Skip to content

Correlation detections

A request can be suspicious without telling the whole story. Correlation detections join related, retained observations into one stable finding so an investigation shows the campaign instead of a row of near-duplicate alerts.

WebDecoy keeps the underlying detections intact. A correlation finding adds a bounded summary, representative stages, and links back to the source evidence; it does not rewrite the requests or turn an inference into an observed fact.

Finding What it brings together
Attack progression Ordered behavior such as reconnaissance, sensitive-resource discovery, credential probing, verified exploit attempts, and deception hits. Related follow-ons update one campaign finding instead of creating sibling alerts.
Multi-vector exploit campaign Distinct, independently verified exploit families observed from one strong actor.
Adaptive retry after denial A proven edge denial followed by meaningfully changed attempts against the same route family.
Crawler identity rotation Conflicting crawler claims used while probing sensitive resources.
Multi-CMS surface enumeration A compact sweep across independent administrative, discovery, content, and upload surfaces.
Authentication activity Password spraying, distributed failures against one protected subject, and elevated risk when trusted authentication outcomes follow a failure cluster.
Authorization activity Object enumeration and progressions that reach a trusted policy mismatch, unexpected allow, or decoy object.
Sensitive-resource campaign Related path variants, resource families, and recurring bursts of sensitive-file probing.
Confirmed SSRF callback A trusted SDK initiation joined to its matching single-use canary callback.
Honey-credential reuse Retrieval of an inert, property-scoped decoy credential followed by its exact presentation to a trusted integration.
Campaign progression A short-lived request cohort whose retained observations advance through multiple behavioral stages.
Framework attribution Independent, inspectable evidence that associates observed activity with a framework family at an explicit evidence tier.

This catalog describes customer-visible outcomes, not the rulebook. WebDecoy does not publish matching expressions, scoring weights, thresholds, signature sets, fingerprint construction, suppression logic, or internal evaluation order.

Depending on the finding type and available sensor coverage, a correlation card can show:

  • a stable finding ID and the retained detection that anchors it;
  • first and last observed times within the bounded campaign window;
  • an identity basis, such as a non-shared strong actor or a property-scoped trusted subject;
  • bounded aggregate counts and coverage limitations;
  • a short representative stage sequence; and
  • links to the source detections you are authorized to inspect.

Later qualifying observations update the same finding when they belong to the same bounded campaign. Separate sites, identities, or campaign windows remain separate. This reduces alert duplication without hiding the evidence that produced the rollup.