Glonce

C2PA camera authentication broken on Android, researcher shows

securityThe Glonce editorsHacker News🔥78Glonce rating78Impact82Novelty68Relevance98Surprise65Players72Credibility88

Security researcher David Buchanan, known as retr0id, published findings on 25th August 2026 showing that C2PA, the standard meant to cryptographically prove an image or video came from a real camera, is broken on Android in a way he says cannot be realistically patched. C2PA camera apps rely on Android's Key Attestation and Google Play Integrity to stop a device owner from tampering with the app so it signs arbitrary files instead of genuine sensor data. Buchanan demonstrates that gaining root on the device breaks that entire trust model: once an attacker has root, the C2PA signing keys are still locked inside hardware security (StrongBox, backed by the Titan M2 chip on newer Pixels) so they cannot be extracted, but the attacker does not need the raw keys. They can simply ask StrongBox to sign whatever data they like while impersonating the camera app, producing a forged image or video that verifies as an authentic, unedited capture.

Two routes give an attacker that root access. The first is a low-cost hardware fault-injection (glitching) attack, which Buchanan says cannot be patched at all on affected hardware. The second is a software privilege-escalation exploit; he states that a one-click root exploit currently exists in the wild for fully-patched Google Pixel devices, tracked as CVE-2026-43499, meaning no hardware attack is even required. He built a client/server tool called keystork to automate signing arbitrary files through the KeyStore API once root is obtained, and used it to produce a demo AI-generated image that Google's Pixel Camera app certifies as a real unedited photo, and a YouTube video whose infobox claimed it was camera-captured when it was not.

Buchanan targeted the Pixel Camera app specifically because Google states it achieved Assurance Level 2, the highest rating in the C2PA Conformance Program, a level currently only achievable on Android. He argues that every other Android "C2PA Camera" app he checked relies on the same two mechanisms, Key Attestation or Play Integrity, and is broken the same way, and that unlike Pixel-specific hardware attacks, the software exploit works on the cheapest, most vulnerable device in the Android ecosystem. He separately notes that Samsung devices with RKP (Real-time Kernel Protection) enabled resist his current hardware exploit, though he has unimplemented plans to work around it.

He reported the issue to Google at least 90 days before publication. Google awarded him a $7500 bug bounty but closed the report with the status "Won't fix (infeasible)", telling him that hardware glitching and side-channel attacks are formally out of scope for its bounty program even though the team found the findings valuable. Buchanan concludes that since the most obvious attack vector is out of scope for Google's Vulnerability Reward Program, the program does not meaningfully protect Android C2PA implementations. He also notes, separately, that Meta had already patched the same CVE-2026-43499 privilege-escalation bug on its Quest VR headsets near the start of the month, specifically to stop players from cheating in VR games, which he contrasts with Google's flagship Pixel devices still lacking a patch. Buchanan credits Dr. Neal Krawetz of Hacker Factor, who has criticized C2PA's effectiveness for years, with introducing him to the topic and helping coordinate the disclosure, and thanks the Provenance and Authenticity Standards Assessment Working Group (PASAWG) for its related research. He adds that Apple is rumored to be building its own media provenance system, not yet released.

Key facts

  • Rooting an Android device, via either a low-cost hardware fault-injection attack or a software exploit, defeats the Key Attestation and Play Integrity checks that C2PA camera apps depend on, letting an attacker sign arbitrary files, including AI-generated images and video, as genuine camera captures.
  • A one-click root exploit, CVE-2026-43499, currently exists in the wild for fully-patched Google Pixel devices, so no hardware attack is even required to forge C2PA-signed media today.
  • The Pixel Camera app holds Assurance Level 2, the highest rating in the C2PA Conformance Program and the only level currently achievable on Android, yet the researcher used it to certify a fake AI image as an authentic unedited photo.
  • Google awarded a $7500 bug bounty but closed the vulnerability report as "Won't fix (infeasible)" and classifies hardware glitching and side-channel attacks as out of scope for its bounty program.
  • Meta had already patched the same CVE-2026-43499 flaw on its Quest headsets near the start of the month to stop VR game cheating, while Google's own Pixel devices remained unpatched at time of publication.

Why it matters

C2PA is the industry's flagship answer to AI-generated fakes: a cryptographic chain meant to let anyone verify that a photo or video came straight from a real camera sensor, untouched. This research shows that chain relies entirely on Android's root-detection mechanisms holding, and those mechanisms do not hold, not against a patched flaw that already has a public one-click exploit, and not against a class of hardware attacks that the researcher says cannot be patched at all. A verification badge that can be forged is arguably worse than no badge, because it lends false confidence.

Who it affects

Anyone relying on C2PA provenance labels to judge whether Android-captured images or video are authentic, including platforms and viewers that trust the Pixel Camera app's Assurance Level 2 status. It also affects other Android "C2PA Camera" apps that lean on the same Key Attestation or Play Integrity mechanisms, which the researcher says are vulnerable in the same way, and it puts Google in the position of having formally closed a report on its own flagship security claim.

How to use it

The researcher has published a client/server tool, keystork, and a proof-of-concept script for signing arbitrary images through the Pixel Camera app's keys once a device is rooted, along with a pointer to the Root My Pixel tool for reproducing the root itself; he says he personally tested this on Pixel 8a and 9a devices. He also lists a public C2PA conformance registry where any implementation using Android_KeyAttestation or Google_PlayIntegrity for its attestation is likely vulnerable the same way.

How solid is it

The claims come from the researcher's own working exploits, demonstrated against Google's own top-rated implementation with a real forged image and video, and reported to Google at least 90 days before publication under coordinated disclosure. Google's own response, a $7500 bounty alongside a formal 'Won't fix (infeasible)' closure, corroborates that the underlying flaw is real and that Google does not currently see a practical fix.

Risks and caveats

The article does not name how many other Android C2PA camera apps exist beyond Pixel Camera, nor does it give a full list or count of device models proven vulnerable to the hardware attack; the complete technical details of the hardware fault-injection method are deliberately withheld for a future article. Samsung devices with RKP enabled currently resist the researcher's specific hardware exploit, though he states he has unimplemented plans to work around that mitigation, and no date is available for when Apple's rumored separate provenance system might ship or how it would fare.

“The most obvious (to me) C2PA attack vector is out of scope for Google's VRP. Thus, the VRP does not meaningfully protect Android C2PA implementations.”

— David Buchanan, security researcher