AI summary by TwinMind
This GitHub CSS Exploit Is WILD
Theo - t3․gg
6m 25s
Initial Discovery and Disclosure
- Cloud discovered CSS injection vulnerability on GitHub using math mode
- Posted on Twitter initially after failed attempts at XSS exploitation
- IP grabbers blocked by CORS policies
- Responsible disclosure after realizing potential security risks
- Giant images could mimic GitHub interface
- Button bindings could trigger unwanted POST requests
- Social engineering via crafted SVGs resembling GitHub sites
Technical Details of First Exploit
- LaTeX math syntax vulnerability allows CSS injection
- Unicode escape sequences embedded in math code
- Images hosted on GitHub domains can be leveraged
- Specific payload structure:
- Uses
\cecommand (chemical equations in LaTeX) - Encoded slash bypasses checks
- Unicode Goomba font for context escape
- x00 target with four zeros creates empty Unicode character
- Uses
Exploitation Timeline and Patches
- Initial payload posted on Twitter, gained attention
- Community experimentation revealed security risks
- GitHub security contacted for responsible disclosure
- First patch implemented - blocked Unicode macros in math syntax
- Second exploit found using HTML escaped backslash workaround
- Multiple iterations of bypasses discovered
- Final patches implemented late Friday night
- All known exploitation methods now patched
Community Response and Impact
- Reminiscent of MySpace/Tumblr customization era
- "Weeb hacker community" created anime-themed profiles
- Various creative implementations beyond basic images
- GitHub's unusually quick response compared to typical bug handling
- Positive reception of responsible disclosure approach
