Toolkit tests PRNU camera attribution against fingerprint-copy forgeries
Original title: Evaluating the Robustness of PRNU Source-Camera Attribution to the Fingerprint-Copy Attack: An Open Toolkit with Triangle-Test Detection
Triangle-test detection caught every forgery at a cohort-calibrated 5% false-positive point in the benchmark.
Researchers released prnucf, an MIT-licensed toolkit that implements the additive fingerprint-copy attack on Photo-Response Non-Uniformity camera attribution and pairs it with a triangle-test detector and cohort-calibrated decision procedure. The 2011 attack estimates a target camera fingerprint from public images and injects a scaled copy into an unrelated photo so it can clear a Peak-to-Correlation-Energy threshold. On a controlled multi-camera Monte-Carlo benchmark, the attack reached an arbitrary target PCE at a mean PSNR of 59.2 dB, described as visually lossless. The triangle test separated forgeries from innocent images with an area under the ROC curve of 0.999.
Unconfirmed: Authors note limits of synthetic validation and that GAN-based fingerprint-transfer detection is unsolved.
- The toolkit is released under the MIT license.
- Mean PSNR for target-PCE forgeries was 59.2 dB.
- Triangle-test AUC was 0.999 on the benchmark.
- Authors say a fixed significance cut-off is unsafe in casework.
- GAN-based fingerprint-transfer detection remains an open problem.
Impact: Forged PRNU can falsely attribute a photo to a camera
From the original article
1 min read at arxiv.orgPhoto-Response Non-Uniformity (PRNU) is the de facto standard trace for sensor-level source-camera attribution, and a high Peak-to-Correlation-Energy (PCE) between a questioned image and a camera reference fingerprint is routinely treated as strong evidence of common origin.
By Zeno J. M. H. Geradts · arxiv.org
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.