Gespeichert in:
Bibliographische Detailangaben
1. Verfasser: Reed, Alex
Format: Recurso digital
Sprache:
Veröffentlicht: Zenodo 2024
Schlagworte:
Online-Zugang:https://doi.org/10.5281/zenodo.18501209
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866902239759564800
author Reed, Alex
author_facet Reed, Alex
contents <p>A genuine raw image is a measurement of photon to electron conversion on silicon, not merely a container file with camera tags. This paper presents a physics based forensic method that tests whether the active image payload is electrically coupled to the Optical Black reference pixels on the same sensor die, detecting “re-wrapped" or "injected" payloads, which is the core threat model. We define the Voltage Dead Zone as a statistically significant gap between the Optical Black baseline distribution and the darkest values in the active image area when evaluated in linear raw space with automatic scaling disabled. We also define histogram indicators that detect algorithmic noise injection, including Gaussian Sterility where the floor noise distribution is an unusually smooth Gaussian with suppressed heavy tails. For exposures captured at standard ISOs, where analog gain is low and dynamic range is high, the joint presence of a large Voltage Dead Zone and Gaussian Sterility provides a strong and testable basis to conclude that an image set is synthetic or that an external payload has been injected into a raw container.</p> <p>Independent forensic results confirming that these 19 "cloud images" originated from no known CMOS sensor further validate the robustness of physics-based sterility in forgery detection.</p> <p><a name="bookmark=id.30j0zll"></a><em>Index Terms</em>—CMOS image sensors, digital forensics, image authentication,, optical black, sensor noise, synthetic media detection.</p> <p>Version 3.11; Feb 2 2026: Addendum iincluded</p>
format Recurso digital
id zenodo_https___doi_org_10_5281_zenodo_18501209
institution Zenodo
language
publishDate 2024
publisher Zenodo
record_format zenodo
spellingShingle Forensic Authentication of RAW Imagery Using Optical Black and CMOS Sensor Noise Analysis
Reed, Alex
CMOS image sensors, digital forensics, image authentication, Canon, optical black, sensor noise, synthetic media detection.
<p>A genuine raw image is a measurement of photon to electron conversion on silicon, not merely a container file with camera tags. This paper presents a physics based forensic method that tests whether the active image payload is electrically coupled to the Optical Black reference pixels on the same sensor die, detecting “re-wrapped" or "injected" payloads, which is the core threat model. We define the Voltage Dead Zone as a statistically significant gap between the Optical Black baseline distribution and the darkest values in the active image area when evaluated in linear raw space with automatic scaling disabled. We also define histogram indicators that detect algorithmic noise injection, including Gaussian Sterility where the floor noise distribution is an unusually smooth Gaussian with suppressed heavy tails. For exposures captured at standard ISOs, where analog gain is low and dynamic range is high, the joint presence of a large Voltage Dead Zone and Gaussian Sterility provides a strong and testable basis to conclude that an image set is synthetic or that an external payload has been injected into a raw container.</p> <p>Independent forensic results confirming that these 19 "cloud images" originated from no known CMOS sensor further validate the robustness of physics-based sterility in forgery detection.</p> <p><a name="bookmark=id.30j0zll"></a><em>Index Terms</em>—CMOS image sensors, digital forensics, image authentication,, optical black, sensor noise, synthetic media detection.</p> <p>Version 3.11; Feb 2 2026: Addendum iincluded</p>
title Forensic Authentication of RAW Imagery Using Optical Black and CMOS Sensor Noise Analysis
topic CMOS image sensors, digital forensics, image authentication, Canon, optical black, sensor noise, synthetic media detection.
url https://doi.org/10.5281/zenodo.18501209