The Arrest Was Real. But Was the Photograph?
Two AFP-released photographs, the same camera-class label, and very different forensic likelihood ratios. What SPOT found in the Kazem Hamad case.
Topic
SPOT
18 items from the laboratory.
AI-Generated CSAM: When Everything Is a Priority, Nothing Is
Detecting synthetic imagery is only part of the problem. The greater challenge is ensuring investigative resources reach the children who need protecting.
OpenAI just said there may be “no scientific way” to detect AI-generated images. That’s bullshit. Here’s why.
The absence of a universal, infallible detector does not mean image origin is beyond scientific examination. Forensic science works by measuring evidence and defining the limits of an inference.
AI-Generated Images as Evidence: Authentication, Provenance and Forensic Examination
Forensic examination starts with the disputed proposition, the supplied file and the history of any transformations.
Are AI Image Detectors Reliable? Why Accuracy Alone Is Not Enough
Detector suitability depends on error rates, threshold, calibration, prevalence and how closely validation resembles the material being examined.
C2PA and Content Credentials: What Happens When Provenance Metadata Is Missing?
A missing Content Credential is a missing provenance record, not a classification of an image as synthetic.
Can AI-Generated Images Be Reliably Detected? A Forensic Perspective
Image-origin assessment depends on the evidence available, the competing hypotheses and the conditions under which a method was tested.
Deepfake Detection vs Synthetic-Media Forensics: What Can Actually Be Established?
Deepfake is a broad label. A forensic conclusion must identify the medium, alleged alteration and origin proposition actually examined.
How Can You Tell Whether an Image Was AI-Generated?
A practical reading of visual clues, metadata, Content Credentials and forensic measurements when the origin of an image is questioned.
Synthetic-Media Forensics for Investigators: A Practical Framework
An evidential workflow for preserving questioned media, defining origin propositions and reporting uncertainty.
AIPIO 2026: forensic science in support of intelligence
AIPIO Intelligence 2026 sharpened a question for RHEM Labs: can forensic information about image origin reach analysts early enough to inform a decision?
ADSTAR 2026: from forensic research to useful capability
ADSTAR changed how RHEM Labs understands the path from a working forensic method to useful capability: mission need, problem ownership, experimentation and operational learning.
SPOT: from research to representative evaluation
The current stage of Sensor Pattern Origin Testing and the role of operational users, technical partners and independent evaluators.
Challenges in authenticating synthetic media: forensic insights from 6,000 DALL·E-generated images
Experimental research evaluating image-origin characteristics in camera and DALL·E-generated images.
Determining Image Sensor Temperature Using Dark Current
Research on dark current and image-sensor temperature, forming part of the laboratory’s sensor-pattern research foundation.
Reverse Engineering the Raspberry Pi Camera V2: A study of Pixel Non-Uniformity using a Scanning Electron Microscope
Research into pixel non-uniformity in the Raspberry Pi Camera V2, using scanning electron microscopy.
An analysis of optical contributions to a photo-sensor’s ballistic fingerprints
Research examining optical contributions to image-sensor fingerprints.
Towards a unified theory of sensor pattern noise
Doctoral research on dark current, lens effects and temperature bias in CMOS image sensors.