About BluePaper

BluePaper is a document sanitizer from Bluethroat Labs. You upload an untrusted document and receive a PDF rebuilt from pixels, plus a report of raw byte indicators found in the original. It is a network-service fork of Dangerzone, the open-source tool from Freedom of the Press Foundation.

By destruction, not detection. The original is opened only inside a fresh, network-isolated sandbox that renders each page to raw pixels. The sandbox is deleted, and a trusted worker builds a new PDF from those pixels. Pixels cannot carry JavaScript, macros, embedded executables, or parser exploits, so those do not reach the safe PDF.

See the full lifecycle

Everything that is not visible on the rendered page: JavaScript, macros, embedded files and attachments, form fields, clickable links, hidden layers, and most metadata.

Visible content remains as an image, including a phishing link or QR code printed on the page. Text is not selectable unless you choose an OCR language.

• PDF (.pdf)
• Microsoft Word, Excel, and PowerPoint (.docx, .doc, .docm, .xlsx, .xls, .pptx, .ppt)
• OpenDocument text, spreadsheet, presentation, and graphics (.odt, .ods, .odp, .odg)
• Hancom HWP and HWPX (.hwp, .hwpx)
• EPUB (.epub)
• Images: JPEG, PNG, GIF, SVG, BMP, TIFF, PNM, PBM, and PPM

Uploads are typically limited to 32 MiB. Other types are rejected before a sandbox is created.

Only if you pick an OCR language. Without OCR, every page is an image. With OCR, Tesseract adds a text layer on the trusted side, after rasterization, so it cannot reintroduce active content. OCR text can contain recognition errors.

BluePaper scans the original bytes for a fixed, versioned list of raw indicators such as /JavaScript, /OpenAction, /Launch, vbaProject.bin, and DDE markers. Each hit shows the indicator, a count, and the first byte offset.

The scan uses no document parser and does not decompress streams. It tells you what strings were present, not what the document would have done.

See the full catalog

No. Compression and other encodings can hide those strings, so absence of patterns is not a malware verdict. The pixel rebuild is what makes the output safe, and it runs whether or not the scan finds anything.

No. The same bytes can occur in metadata or other inert data. A hit means the indicator appeared in the original bytes, not that an active construct was present or that the file was hostile.

No. BluePaper never labels a file as malicious or clean. It gives you a rebuilt copy that cannot carry active content, plus informational byte indicators. Use it when you need to read a document you do not trust, not to decide whether the original is safe to open.

No safe PDF is produced, and the report says so. A failure is not a malware verdict: timeouts, sandbox limits, and unreadable files all fail. If the byte scan finished, its hits are still shown, which can help when a file was too hostile or broken to render.

Only inside an Azure Container Apps sandbox: a hardware-isolated microVM created from a clean image for your conversion, with all outbound network access denied. The worker copies the document in and the pixel stream out, then deletes the sandbox. It is never snapshotted or reused.

The API that receives your upload never opens it with a document parser.

The original, safe PDF, and report are stored in the operator’s Azure storage. Delete removes the safe PDF, the report, and the job record. The original is stored under its SHA-256 hash, and Delete does not currently remove it.

Anyone who has a console conversion’s id can read and delete that conversion, so treat the id as a secret. Do not upload documents you are not authorized to share.

Yes, visually. Draw or type a signature and place it on the preview. The signed copy is built in your browser and downloaded; the stored safe PDF stays unchanged. It is a visual mark, not a cryptographic signature.

Usually seconds to a few minutes, depending on page count, file type, and OCR. Each conversion runs under an operator-configured time budget, 10 minutes by default. The console polls and shows the current stage.

The web console is free to use after a Cloudflare Turnstile human check. BluePaper is licensed under AGPL-3.0. The corresponding source for the running service is on GitHub, and GET /v1/source returns the deployed revision.

Dangerzone is a desktop app that converts documents on your own computer inside a local container. BluePaper keeps Dangerzone’s converters but runs them as a web service and REST API, with each conversion in a disposable cloud microVM.

The trade-off: with BluePaper your file leaves your machine. For documents that must not leave your computer, use Dangerzone locally. BluePaper is an independent fork and is not a Freedom of the Press Foundation product.

Yes, through the REST API with an integrator API key from the operator. The web console requires a human check and must not be automated. For agents describes the workflow, and the OpenAPI specification is at /openapi.json.