Skip shared infrastructure in domain matching

The deterministic tracker-mapping heuristics group patterns by shared
initiator domain, but tag managers, customer-data platforms, and
generic CDNs (Google Tag Manager, Segment, cloudfront.net, ...)
initiate trackers for many unrelated vendors. Grouping on such a
domain mis-attributes one vendor's tracker to another.

Add uri.FilterSharedInfrastructureDomains backed by a curated eTLD+1
denylist and apply it once in resolveDeterministic, so sibling
grouping, catalog domain matching, and the sibling re-enqueue cascade
all ignore shared-infrastructure hosts. Vendor-specific domains such
as google-analytics.com are intentionally kept as a same-vendor
signal. The agent path is unchanged: it still sees observed domains,
now with a prompt caveat about shared infrastructure.

Update the two sibling tests that used googletagmanager.com as the
initiator domain to a vendor domain, since that host is now stripped
before grouping.

Signed-off-by: Émile Ré <emile@probo.com>
This commit is contained in:
Émile Ré
2026-05-29 20:22:46 +02:00
parent 54c05ebe6a
commit 323fe4b5c3
4 changed files with 192 additions and 16 deletions

View File

@@ -280,10 +280,11 @@ func (h *trackerMappingHandler) Process(ctx context.Context, tp coredata.Tracker
// deterministicResult carries the outcome of the pure-SQL catalog
// signals (existing link, pattern, sibling origin, domain overlap) from
// the read phase to the agent and persist phases. domains holds the
// unfiltered initiator domains observed for the pattern (used by the
// sibling re-enqueue cascade); commonThirdPartyPreexisted records
// whether a catalog third party was already known before this run, so
// the cascade only fires when this run is the one that resolved it.
// observed initiator domains for the pattern with shared-infrastructure
// hosts removed (used by the sibling re-enqueue cascade);
// commonThirdPartyPreexisted records whether a catalog third party was
// already known before this run, so the cascade only fires when this run
// is the one that resolved it.
type deterministicResult struct {
origin string
commonPatternID *gid.GID
@@ -312,9 +313,6 @@ func (h *trackerMappingHandler) resolveDeterministic(
return res, fmt.Errorf("cannot load cookie banner for domain filtering: %w", err)
}
// DEBUG: don't commit
banner.Origin = "https://t.probo.com"
res.origin = banner.Origin
if tp.CommonTrackerPatternID != nil {
@@ -349,16 +347,22 @@ func (h *trackerMappingHandler) resolveDeterministic(
return res, err
}
res.domains = loaded
// Shared tracker-delivery infrastructure (tag managers, CDPs, generic
// CDNs) initiates trackers for many unrelated vendors, so a shared
// initiator domain among them is not a same-vendor signal. Strip them
// once here so no downstream domain-overlap heuristic (sibling
// grouping, catalog domain match, or the sibling re-enqueue cascade)
// can group unrelated trackers on, say, a common googletagmanager.com.
res.domains = uri.FilterSharedInfrastructureDomains(loaded)
// Sibling matching is an org-local co-occurrence signal: two
// patterns served from the same origin on the same banner are likely
// the same vendor, even when that origin is the site's own
// (first-party) host — a tracker proxied through first-party still
// co-occurs with its siblings. So it intentionally runs on the
// unfiltered domains; the ambiguity guard in
// resolveThirdPartyFromSiblings prevents grouping unrelated
// first-party scripts.
// co-occurs with its siblings. So it intentionally keeps first-party
// domains (shared infrastructure was already removed above); the
// ambiguity guard in resolveThirdPartyFromSiblings prevents grouping
// unrelated first-party scripts.
siblingMatch, err := h.matchBySiblingOrigin(ctx, tx, tp, res.domains)
if err != nil {
return res, fmt.Errorf("cannot match by sibling origin: %w", err)