WatermarkScanner™
Unicode Steganography & Hidden Mark Detection
WatermarkScanner™ performs deep character-level scanning to detect 7 types of hidden steganographic markers embedded in text: zero-width invisible characters, homoglyph substitutions (look-alike Unicode), invisible format characters, spacing steganography, trailing space patterns, statistical token bias, and metadata markers. It exposes the exact position, type, and content of every hidden mark.
How It Works
Character-Level Scan
Iterates through every Unicode code point in the text, classifying each by Unicode category. Characters in the Cf (format), Zs (space separator) non-standard, or Mn (non-spacing mark) categories outside expected positions are flagged as suspicious.
Homoglyph Detection
Checks every word character against a comprehensive homoglyph table mapping visually identical characters across scripts (Cyrillic "а" vs Latin "a", Greek "ο" vs Latin "o", etc.). Detects > 500 known homoglyph pairs.
Spacing Pattern Analysis
Analyzes inter-word and intra-word spacing for non-standard space characters (U+2002–U+200A, U+3000, U+FEFF) that can encode binary data. Decodes detected spacing steganography patterns.
Clean-Text Output
After detection, applies surgical character cleaning to remove or normalize all detected markers while preserving the entire visible text content. Outputs both a clean text and a forensic report listing all removed markers with positions.
Key Metrics
Use Cases
- Removing hidden tracking marks from AI-generated text before publishing
- Detecting whether received text contains invisible identification data
- Cleaning homoglyph-poisoned text for NLP pipelines
- Forensic analysis of suspicious text documents
Try WatermarkScanner™ Now
Powered by our proprietary technology stack. Available directly in the TextHumanize tools and via REST API.