## Summary Two related authorization gaps let a host that should no longer be trusted obtain a fresh, valid Nebula certificate, because nebula-mgmt does not re-evaluate revocation/authorization state at certificate *issuance* time — only at poll time. ## 1. Blocklist not enforced at sign / re-enroll time `internal/api/enroll.go:128` calls `caMgr.Sign(...)` without consulting the blocklist. The blocklist is only checked in the poll path (`internal/api/updates.go:57`, `fingerprintInBlocklist`). The blocklist is keyed by certificate *fingerprint* (`internal/store/sqlite.go`), so a re-enrollment produces a new fingerprint that is not in the blocklist. `mintEnrollmentTokenForHost` (`internal/api/hosts.go:491`) authorizes the caller via `canAccessHost` but does **not** check the host status. There is no guard preventing a `blocked` host from transitioning back to `enrolled` (`internal/store/sqlite.go`, `enrollHostInTx` updates status unconditionally). **Impact:** A host that an operator has blocked can be silently un-blocked by issuing a new enrollment token and re-enrolling — it receives a fresh certificate (new fingerprint) that passes all subsequent poll-time blocklist checks. Revocation is therefore not durable. Requires an operator action (minting a re-enroll token), so this is an integrity/operational-revocation failure rather than an unauthenticated bypass. ## 2. Renewal does not re-validate operator / CA status Auto-renewal at poll time (`internal/api/updates.go:285-319`, `signHostCert`) reads `host.Name`, `host.Groups`, `host.NebulaIPs` from the DB and re-signs without checking whether the owning operator is still active or the CA still valid. `DisableOperator` (`internal/store/sqlite_operators.go`) revokes sessions and API keys but does not retire the operator's CAs, and `pki/signer.go` checks only CA cert time-expiry, not operator/CA status. **Impact:** A host enrolled under an operator who is later disabled continues to renew its certificate indefinitely. Offboarding an operator does not cut off the hosts they provisioned. ## Affected versions Latest tagged release (v0.3.6) and `main`. ## Remediation 1. Call a blocklist/status guard inside `handleEnroll` and `signHostCert` **before** `caMgr.Sign(...)`; refuse issuance for a host whose status is `blocked` or whose previous fingerprint is on the blocklist. Require an explicit unblock before re-enroll. 2. At renewal, re-resolve the owning operator/CA status and reject renewal if the operator is disabled or the CA retired (force re-enrollment instead). ## Discovery Found during an internal source + offensive security audit (tracking issue #178). Adversarially cross-verified against the code paths above.
## Summary Two related authorization gaps let a host that should no longer be trusted obtain a fresh, valid Nebula certificate, because nebula-mgmt does not re-evaluate revocation/authorization state at certificate *issuance* time — only at poll time. ## 1. Blocklist not enforced at sign / re-enroll time `internal/api/enroll.go:128` calls `caMgr.Sign(...)` without consulting the blocklist. The blocklist is only checked in the poll path (`internal/api/updates.go:57`, `fingerprintInBlocklist`). The blocklist is keyed by certificate *fingerprint* (`internal/store/sqlite.go`), so a re-enrollment produces a new fingerprint that is not in the blocklist. `mintEnrollmentTokenForHost` (`internal/api/hosts.go:491`) authorizes the caller via `canAccessHost` but does **not** check the host status. There is no guard preventing a `blocked` host from transitioning back to `enrolled` (`internal/store/sqlite.go`, `enrollHostInTx` updates status unconditionally). **Impact:** A host that an operator has blocked can be silently un-blocked by issuing a new enrollment token and re-enrolling — it receives a fresh certificate (new fingerprint) that passes all subsequent poll-time blocklist checks. Revocation is therefore not durable. Requires an operator action (minting a re-enroll token), so this is an integrity/operational-revocation failure rather than an unauthenticated bypass. ## 2. Renewal does not re-validate operator / CA status Auto-renewal at poll time (`internal/api/updates.go:285-319`, `signHostCert`) reads `host.Name`, `host.Groups`, `host.NebulaIPs` from the DB and re-signs without checking whether the owning operator is still active or the CA still valid. `DisableOperator` (`internal/store/sqlite_operators.go`) revokes sessions and API keys but does not retire the operator's CAs, and `pki/signer.go` checks only CA cert time-expiry, not operator/CA status. **Impact:** A host enrolled under an operator who is later disabled continues to renew its certificate indefinitely. Offboarding an operator does not cut off the hosts they provisioned. ## Affected versions Latest tagged release (v0.3.6) and `main`. ## Remediation 1. Call a blocklist/status guard inside `handleEnroll` and `signHostCert` **before** `caMgr.Sign(...)`; refuse issuance for a host whose status is `blocked` or whose previous fingerprint is on the blocklist. Require an explicit unblock before re-enroll. 2. At renewal, re-resolve the owning operator/CA status and reject renewal if the operator is disabled or the CA retired (force re-enrollment instead). ## Discovery Found during an internal source + offensive security audit (tracking issue #178). Adversarially cross-verified against the code paths above.
