Menu
Independent Tor encyclopedia Guides
Guides

Photo and PDF Metadata (MAT2)

What EXIF, GPS, and author fields leak in photos and PDFs, how MAT2 writes a cleaned copy, and why that is not the same as opening an attachment.

Author
Published
Last reviewed
Level
Beginner
Reading time
5 min
What a file leaks versus what MAT2 strips Three steps. A saved photo or PDF can carry GPS, camera, and author fields. MAT2 writes a cleaned sibling and leaves the original. Send only the cleaned copy — scrubbing does not make a hostile file safe to open while online. 01 As saved GPS, camera, author Tags ride with the bytes 02 Run MAT2 Writes filename.cleaned.* Original stays on disk 03 Send cleaned Share only thecleaned copy Hostile files stay hostile 01 As saved GPS, camera, author Tags ride with the bytes 02 Run MAT2 Writes filename.cleaned.* Original stays on disk 03 Send cleaned Share only the cleaned copy Hostile files stay hostile
MAT2 writes a sibling file (photo.cleaned.jpg). Keep the original offline if you need it. Metadata scrubbing is not a license to open someone else’s DOC or PDF in a host app while online.

Short answer: Photos and office files carry data about the file — GPS, camera model, author, timestamps — that encryption and Tor do not remove. MAT2 (Metadata Cleaner in Tails) writes a cleaned copy. It does not anonymize what is in the picture, and it does not make a downloaded PDF safe to open in Word while you are online.

Two different jobs sit next to each other. Strip metadata before you send a file you created. Do not open a file someone else sent you in a host app on a live network: don’t open downloaded files while online.

What a file still says

Tails and EFF both list the usual tags: where a photo was taken, when, which camera, whose name is in a PDF or Word document. Whonix adds that JPEG EXIF inside a PDF can survive if the image was embedded whole. The metadata glossary is the short definition.

MAT2’s own notes: there is no reliable way to detect every metadata system in a complex format. If --show prints nothing, the file is not therefore clean. Do not wait for a red flag. Clean it if you will share it.

Procedure (files you authored)

  1. On Tails — Apps → Accessories → Metadata Cleaner. Select the file. Save the cleaned output. Send that copy, not the original.
  2. On a Linux workstation with MAT2 — mat2 photo.jpg writes photo.cleaned.jpg beside the original. Full mode can rasterize a PDF (text no longer selectable). Lightweight mode (mat2 -L …) removes some tags and keeps more of the original layout.
  3. Before embedding images in a .docx — clean each image first. Tags inside embedded JPEGs survive otherwise.
  4. Send only the cleaned sibling — via OnionShare or a newsroom SecureDrop, not as an afterthought on a mailbox.

Whonix ships MAT2 on the workstation. Privacy Guides lists MAT2 and the Linux GUI Metadata Cleaner.

MAT2 does not defeat watermarks, printer tracking dots, or steganography. It does not clean the content.

Attachments you do not open

A cleaned file you authored is one problem. A DOC or PDF that arrived over Tor is another. Opening it in Acrobat or Word can fetch images and fonts off Tor and show your real IP. That is the Tor Browser warning — not MAT2’s job.

Dangerzone (Freedom of the Press Foundation) converts an untrusted document to pixels in a sandbox, then rebuilds a PDF. Tails points at Dangerzone for other people’s files, and at Metadata Cleaner for yours. Use both ideas; do not mix them up.

Sources

See also: don’t open downloaded files while online, OnionShare.