What hidden information can a PDF leak before you send it?
A PDF file is a structured binary container that often stores author names, software versions, embedded attachments, reviewer notes, and earlier saved drafts invisibly behind the visible pages. Scanning the file before emailing or publishing it exposes those hidden layers so you can strip them locally with 0 bytes sent over the network.
Inspecting Info dictionary keys and XMP metadata streams
Authoring tools routinely stamp your computer username, software build, and exact creation timestamps into the document’s Info dictionary. According to the Library of Congress PDF 1.7 Format Description, ISO 32000 stores these details across 8 standard Info keys (Author, Creator, Producer, Title, Subject, Keywords, CreationDate, and ModDate) as well as structured XMP XML packets.
Attorneys and compliance teams frequently clean legal brief PDF metadata before filing to prevent opposing counsel from seeing draft timestamps or workstation paths. When you want to replace an author or title rather than wipe every property, you can review and update document properties selectively.
Detecting embedded files, scripts, and reviewer comments
Beyond metadata, PDFs can carry hidden EmbeddedFiles attachments, PieceInfo application blobs, document-level JavaScript actions, and reviewer comment annotations that expose internal feedback. Knowing how to tell if a PDF has hidden scripts or attachments takes one drop here because the scanner walks OpenAction, AA triggers, bookmarks, and page annotations automatically.
Flagging each item with a High, Medium, or Low risk badge lets you audit exactly what is inside before clicking Remove items. Review each listed author name, attached filename, or external domain so you never accidentally strip a link or annotation you intended to keep.
All nine risk categories and what the cleaner removes
Each finding is classified across 9 risk categories by whether the cleaner can strip it automatically without altering visible page layout. The table below maps all 9 scanned categories under Library of Congress PDF 2.0 Format Description structural rules, where XMP metadata replaces legacy Info keys.
| Finding category | What it inspects | Auto-removable | Handling during Clean PDF |
|---|---|---|---|
| Metadata & XMP | 8 Info keys, XMP streams, PieceInfo | Yes (checkbox) | Clears Info dict and deletes Metadata/PieceInfo |
| Attachments | EmbeddedFiles and FileAttachment annots | Yes (checkbox) | Removes embedded file trees and annotations |
| JavaScript | OpenAction, AA triggers, outline/field JS | Yes (checkbox) | Strips script actions from catalog and pages |
| Comments | Non-link, non-widget page annotations | Yes (checkbox) | Deletes review notes while keeping links intact |
| Edit history | Multiple %%EOF incremental trailers |
Automatic on clean | Purged when rewriting object graph from scratch |
| Forms & layers | AcroForm fields and optional content groups | Informational | Directs you to flatten before sharing |
For documents containing fillable fields or toggleable CAD layers, running our browser-based PDF flattening tool beforehand locks visible values onto the page so no hidden state remains.
Why incremental PDF saves leave earlier versions intact
Every time a desktop PDF editor performs a quick save, it appends modified objects to the end of the file rather than rewriting the whole document. That incremental design explains why people can see earlier versions of a PDF even after paragraphs or signature boxes appear deleted.
Counting %%EOF markers to spot appended edit history
The scanner inspects raw PDF bytes for %%EOF end-of-file markers to count how many incremental revisions are stacked inside the document. Finding 2 or more %%EOF markers warns you that previous versions of the file may still be recoverable with a hex viewer.
Rebuilding the PDF object graph from scratch
Zeroing out a metadata string in place does not satisfy true sanitization if abandoned objects stay in the file body. Following the core principle of NIST SP 800-88 Rev. 1 Guidelines for Media Sanitization, effective sanitization must purge residual data structures rather than merely hiding pointers.
When you click Clean PDF or Remove items, the cleaner walks every reachable object from the root catalog, drops unreachable revisions, and writes a fresh single-revision PDF using object streams. Afterward, if your document also contained blacked-out text boxes, you can verify that blacked-out boxes actually removed underlying text.
Combining metadata cleaning with redaction and encryption
Metadata cleaning and visual redaction solve two different privacy problems. Stripping metadata removes invisible file properties, comments, scripts, and edit history while leaving every visible word on the page intact.
Redacting visible text before a final privacy sweep
If your PDF still shows confidential account numbers, addresses, or names in the visible page text, scrub them permanently before running a final privacy scan. Cleaning metadata alone never hides visible body text, whereas combining visual redaction with object-graph pruning protects both layers.
Confirming that both visible text and hidden document properties are clean before hitting send prevents accidental data leaks in legal, HR, and financial workflows.
Archiving and locking cleaned PDFs for distribution
Once the scanner reports 0 remaining findings, you can archive cleaned PDFs in long-term PDF/A format or add password encryption before sending. According to the MDN Web Workers API Documentation, both the scan and the object-graph rewrite execute inside an isolated browser worker thread with 0 bytes transmitted anywhere.
Saving the cleaned copy under a distinct filename keeps your original working draft intact on disk while ensuring only the sanitized revision leaves your device. Checking the output size in the result panel also confirms that unreachable revision history has been pruned away.
Tip: Always run Privacy scanner as your final check after redacting or flattening a document, since editing a PDF in another app can re-introduce
Creator,ModDate, and incremental%%EOFhistory.