Update github.com/forgekeep/nebula-mesh to 0.3.7 if you use the affected versions. Test the change in a non-production environment first.
Local check
hol-guard supply-chain scannebula-mesh: Host revocation is not durable - blocked/offboarded hosts can regain a valid certificate affects github.com/forgekeep/nebula-mesh (go). Severity is medium. ## Summary Two related authorization gaps let a host that should no longer be trusted obtain a fresh, valid Nebula certificate, because nebula-mgmt does not re-evaluate revocation/authorization state at certificate *issuance* time — only at poll time. ## 1. Blocklist not enforced at sign / re-enroll time `internal/api/enroll.go:128` calls `caMgr.Sign(...)` without consulting the blocklist. The blocklist is only checked in the poll path (`internal/api/updates.go:57`, `fingerprintInBlocklist`). The blocklist is keyed by certificate *fingerprint* (`internal/store/sqlite.go`), so a re-enrollment produces a new fingerprint that is not in the blocklist. `mintEnrollmentTokenForHost` (`internal/api/hosts.go:491`) authorizes the caller via `canAccessHost` but does **not** check the host status. There is no guard preventing a `blocked` host from transitioning back to `enrolled` (`internal/store/sqlite.go`, `enrollHostInTx` updates status unconditionally). **Impact:** A host that an operator has blocked can be silently un-blocked by issuing a new enrollment token and re-enrolling — it receives a fresh certificate (new fingerprint) that passes all subsequent poll-time blocklist checks. Revocation is therefore not durable. Requires an operator action (minting a re-enroll token), so this is an integrity/operational-revocation failure rather than an unauthenticated bypass. ## 2. Renewal does not re-validate operator / CA status Auto-renewal at poll time (`internal/api/updates.go:285-319`, `signHostCert`) reads `host.Name`, `host.Groups`, `host.NebulaIPs` from the DB and re-signs without checking whether the owning operator is still active or the CA still valid. `DisableOperator` (`internal/store/sqlite_operators.go`) revokes sessions and API keys but does not retire the operator's CAs, and `pki/signer.go` checks only CA cert time-expiry, not operator/CA status. **Impact:** A host enrolled under an operator who is later disabled continues to renew its certificate indefinitely. Offboarding an operator does not cut off the hosts they provisioned. ## Affected versions Latest tagged release (v0.3.6) and `main`. ## Remediation 1. Call a blocklist/status guard inside `handleEnroll` and `signHostCert` **before** `caMgr.Sign(...)`; refuse issuance for a host whose status is `blocked` or whose previous fingerprint is on the blocklist. Require an explicit unblock before re-enroll. 2. At renewal, re-resolve the owning operator/CA status and reject renewal if the operator is disabled or the CA retired (force re-enrollment instead). ## Discovery Found during an internal source + offensive security audit (tracking issue #178). Adversarially cross-verified against the code paths above.
