trojan-source

Closing the localization gap in open source: evidence on how Indic and other under-served locales get reviewed upstream, plus i18n tooling — including i18n-security-lint, a CI scanner for defects in translated strings.

Detect and remove invisible Unicode security hazards. Trojan Source detection, CI checks, and safe RTL handling.

Scan code for invisible bidirectional Unicode characters (Trojan Source attack prevention, CVE-2021-42574)

Research-only AI watermark & provenance robustness toolkit: local reverse proxy (OpenAI/Anthropic/Gemini) + CLI stripping C2PA/EXIF/XMP, zero-width & homoglyph Unicode, KGW text watermarks, DWT/Tree-Ring image stego, AudioSeal, PDF/DOCX/PPTX/XLSX metadata, and Trojan Source (CVE-2021-42574) code scanning.

a modular offensive security framework designed for executing Unicode-based attacks, like those seen in the "GlassWorm" compromises

Desktop scanner for hidden marks and threats in Python code — invisible Unicode, steganography, homoglyphs, secrets and obfuscated code. PyQt6, RU/EN.

A security scanner designed to detect invisible Unicode vulnerabilities, BiDi overrides, and homoglyph attacks in source code to prevent Trojan Source exploits.

Find the ink you can't see. Inspects and cleans invisible Unicode, homoglyphs, and container metadata — deciding per occurrence whether a codepoint is a hidden mark or real content. Zero dependencies.

Pre-commit hook to detect and fix non-ASCII Unicode characters (smart quotes, invisible chars, Trojan Source attacks)

Check a repo before you point a coding agent at it — and check your own repo in CI for instructions nobody can see.

Local, zero-dependency toolkit for AI-text and AI-image provenance: inspect hidden characters, document how you wrote something, read C2PA credentials without pretending to verify them.

Finds the characters in text that you cannot see: zero width spaces, bidi overrides, tag payloads and homoglyphs. CLI, HTTP API and browser build