Indicators¶
Overview¶
An indicator rule asks one question: does a declared surface match an entry in
signed external data? The mechanism is membership, not a heuristic - the
answer is exact equality against a list TrustSight did not invent. H056 reads
the shipped iocs.toml and any locally imported baseline; the IOC federation
stage reads a publisher's signed indicator set. They are the same mechanism, so
H056 is filed here and keeps its id; new federation matches will take I ids.
Indicator matches are outside the heuristic score and never add points. A
confirmed match is FATAL; the rest map to a severity by the indicator's
confidence tier. See IOC Federation for the transport and the
update cadence.
See Rule Mechanics for the field table, the severity weights and the reserved identifier ranges.
Threat context¶
An indicator is a wanted poster. It does not reason about behaviour; it is a specific, named observable - a face, a plate number, a fingerprint - that someone with authority has put on a list. The indicator family matches those specific observables rather than shapes: a known-bad host, a known campaign's literal, a specific metadata endpoint. These are the atomic facts that the collective correlation rules (G-series) aggregate, and the ones a maintainer can override when a stated use is legitimate.
A wanted poster needs a signature and a date for the same reason. Indicators age: a host that was malicious last year may be a recycled CDN today, so indicator findings carry provenance and date, and the override-tolerance logic exists to let a known-good explanation suppress one without suppressing the class.
Detection mechanisms¶
The rules that make up this category are listed below, each with the series that implements it and the severity it carries. Scoring, matching and identifier allocation are shared across every category: see Rule Mechanics.
Individual rules¶
| Rule | Name | Severity |
|---|---|---|
| H056 | Known Indicator of Compromise | FATAL |
H056: Known Indicator of Compromise¶
- Severity: tiered by the indicator's confidence:
confirmedis FATAL,highis CRITICAL,mediumis HIGH, an untiered entry is MEDIUM. Apkgbuild_patternentry is capped at CRITICAL whatever its tier, because a regex names a shape rather than an exact artefact. -
Category:
ioc -
Condition: A declared surface exactly matches an entry in
iocs.toml, or an added line matches apkgbuild_patternentry.
Five surfaces are read: the package's own name; names added to
depends/makedepends/optdepends/checkdepends/provides/replaces/
conflicts; the host of any URL and any bare host token; any hex digest of
digest length; and a file the diff added or modified, or the recipe names via
install= or a local source=() entry. A pkgbuild_pattern entry is matched
as a regular expression against added lines; an invalid or
catastrophic-backtracking pattern is refused at load rather than run.
The IOC federation stage is separate and reports attributed matches through
PackageFact.ioc_matches; see IOC Federation.
Matching is exact. Normalisation is limited to what is not part of the identity: case, a trailing root dot, IDNA spelling, surrounding quotes. A host is never stripped of a subdomain and a name is never stemmed, because that would turn equality into resemblance. A malformed entry is dropped with a warning rather than coerced.
The shipped list is empty. TrustSight does not invent indicators, so a fresh
install cannot fire H056 at all, and a miss is uninformative. trustsight corpus
pivot <indicator> inverts the rule: given one indicator, it lists every corpus
package referencing it, reading only stored material and never the network.
A confirmed indicator cannot be suppressed through overrides.json.
Interpretation¶
- The finding is a concrete match, with the matched observable and its provenance (which corpus or campaign it came from).
- An indicator match is strong evidence of the specific past activity; it is not proof the current diff is malicious, because observables are reused.
- An indicator suppression must be terminal-visible: when a reviewer overrides one, the report still shows what was suppressed and why, so a stale override cannot hide a re-appearance.