Skip to content

Managing Watermarks

Enterprise documents often contain watermarks ("Confidential", "Draft", "Property of X") injected by Document Management Systems (DMS) or third-party tools (like DocuSign or Adobe).

While usually benign, watermarks can sometimes trigger security alerts in DocFirewall because they use techniques similar to obfuscation attacks, such as:

  1. Invisible Text (T3/T9): Placing text in hidden layers or using white-on-white text for tracking.
  2. Active Content (T2): Using JavaScript to dynamically render the current date or username.

Will my watermarks be flagged?

Visible Watermarks (Safe)

Standard, visible text overlays (e.g., a gray "DRAFT" diagonal text) are generally safe. They are treated as normal document content.

Hidden / Tracking Watermarks (Risk: High)

Some systems inject invisible metadata or hidden text layers to track document leakage. * Trigger: T3 (Obfuscation) or T9 (ATS Manipulation). * Reason: DocFirewall detects text that is present in the file structure but not visible to the user, assuming it might be a prompt injection or SEO poisoning attempt.

Dynamic Watermarks (Risk: High)

Watermarks that use PDF Scripts (JavaScript) to update automatically. * Trigger: T2 (Active Content). * Reason: Executable code in PDFs is a primary vector for malware.

Smart Watermark Bypass

DocFirewall v0.2+ includes a allow_hidden_watermarks setting. When enabled (default), the scanner will ignore hidden text anomalies if the text matches common watermark patterns.

Keywords ignored: confidential, draft, internal use, property of, copyright, do not copy, generated by.

Configuration

To disable this behavior (make strict):

config.allow_hidden_watermarks = False