gcp-wif-add-workload-user-binding
setIamPolicy on a ServiceAccount lets a principal add/loosen the workloadIdentityUser binding that maps an external WIF subject to the SA.
match
A conjunctive graph pattern. Variables (?x) bind node ids; every clause must hold.
where
node_type(?sa) ==
ServiceAccount
?principal has EFFECTIVE iam.serviceAccounts.setIamPolicy on ?sa
a WIF pool/provider exists whose principalSet:// members could be bound to ?sa
emit
| source type | Identity |
|---|---|
| target type | ServiceAccount |
| source | ?principal |
| target | ?sa |
| permissions | iam.serviceAccounts.setIamPolicy roles/iam.workloadIdentityUser |
| conditions | iam_permission |
| state logic | ACTIVE for the trust-write when the principal has effective iam.serviceAccounts.setIamPolicy on ?sa and no deny policy denies it. Becomes a realized federation only once a WIF provider accepts an attacker-controllable subject the new/loosened principalSet:// binding resolves — chain via rule 1. BLOCKED if an IAM deny policy denies setIamPolicy at the SA scope. |
Narrative
{principal.name} can set the IAM policy of {sa.name} (iam.serviceAccounts.setIamPolicy) and add or broaden a roles/iam.workloadIdentityUser binding for a principalSet:// member an attacker-controlled external subject satisfies - wiring WIF impersonation of {sa.name}.
Raw rule rules/derived/gcp/wif.yaml
id: gcp-wif-add-workload-user-binding
emits: CanModifyTrust
description: setIamPolicy on a ServiceAccount lets a principal add/loosen the workloadIdentityUser binding
that maps an external WIF subject to the SA.
applies_to:
- gcp
match:
- - principal: null
- CanModifyTrust
- sa: null
where:
- node_type(?sa) == ServiceAccount
- ?principal has EFFECTIVE iam.serviceAccounts.setIamPolicy on ?sa
- a WIF pool/provider exists whose principalSet:// members could be bound to ?sa
emit:
source_type: Identity
target_type: ServiceAccount
source: ?principal
target: ?sa
permissions:
- iam.serviceAccounts.setIamPolicy
- roles/iam.workloadIdentityUser
conditions:
- iam_permission
state_logic: "ACTIVE for the trust-write when the principal has effective iam.serviceAccounts.setIamPolicy\
\ on ?sa and no deny policy denies it. Becomes a realized federation only once a WIF provider accepts\
\ an attacker-controllable subject the new/loosened principalSet:// binding resolves \u2014 chain\
\ via rule 1. BLOCKED if an IAM deny policy denies setIamPolicy at the SA scope."
confidence: min(contributing_confidences) * 0.95
derived_from:
- ?principal CanModifyTrust ?sa (Phase-2 gcp-sa-set-iam-policy, iam.serviceAccounts.setIamPolicy)
false_positive_note: "Config-side primitive only \u2014 adding a workloadIdentityUser binding yields\
\ federation only if a pool provider will ALSO accept a subject the attacker controls (rule 1 / rule\
\ 3). This CanModifyTrust reuses the same iam.serviceAccounts.setIamPolicy permission as Phase-2 gcp-sa-set-iam-policy\
\ (which already emits CanModifyTrust for the SA-token-creator self-grant case); dedupe on (CanModifyTrust,\
\ ?principal, ?sa) and do not double-count weight. This rule exists to make the WIF-binding-write\
\ intent explicit in the narrative."
narrative: "{principal.name} can set the IAM policy of {sa.name} (iam.serviceAccounts.setIamPolicy)\
\ and add or broaden a roles/iam.workloadIdentityUser binding for a principalSet:// member an attacker-controlled\
\ external subject satisfies \u2014 wiring WIF impersonation of {sa.name}."