The browser blocks application network requests
The deployed site is configured to send a Content Security Policy with a connect-src 'none' rule. That tells a supporting browser not to let the application open fetch, XMLHttpRequest, WebSocket or similar connections. Selected file bytes and extracted metadata therefore have no application network destination.
Content-Security-Policy: … connect-src 'none'; …Page requests still reach the hosting provider when the browser downloads the site itself. Those ordinary requests do not include a file selected later inside the tool. Read the complete privacy policy for the distinction.
Local input
The browser File and Blob APIs read the selected file. ImageHush does not create an upload form or application endpoint.
Self-hosted processors
Image and document libraries are served from ImageHush and loaded only when their formats need them.
Separate output
The original is not overwritten. The cleaner copy is represented by a temporary browser object URL for download.
Released temporary URLs
Temporary preview and output URLs are revoked when a file is replaced or the page is closed.
Every cleaner copy is scanned again
ImageHush does not enable the download immediately after writing a file. It runs the supported scanner over the generated bytes and stops if a targeted metadata group is still detected. This verifies the behavior the product claims; it is not a forensic guarantee about every private signal a format could contain.
Current automated test coverage
The image fixtures exercise JPEG EXIF, extended XMP, IPTC, comments, orientation handling, GIF preservation, TIFF decoding, HEIC and AVIF detection, and GPS coordinate parsing. Document fixtures create and clean PDF, DOCX, XLSX and PPTX files, reopen the result, check targeted properties and confirm retained visible content.
What is checked
- Images
- Supported metadata detection, binary cleanup paths, export paths and post-cleanup verification.
- Office
- Core, application and custom properties, thumbnails and supported attribution fields.
- Document information, catalog XMP removal, page retention and signature blocking.
- Build
- Static routes, canonicals, headings, structured data, assets, internal links and security policy.
Read the complete fixture-by-fixture test report or download its machine-readable JSON version.
Important boundaries
Metadata cleaning is not document redaction, malware scanning, authenticity analysis or anonymity. Visible faces, signs and locations stay visible. External services may retain their own records. Signed, encrypted and macro-enabled documents are blocked where rewriting would be unsafe.
Questions, corrections and reproducible format reports can be sent to [email protected]. Product principles and maintenance information are available on the About page.