to see which other apps are allowed to access certain information.

Google attempted to address this problem by releasing a partial mitigation in September and has planned an additional security patch in December, according to the researchers. A media outlet, specialising in tech, has reported that the company has seen no evidence of exploitation in the wild.

Security experts say the attack underlines the limits of platform assurances that one app cannot access another’s data. For now, users will need to be more careful with the apps they download.

AI was used to assist in research and structure.