Why Black Highlighter Redaction Fails to Secure PDFs

The illusion of coverage

Drawing a black rectangle over sensitive text feels intuitive. The information disappears from view, so it must be gone. This assumption has caused data breaches affecting millions of people, including high-profile failures by government agencies, law firms, and corporations who should have known better.

The problem is fundamental: PDFs store text as data objects separate from their visual appearance. When you use a highlighter tool, annotation feature, or draw a shape over text, you're adding a new visual layer on top of existing data. The original text remains fully intact in the file structure, trivially extractable by anyone with basic technical knowledge.

How PDF structure exposes "hidden" text

A PDF file contains multiple distinct components:

• Content streams: The actual text characters, their positions, and encoding
• Annotation objects: Comments, highlights, form fields, and drawn shapes
• Appearance streams: Visual rendering instructions
• Metadata: Author, creation date, software used, revision history

When you highlight text black using tools like Preview's markup, Microsoft Word's drawing tools, or most PDF annotation features, you create an annotation object that sits in a separate layer. The content stream underneath remains unchanged.

Extraction methods that defeat black boxes

Select and copy: In many cases, you can simply click and drag to select text beneath a black rectangle. The selection highlights through the overlay, and copied text pastes normally.

Text extraction tools: Command-line utilities pull all text regardless of visual layers:

pdftotext document.pdf - | grep -i "social security"

strings document.pdf | grep -E "[0-9]{3}-[0-9]{2}-[0-9]{4}"

PDF readers' accessibility features: Screen readers and text-to-speech tools read the underlying content, ignoring visual annotations.

Layer manipulation: Opening the PDF in Illustrator, Inkscape, or even some PDF editors allows moving or deleting the black rectangle to reveal text beneath.

Programmatic extraction: A few lines of Python with libraries like PyMuPDF or pdfplumber extract every text object:

import fitz
doc = fitz.open("redacted.pdf")
for page in doc:
    print(page.get_text())

Real-world redaction failures

These aren't theoretical risks. Documented incidents include:

TSA airport security procedures (2009): A "redacted" PDF released under FOIA contained black boxes over screening procedures. Copying and pasting revealed the complete text, exposing security protocols.

Paul Manafort court filing (2019): Attorneys submitted a PDF with black rectangles over text. Journalists extracted the hidden content within minutes, revealing details prosecutors had sought to protect.

Australian government budget documents (2016): Sensitive financial figures "redacted" with black highlighting were extracted by journalists, exposing protected budget information.

Countless legal filings: Court records regularly contain improperly redacted documents where Social Security numbers, addresses, and financial details remain extractable.

Why common tools create false security

Preview on macOS

The markup toolbar's highlighter and shape tools create annotation layers. Even the "Redact" option in some versions doesn't flatten content—it requires an explicit "Apply Redactions" step that users often skip. See How to Redact a PDF Offline on a Mac (Securely) for the correct workflow.

Microsoft Word

Opening a PDF in Word, drawing black shapes, and re-exporting creates a new PDF with the original text intact beneath shapes. Word's drawing tools never modify the underlying content stream.

Basic PDF editors

Many free PDF tools advertise "redaction" but only add visual overlays. Unless the tool explicitly states it removes text from the content stream and flattens the document, assume it doesn't.

Print-to-PDF after highlighting

This sometimes works accidentally—if the print driver rasterizes the page, text becomes pixels. But it's unreliable, may preserve text in some implementations, and doesn't address metadata. It's not a redaction method; it's hoping for a side effect.

What actual redaction requires

Secure redaction must accomplish three things:

1. Remove text from content streams: The character data must be deleted, not covered
2. Flatten or rasterize the visual layer: Ensure no recoverable vector text exists
3. Strip metadata and hidden content: Remove revision history, comments, embedded files, and document properties

Verification steps

After any redaction, verify your work:

# Extract all text - should show nothing sensitive
pdftotext redacted.pdf - | less

# Search for patterns you redacted
strings redacted.pdf | grep -iE "(ssn|password|confidential)"

# Check metadata
exiftool redacted.pdf

Open the file in a different PDF reader than you used for redaction. Try selecting text in redacted areas. If you can highlight or copy anything, the redaction failed.

Secure redaction methods

Dedicated redaction tools

Adobe Acrobat Pro's redaction feature (not the highlighter—the actual Redact tool under Tools) removes content when you apply redactions. It also offers "Remove Hidden Information" to strip metadata.

For offline alternatives that don't require subscriptions, see How to Redact a PDF on Windows Without Adobe Acrobat.

Rasterization approach

Converting pages to images eliminates text objects entirely:

# Convert to images, then back to PDF
pdftoppm -png -r 300 input.pdf page
convert page-*.png -quality 100 output.pdf

This destroys all text data but also eliminates searchability and increases file size. For documents where you need to redact specific sections while keeping other text searchable, use proper redaction tools instead.

Metadata removal

Even after content redaction, metadata can expose sensitive information. Author names, revision history, and software details persist unless explicitly removed:

exiftool -all= redacted.pdf

qpdf --linearize --remove-unreferenced-resources=yes input.pdf output.pdf

The guide Free Local PDF Metadata Scrubber: A Complete Guide covers this process thoroughly.

Key takeaways

• Black highlighters, shapes, and annotation tools add visual layers without removing underlying text
• PDF content streams store text separately from visual appearance
• Basic command-line tools extract "hidden" text in seconds
• Proper redaction requires removing text from content streams, not covering it
• Metadata and revision history can expose information even after visual redaction
• Always verify redactions with text extraction tools before sharing documents

Verify your work

Before sending any redacted document:

1. Open in a different application than you used for redaction
2. Attempt to select text in redacted areas
3. Run pdftotext or strings and search for sensitive patterns
4. Check metadata with exiftool
5. If the document contains truly sensitive information, have a second person verify independently

The extra five minutes of verification prevents the permanent exposure that comes from a redaction failure. Once a document with extractable "redacted" content reaches the internet or an adversary, the information cannot be recalled.

Related guides

• Securely Black Out Text on Tax Forms Before Sending
• Is It Safe to Redact PDFs Online? A Risk Assessment
• How to Redact a PDF on Windows Without Adobe Acrobat