You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
attachments: strip EXIF metadata from image attachments #384
Lagebuch does not strip EXIF metadata from attached images. A repo-wide search for exif over src/ returns nothing; image bytes go into the .files folder verbatim and into the PDF report verbatim.
Why it matters
A photo taken with a Diensthandy at the Einsatzstelle typically carries GPS coordinates, the exact capture timestamp, and the device model. All of that travels with the attachment: into the incident's .files folder, over the sync connection to every joined device, and into the exported PDF that goes to the Akte and the Kreisbrandinspektion.
The coordinates are usually the address of a private property, and often the address of the person whose name is already in the CO-Messprotokoll. Nothing about the app's UI suggests a photo carries anything beyond what is visible in it.
docs/datenschutz-und-sicherheit.md (#380) names this as a known limit, which is the honest short-term answer, but stripping is the right one.
Shape
Strip on ingest, in IncidentFile's add path, so it applies uniformly to the local store, the sync payload and the PDF — rather than at export, which would leave the stored copy untouched.
Decide deliberately what to keep: orientation must survive or photos display rotated. Everything else (GPS, timestamps, device, serial numbers, thumbnails — which can themselves contain a pre-edit image) can go.
PNG/WebP/GIF carry metadata chunks too and should be handled in the same pass. Merged PDF attachments have their own metadata; out of scope here unless it is cheap.
Taking this off the v0.9 milestone and leaving it open as a documented known limit.
There is no image library in Directory.Packages.props. Stripping EXIF leaves two options, and neither is good:
Re-encode through Skia: lossy, orientation has to be applied by hand, and these photos are evidence.
Hand-write a JPEG/PNG/WebP/GIF segment stripper: that means parsing untrusted input (an attachment can come from a sync peer), exactly the kind of code the threat model in AGENTS.md says to avoid.
The gain is also smaller than it looks: the GPS position in an Einsatzstelle photo is usually the Einsatzadresse, which the Bericht already names. docs/datenschutz-und-sicherheit.md and SECURITY.md say plainly that EXIF is not removed. Worth revisiting if a Datenschutzbeauftragter actually asks for it, or if a small, maintained metadata-only library turns up.
Closing as not planned. Going through the fields one by one, there is no privacy gain that would justify the cost:
Field
What it adds beyond the Bericht
GPS
Nothing for a photo of the Einsatzstelle. It is the Einsatzadresse, which the Bericht names.
Capture time
Nothing. The ETB is timestamped to the minute.
Device model
Identifies almost nobody.
Owner / serial number
Rarely written by phones. Dedicated cameras do write them.
Embedded thumbnail
Only differs from the image with old cameras or desktop editors. Phone editors regenerate it.
The personal data that matters is in the pixels: faces, number plates, the inside of a home. Stripping EXIF does nothing about that. The processing is lawful (Art. 6(1)(e) DSGVO with the Feuerwehrgesetz), and the recipients (Akte, Kreisbrandinspektion, the Einsatz's own devices) are authorised. The only hook left is data minimisation, and the table shows how little there is to minimise. Against that stand the two costs from the earlier comment: re-encoding evidence, or hand-parsing untrusted image formats.
#777 rewrites the privacy page and SECURITY.md, which both still called this "planned", to describe it as a deliberate limit. It also points to the existing way to attach an image without metadata: paste it, which re-encodes it as PNG.
Reopen if a Feuerwehr's Datenschutzbeauftragter asks for it.
Activity
docs: link the tracking issues from the known limits
docs: link the tracking issues from the known limits
docs: add the German Datenschutz- und Sicherheitsseite (#380)
DeepDiver1975 commentedon Oct 17, 2026
Taking this off the v0.9 milestone and leaving it open as a documented known limit.
There is no image library in
Directory.Packages.props. Stripping EXIF leaves two options, and neither is good:The gain is also smaller than it looks: the GPS position in an Einsatzstelle photo is usually the Einsatzadresse, which the Bericht already names.
docs/datenschutz-und-sicherheit.mdand SECURITY.md say plainly that EXIF is not removed. Worth revisiting if a Datenschutzbeauftragter actually asks for it, or if a small, maintained metadata-only library turns up.🤖 Generated with Claude Code
chore: prepare release 0.9.0 (#692)
DeepDiver1975 commentedon Oct 26, 2026
Closing as not planned. Going through the fields one by one, there is no privacy gain that would justify the cost:
The personal data that matters is in the pixels: faces, number plates, the inside of a home. Stripping EXIF does nothing about that. The processing is lawful (Art. 6(1)(e) DSGVO with the Feuerwehrgesetz), and the recipients (Akte, Kreisbrandinspektion, the Einsatz's own devices) are authorised. The only hook left is data minimisation, and the table shows how little there is to minimise. Against that stand the two costs from the earlier comment: re-encoding evidence, or hand-parsing untrusted image formats.
#777 rewrites the privacy page and SECURITY.md, which both still called this "planned", to describe it as a deliberate limit. It also points to the existing way to attach an image without metadata: paste it, which re-encodes it as PNG.
Reopen if a Feuerwehr's Datenschutzbeauftragter asks for it.
🤖 Generated with Claude Code
docs: document EXIF in attachments as a deliberate limit (#777)