Repository navigation
This release requires pySigma 2.0.1 and fixes chained correlations on SQLite 3.54. If you run chained correlations as standalone SQL on SQLite 3.54 or later, regenerate those rulesets.
pySigma 2.0.1
- Requires pySigma ≥ 2.0.1, < 3.
- pySigma 2.0.1 limits
|windashto three expandable dashes per value and placeholder expansion to 200 results. Rules beyond those limits now fail to parse or process. |refollows the rule's regex policy: withpolicy=TrustedPolicy, patterns are checked by Python'sreinstead of RE2. Values expanded from placeholders are still checked by RE2, because pySigma drops the policy during expansion.
Fixes
- On SQLite 3.54 and later, the standalone SQL of every chained correlation failed with "VIEWs and/or subqueries nested too deep". It now runs.
execute_plan(), which Zircolite uses, was not affected. - Very large chains, such as four stacked correlations, still exceed SQLite 3.54's nesting limit as standalone SQL; run them with
execute_plan(). See correlations. - With
collect_errors=True, an unsupported or invalid correlation is recorded and skipped, as detection rules already were, instead of aborting the conversion.
Changes
- Correlation errors are Sigma errors that name the rule (
SigmaCorrelationRuleError,SigmaTimespanError,SigmaCorrelationConditionError) instead of plainValueError. They still subclassValueError. observation_endis validated when the backend is constructed, not at conversion.execute_plan()raisesValueErrorfor a negativelimit. Before, a negativelimitreturned every alert.
Documentation
- The README links to the docs on the SigmaHQ repository, and each fact is stated once.
Testing
- CI tests SQLite 3.54.1 alongside the minimum, 3.38.0, and the corpus audit now covers keyword rules.
- The project lints and formats with ruff instead of black.
Full Changelog: v2.1.0...v2.1.1