* [New Rules] New Terms rules for malicious Python/Pickle model activity on macOS
Adds three new_terms SIEM detection rules to close the detection gap identified in ia-trade-team#666 where malicious pickle/PyTorch model files execute arbitrary commands via Python deserialization without triggering existing GenAI-parent-gated endpoint rules.
Co-authored-by: Cursor <cursoragent@cursor.com>
* Address PR feedback: broaden descriptions and simplify process.name
- Update descriptions across all three rules to not over-attribute to
pickle/PyTorch — these rules detect any malicious Python activity
(scripts, compromised dependencies, model deserialization, etc.)
- Simplify process.name from explicit enumeration to python* wildcard
since KQL matching is case-insensitive
- Update investigation guides to reflect broader scope of potential
attack vectors
Made-with: Cursor
* Apply suggestion from @DefSecSentinel
* Apply suggestion from @DefSecSentinel
* Apply suggestion from @DefSecSentinel
---------
Co-authored-by: Cursor <cursoragent@cursor.com>
(cherry picked from commit 49b660a135)
Colson Wilhoit
c7c889f748