Open-source AI Security Operations Center: alert fusion, LLM-agent triage, MITRE ATT&CK investigation, and a replayable decision ledger for every agent step. Self-hostable, runs with no API keys, MIT licensed. Ships an MCP server for Claude, Cursor and Continue.
Detection content — truth table
Generated by
scripts/detection_truth_table.py.--checkgates this file invalidate-detections.yml, so the numbers below can never quietly diverge from what the engine actually runs. Do not edit by hand.
A rule is executable when the detection engine actually loads it — when
its id appears in services/fusion/app/data/detection_ruleset.json, the
compiled ruleset DetectionEngine reads at startup.
That is the only definition that means anything, because the YAML under
detections/ is a generated projection of the Python spec modules in
scripts/detection_specs*.py — the engine never reads it. Editing a
condition: block or flipping enabled: has no effect on what fires. The
lever that reaches the engine is adding a spec and re-running
scripts/export_detection_ruleset.py.
Headline
| metric | count |
|---|---|
| rules on disk (total) | 7023 |
| executable (loaded by the engine) | 2538 |
| non-executable (provenance/coverage only) | 4485 |
| — of which: enabled, but no compiled spec | 0 |
windowed rules in windowed_ruleset.json |
76 |
The windowed figure is its own row rather than part of the headline because
a windowed rule is a different kind of thing: it counts events or distinct
values for one entity over a sliding window and fires on a threshold, so it
has no YAML under detections/ and no entry in the marketplace index. Adding
it to the executable total would make two incomparable numbers into one. The
engine carries three further built-in rules declared in
services/fusion/app/services/windowed_detection.py, which are not in the
JSON and so are not in the figure above.
50 of those rules were translated out of the stateless corpus by depth plan
3.1: they named a counter no source emits, so they were loaded, counted as
executable and could never fire. Each one's threshold and window are derived
from the original's own clauses and replayed through the real engine
(services/fusion/tests/test_windowed_translation_replay.py). The other 24 of
that family could not be expressed as a count over a window and are refused
with a reason that travels into each rule's quarantine_reason.
Four more are Sigma correlation documents compiled by
scripts/sigma_correlation.py. They are first-party content, not imported
coverage: no upstream correlation rule is vendored in this tree, so the
import path had no corpus and the rules under detections/sigma-correlations/
were hand-authored in the upstream format to give it one.
By tier
| tier | on disk | executable | enabled but not loaded |
|---|---|---|---|
| car (imported) | 99 | 0 | 0 |
| chronicle (imported) | 877 | 0 | 0 |
| community | 1 | 0 | 0 |
| native | 885 | 744 | 0 |
| sigma (imported) | 3132 | 1770 | 0 |
| splunk (imported) | 2005 | 0 | 0 |
| threat-anomaly | 24 | 24 | 0 |
Why rules are non-executable
| reason | count |
|---|---|
enabled: false |
272 |
under _quarantine/ (untranslated on import) |
4213 |
How to read the README claim
The imported corpus is large (7023 rules on disk) and valuable as a provenance-tracked ATT&CK-mapped library, but the number that matters operationally is 2538 executable rules — the ones the engine loads and fires against live telemetry. The README and marketplace must cite the executable figure when describing detection coverage, and may cite the on-disk figure only when explicitly describing the imported library.
Two things this table cannot tell you, stated so nobody infers them:
- A loaded rule still has to match on a field some connector emits. That
property is enforced separately by
scripts/check_detection_fields.py. - Executable is not the same as tuned.
false_positives:is prose and is not machine-checked; there is no per-rule false-positive-rate gate.