2 PDF Metadata Cleanup Methods for Java - Sanitize vs Author Removal Leave feedback

PDF metadata cleanup is a GroupDocs.Metadata workflow for Java that removes identity fields from the Info dictionary and the XMP packet of a PDF before the file leaves your service. Author, Creator, Producer and Keywords are written by the tools that produce the document, and they survive “Save As”, printing to PDF and most browser downloads. A server-side step is the only place where you can check every outgoing file the same way.

Real pipelines need one of two cleanup levels. Either everything goes, because the file crosses a trust boundary, or only the people go, because Title and Subject are still needed for search. GroupDocs.Metadata for Java supports both on the same Metadata object.

This guide compares 2 cleanup methods for PDF documents, plus an inspect step before the cleanup and a verify step after it.

You will see when to call sanitize() for a full wipe, when to call removeProperties for Author-style fields only, how to get a baseline with findProperties, and how to check Author and Person.Creator after save.

In Java, findProperties and removeProperties take a Specification object instead of a lambda. Specifications are combined with .or(...) and .and(...), and the same instance can be reused for removal and for verification.

Prerequisites:

• Java 8 or later
• GroupDocs.Metadata for Java 26.11 from the GroupDocs Maven repository
• A PDF that still carries Info dictionary or XMP identity fields

Overview: Calls sanitize() on a loaded PDF to clear the metadata packages GroupDocs.Metadata detects, including the Info dictionary and XMP, and saves a new file.

Open the PDF with Metadata, call sanitize(), log the number of removed properties, and save the result to a new path. You don’t list field names: the call works on whole packages.

Since version 26.11, sanitize() also removes annotations, attachments, bookmarks, form fields and digital signatures from PDF documents. Before 26.11 it removed annotations only.

• Links for partners or public downloads
• Document exchange between tenants
• “No author trail” policies

Pros:

• One call covers the Info dictionary and XMP
• Small code that is easy to wrap in a service method
• Returns a count you can write to the log

Cons:

• Removes Title, Subject and other descriptive fields you may still need
• Removes attachments, bookmarks, form fields and signatures, which breaks fillable and signed PDFs
• After save, the PDF engine may write its own Producer value
• In evaluation mode the library does not save files, so you need a license

try (Metadata metadata = new Metadata(inputPath)) {
    int removed = metadata.sanitize();
    System.out.println(removed);
    metadata.save(outputPath);
}

Overview: Uses removeProperties with person tags and Author/Creator/Producer name checks, so the identity is removed while Title, Subject and Keywords stay.

Build a specification that matches Tags.getPerson().getCreator(), Tags.getPerson().getEditor() and the common identity property names. WithNameSpecification with ignoreCase = true matches a name regardless of case, so Author also finds the author property of the PDF document properties. removeProperties deletes the matches only.

• Records systems that index Title and Subject
• Drafts that circulate inside the company
• Policies that ban Author but allow Keywords
• Fillable PDFs that must keep their form fields

Pros:

• Selective: keeps the fields that search depends on
• The rule is an explicit object that is easy to review
• The same specification can be reused in the verify step

Cons:

• You have to update the specification when new identity fields appear
• A Producer value can be written again by the PDF engine on save

Specification authorFields = new ContainsTagSpecification(Tags.getPerson().getCreator())
        .or(new ContainsTagSpecification(Tags.getPerson().getEditor()))
        .or(new WithNameSpecification("Author", true))
        .or(new WithNameSpecification("Creator", true))
        .or(new WithNameSpecification("Producer", true));

try (Metadata metadata = new Metadata(inputPath)) {
    int removed = metadata.removeProperties(authorFields);
    System.out.println(removed);
    metadata.save(outputPath);
}

Overview: Read-only listing of Author, Creator, Producer, Title, Subject and Keywords for a before/after baseline.

• First step of any cleanup job
• Investigating “we thought the file was clean” tickets

Specification identityAndContent = new ContainsTagSpecification(Tags.getPerson().getCreator())
        .or(new ContainsTagSpecification(Tags.getTool().getSoftware()))
        .or(new ContainsTagSpecification(Tags.getContent().getTitle()))
        .or(new ContainsTagSpecification(Tags.getContent().getSubject()))
        .or(new WithNameSpecification("Author", true))
        .or(new WithNameSpecification("Creator", true))
        .or(new WithNameSpecification("Producer", true))
        .or(new WithNameSpecification("Keywords", true));

try (Metadata metadata = new Metadata(inputPath)) {
    for (MetadataProperty property : metadata.findProperties(identityAndContent)) {
        System.out.println(String.format("%s = %s", property.getName(), property.getValue()));
    }
}

For a typical PDF the output contains properties from both packages, for example author and producer from the Info dictionary and dc:creator and dc:subject from XMP.

Overview: Re-opens the cleaned PDF and prints true when no Author, Person.Creator or Person.Editor properties are left.

• Assertion in CI after the cleanup
• Telling Author removal apart from engine stamps

Specification authorLeftovers = new ContainsTagSpecification(Tags.getPerson().getCreator())
        .or(new ContainsTagSpecification(Tags.getPerson().getEditor()))
        .or(new WithNameSpecification("Author", true));

try (Metadata metadata = new Metadata(outputPath)) {
    IReadOnlyList<MetadataProperty> leftovers = metadata.findProperties(authorLeftovers);
    System.out.println(leftovers.getCount() == 0);
}

Warning

Any change saved to a signed PDF usually invalidates its digital signatures, whichever method you use. Clean the metadata before the document is signed.

Challenge: A draft contract still shows the name of the lawyer who wrote it in Author, and Keywords contain the internal case number.

Solution: Inspect the file, call sanitize(), then run the verify step before the file is uploaded to the partner portal.

Result: The log contains the number of removed properties, and the verify step prints true.

Challenge: An application form has to go to the archive without the name of the clerk who prepared it. The form fields must stay, and the archive indexes Title and Subject.

Solution: Use removeProperties with the author specification instead of sanitize(), then inspect the file to confirm that Title and Subject are still there.

Result: Author-style fields are removed, while the form fields and the descriptive fields stay in the document.

Re-open the cleaned PDF and look only for Author, Person.Creator and Person.Editor. A Creator or Producer value tagged as Tool.Software can be written by the PDF engine when the file is saved. It names the software, not the author of the document. Treat such values as expected and fail the check only when Author-style properties are left. This keeps the CI check honest and matches what a compliance reviewer asks after a metadata cleanup.

Yes. sanitize() clears the metadata packages GroupDocs.Metadata detects in the PDF, including the Info dictionary and XMP. Inspect your own files before and after the cleanup to confirm it.

Yes. Use removeProperties with a specification that matches only person tags and identity names, as in Method 2.

Without a license GroupDocs.Metadata runs in evaluation mode, which does not save files. Apply a license or a temporary license before you run the cleanup.

GroupDocs.Metadata for Java 26.11. The extended PDF cleanup of sanitize() (attachments, bookmarks, form fields and signatures) requires 26.11 or later.

Use sanitize() when a PDF leaves a trust boundary and nothing identity-related, including attachments and signatures, should go with it. Use removeProperties when Title, Subject or form fields must stay. In both cases, inspect the file before the cleanup and verify Author after it, so operators can see the evidence.

• GroupDocs.Metadata for Java - clean metadata
• GroupDocs.Metadata for Java - remove metadata properties
• GroupDocs.Metadata for Java - working with metadata in PDF documents
• Removing PII from Office documents in Java
• API reference