What image forensics means here
Not the Hollywood version — a disciplined technical reading of the file itself. PictureMatters examines, layer by layer:
- Metadata analysis — EXIF, XMP, IPTC and GPS records: what the file claims about itself, and what has gone missing.
- File structure — the container walked segment by segment: JPEG APP segments, PNG chunks, WebP boxes, comments, color profiles.
- Compression characteristics — JPEG quality estimation, progressive vs baseline encoding, and how the file was likely produced.
- Entropy — a slice-by-slice unpredictability profile that makes appended payloads and unusual regions visible.
- Hashes — a SHA-256 identity and a perceptual fingerprint for near-duplicate comparison.
- Processing traces — software strings left behind by editors, exporters and platforms that touched the file.
- Provenance indicators — AI generator markers, IPTC digital-source declarations, and C2PA / Content Credentials signals.
Signals, not verdicts
The vocabulary matters. PictureMatters reports potential clues: a progressive JPEG hints at a re-save; an appended ZIP hints at hidden content; a generator string hints at the tool that made the file. Each hint has real evidential weight — and real limits. Metadata can be edited or stripped. Missing metadata is normal after sharing and proves nothing. Entropy patterns are statistical, not declarative.
That is why every report separates recorded facts (what is in the bytes) from interpretation (what it may mean), and ends with a limitations note. The methodology page documents each layer and its boundary of knowledge, and the edited-image guide shows how to weigh these signals in practice.
Who uses it
- Journalists and researchers doing first-pass triage on images before deeper verification.
- Photographers checking what survives publishing and what platforms strip.
- Buyers and marketplaces inspecting product photos for signs of re-saving or hidden data.
- Curious people who want to know what their own photos secretly carry.
Frequently asked questions
Can PictureMatters prove an image is fake?
No. It identifies metadata and technical signals that may help evaluate an image — such as a generator marker, appended data or an unusual entropy profile. Signals are evidence to interpret, never proof on their own.
What is an entropy profile?
A measurement of how unpredictable the file's bytes are, slice by slice. Compressed photo data sits in a predictable range; appended archives or odd re-encodes often stand out as dips or spikes. Entropy alone cannot establish the presence of steganography or hidden data — it only flags unusual byte-pattern characteristics worth examining.
What does "trailing data" mean?
Bytes that ride after the image's proper end marker (JPEG EOI or PNG IEND). Sometimes harmless padding; sometimes a hidden ZIP, PDF or other payload. The report flags the length and fingerprints the type.
Which formats can be analyzed?
JPG, PNG and WEBP natively — plus HEIC/HEIF and AVIF, which are converted to JPG inside your browser first. Files up to 20 MB.