Diagonal Artifacts in Samsung Images: PRNU Challenges and Solutions

TL;DR

This work tackles the erosion of PRNU-based camera attribution caused by diagonal artifact patterns in Samsung smartphones. It identifies model-dependent diagonal correlations that produce fingerprint collisions, and shows that PRO-mode raw captures restore PRNU fidelity for S-series devices, though this is not available for all models or forensic cases. The authors also analyze HDR and portrait modes, finding HDR alters artifact localization and portrait bokeh does not inherently trigger false positives, suggesting targeted strategies to mitigate misdetections. The study highlights practical implications for forensic workflows when raw data are accessible and points to future research on artifact suppression and extended analyses across device variants.

Abstract

We investigate diagonal artifacts present in images captured by several Samsung smartphones and their impact on PRNU-based camera source verification. We first show that certain Galaxy S series models share a common pattern causing fingerprint collisions, with a similar issue also found in some Galaxy A models. Next, we demonstrate that reliable PRNU verification remains feasible for devices supporting PRO mode with raw capture, since raw images bypass the processing pipeline that introduces artifacts. This option, however, is not available for the mid-range A series models or in forensic cases without access to raw images. Finally, we outline potential forensic applications of the diagonal artifacts, such as reducing misdetections in HDR images and localizing regions affected by synthetic bokeh in portrait-mode images.

Figures (5)

• Figure 1: Scatter plots of the PCE values obtained for the Galaxy S series devices (a) and A5* series (b).
• Figure 2: Autocorrelation of the estimated PRNUs from different devices: SD06 (a), AD01 (b), and AD07 (c). For clarity, visualization is limited to the central block of size $551\times551$.
• Figure 3: ROC curves comparing performance with raw and default fingerprints (a). TPR and FPR values at $\tau=60$ (b).
• Figure 4: Examples of cross-correlation between the default fingerprint (with diagonal artifacts) and image residuals. Upper panel: non-HDR (a) and HDR (b) images from ALBISANI_2021. Lower panel: non-HDR top-left block showing a peak at $[0,0]$ (c); HDR top-left block showing shifted peaks with maximum at $[5,3]$ (d); HDR bottom-left block with maximum at $[5,-3]$ (e), revealing local translations from HDR construction.
• Figure 5: Example of block-wise cross-correlation ($21\times21$ blocks) between the default fingerprint (with diagonal artifacts) and residuals from Samsung portrait images. Portrait image from ALBISANI_2021 with bokeh effect (a). Corresponding correlation map showing no correlation in bokeh regions (b). Original pre-bokeh image (c), exhibiting uniform correlation across the frame (d).