Photo Privacy Lab
Technical methodology
The product reports only what its parsers and output checks can support. This page documents the current technical boundary.
Scanning
JPEG parsing walks marker segments and TIFF-style EXIF directories with strict offset checks. PNG parsing inventories chunks, including the C2PA caBX container used for Content Credentials. WebP parsing inventories RIFF chunks. The scanner classifies findings rather than executing metadata content.
Content Credentials detection reports that a supported container exists. It does not cryptographically validate the manifest, identify its issuer, or prove who created the image.
Privacy Clean
JPEG cleaning removes EXIF, XMP, IPTC, comment, and embedded-preview segments while preserving compressed scan data and ICC color profiles. When required, a new minimal EXIF block contains only the original orientation value. PNG cleaning also removes a detected C2PA caBX container. PNG and WebP cleaning removes supported privacy-bearing chunks without changing image payload chunks.
Full Flatten and redaction
The browser decodes the displayed image, applies orientation, crop, and visible masks, then encodes a fresh JPEG, PNG, or WebP. The export does not reuse the original metadata blocks or embedded preview.
Verification
Generated bytes are parsed again. Reports include categories and presence outcomes, but omit removed private source values by default. Verification does not analyze faces, signs, reflections, or other visible context.