Files carry hidden passengers. A photo can include the exact GPS coordinates where it was taken; a Word document can include the author's name, employer, and tracked deletions; a PDF can name the machine that produced it. People who would never post their address have posted it inside a JPEG.
This is not theoretical. John McAfee's location in Guatemala was exposed in 2012 by GPS coordinates in a photo published by journalists traveling with him, and an Anonymous-affiliated hacker was identified after posting a photo whose EXIF data pointed to his girlfriend's house, that one has its own Darwin Award entry.
What lives in common formats
- Photos (JPEG/HEIC): EXIF data. GPS coordinates, timestamp, camera or phone model, serial numbers on some cameras, and editing-software traces.
- Office documents (DOCX, ODT, XLSX): author and organization names, total editing time, revision history, comments, and sometimes tracked changes you thought were resolved.
- PDFs: producer application, author fields, creation timestamps; "redactions" drawn as black rectangles over selectable text remain copy-pasteable, a mistake that recurs in court filings yearly.
- Video and audio: device identifiers, timestamps, location on some devices.
- Screenshots: cleaner than photos (no GPS EXIF), but check visible metadata (usernames, tabs, notification previews, Wi-Fi names) which is leaked more often than the invisible kind.
Beyond per-file fields, the content itself fingerprints: printer tracking dots have identified the source of leaked documents, and stylometry (writing style analysis) links texts across identities. Those are worth knowing about even though tooling can't scrub them.
When it matters
Match the effort to the threat model. For everyday photo sharing, the major social platforms already strip EXIF on upload (while, of course, reading it first, and messaging apps vary: sending a photo "as a file" typically preserves everything). For anything where the audience shouldn't learn where you live or who wrote a document (listings, forums, leaks, activism, anything posted pseudonymously) stripping metadata is mandatory, and it must happen before the file leaves your machine.
How to strip it
- Phones: iOS shares photos with location removable at the share sheet (tap Options > Location off); Android's default Photos app can remove location on share. Third-party keyboards and apps vary, verify, don't assume.
- Windows: file Properties > Details > "Remove Properties and Personal Information" covers the basics.
- Cross-platform, thorough: ExifTool is the
reference implementation,
exiftool -all= file.jpgstrips everything strippable. mat2 handles many formats and is packaged in most Linux distributions. - Documents: use the application's own inspector (Word's "Inspect Document," LibreOffice's redaction tools), or export to PDF and strip the PDF, after redacting content with a tool that deletes text rather than covering it.
- The blunt instrument: a screenshot of the sensitive thing, cropped, is often the fastest reliable strip for images, at some quality cost.
Habit, not heroics
The workable practice is one habit: before any file leaves your custody for an audience you don't fully trust, ask what's riding inside it. For recurring workflows (posting listings, publishing documents), make the strip step part of the pipeline so it doesn't depend on remembering, the person in the Darwin Award remembered hundreds of times, and it's the one time he didn't that anyone knows his name.