AI coding agents often install or upgrade packages automatically in go. A medium vulnerability in a dependency can be pulled into a project through a normal install or update without a human reviewing the change, expanding the blast radius from a single package to every agent workspace that depends on it.
| Package | Affected range | Fixed version |
|---|---|---|
| github.com/forgekeep/nebula-meshgo | <0.3.7 | 0.3.7 |
Fixed versions are reported by the source feed; confirm compatibility before updating.
Reported by GitHub Security Advisories (ghsa).
HOL Guard can help your team review package activity against supported protection paths.
Explore HOL GuardUpdate github.com/forgekeep/nebula-mesh to 0.3.7 if you use the affected versions. Test the change in a non-production environment first.
Local check
hol-guard supply-chain scannebula-mesh: Host revocation is not durable - blocked/offboarded hosts can regain a valid certificate affects github.com/forgekeep/nebula-mesh (go). Severity is medium. ## Summary Two related authorization gaps let a host that should no longer be trusted obtain a fresh, valid Nebula certificate, because nebula-mgmt does not re-evaluate revocation/authorization state at certificate *issuance* time — only at poll time. ## 1. Blocklist not enforced at sign / re-enroll time `internal/api/enroll.go:128` calls `caMgr.Sign(...)` without consulting the blocklist. The blocklist is only checked in the poll path (`internal/api/updates.go:57`, `fingerprintInBlocklist`). The blocklist is keyed by certificate *fingerprint* (`internal/store/sqlite.go`), so a re-enrollment produces a new fingerprint that is not in the blocklist. `mintEnrollmentTokenForHost` (`internal/api/hosts.go:491`) authorizes the caller via `canAccessHost` but does **not** check the host status. There is no guard preventing a `blocked` host from transitioning back to `enrolled` (`internal/store/sqlite.go`, `enrollHostInTx` updates status unconditionally). **Impact:** A host that an operator has blocked can be silently un-blocked by issuing a new enrollment token and re-enrolling — it receives a fresh certificate (new fingerprint) that passes all subsequent poll-time blocklist checks. Revocation is therefore not durable. Requires an operator action (minting a re-enroll token), so this is an integrity/operational-revocation failure rather than an unauthenticated bypass. ## 2. Renewal does not re-validate operator / CA status Auto-renewal at poll time (`internal/api/updates.go:285-319`, `signHostCert`) reads `host.Name`, `host.Groups`, `host.NebulaIPs` from the DB and re-signs without checking whether the owning operator is still active or the CA still valid. `DisableOperator` (`internal/store/sqlite_operators.go`) revokes sessions and API keys but does not retire the operator's CAs, and `pki/signer.go` checks only CA cert time-expiry, not operator/CA status. **Impact:** A host enrolled under an operator who is later disabled continues to renew its certificate indefinitely. Offboarding an operator does not cut off the hosts they provisioned. ## Affected versions Latest tagged release (v0.3.6) and `main`. ## Remediation 1. Call a blocklist/status guard inside `handleEnroll` and `signHostCert` **before** `caMgr.Sign(...)`; refuse issuance for a host whose status is `blocked` or whose previous fingerprint is on the blocklist. Require an explicit unblock before re-enroll. 2. At renewal, re-resolve the owning operator/CA status and reject renewal if the operator is disabled or the CA retired (force re-enrollment instead). ## Discovery Found during an internal source + offensive security audit (tracking issue #178). Adversarially cross-verified against the code paths above.
AI coding agents often install or upgrade packages automatically in go. A medium vulnerability in a dependency can be pulled into a project through a normal install or update without a human reviewing the change, expanding the blast radius from a single package to every agent workspace that depends on it.
| Package | Affected range | Fixed version |
|---|---|---|
| github.com/forgekeep/nebula-meshgo | <0.3.7 | 0.3.7 |
Fixed versions are reported by the source feed; confirm compatibility before updating.
Reported by GitHub Security Advisories (ghsa).
HOL Guard can help your team review package activity against supported protection paths.
Explore HOL Guard