Skip to content

Confirmed SSRF correlation

Server-side request forgery is a two-edge event: your application begins a server-side fetch, and a destination controlled for observation receives it. WebDecoy confirms SSRF only when a trusted SDK reports the first edge and the matching WebDecoy canary receives the second while its signed, single-use marker is valid.

A URL-shaped input, a request to a fetch-like route, or a callback alone is not confirmation.

  1. An authenticated, property-scoped SDK calls POST /api/v1/sdk/ssrf/markers with a route template and whether this is a self-test.
  2. WebDecoy returns a short-lived callback URL. The signed marker is scoped to the organization, property, route, expiry, and test mode.
  3. Immediately before the application performs its ordinary fetch, the SDK calls POST /api/v1/sdk/ssrf/initiate with that marker and route template.
  4. If the application reaches the callback URL, WebDecoy records a second edge and creates ssrf.confirmed-callback.v1 on the callback detection.

Each marker permits one initiation and one callback. It cannot be replayed across properties or organizations. The callback endpoint accepts only GET and HEAD, returns an empty response, and never redirects, proxies, fetches a destination, or probes a private address.

The marker lifetime and matching tolerances are bounded but intentionally not published. A stale, replayed, cross-property, or otherwise invalid marker does not create a finding.

Only bounded facts enter correlation evidence: route template, event time, callback protocol/class, redirect count, source coverage, and test mode. The token, arbitrary URL input, request and response bodies, headers, credentials, and secrets are not stored in the finding.

Open Detections, then open the SSRF callback row. A Correlated activity card titled Confirmed server-side request forgery appears directly below the verdict. It links to the initiating SDK detection. Production observations are high signal; signed self-tests are informational and remain excluded from production threat metrics. Correlation does not alter either detection’s threat score.

The MCP search_correlation_findings tool can retrieve ssrf_correlation findings as an aggregate-only projection without marker IDs, route templates, tokens, or per-event evidence.