GHSA-8fh7-7q4q-cq52 · miren.dev/runtime
GHSA-8fh7-7q4q-cq52 · Reported 19 Nov 2026
Full advisory details, affected packages, severity context, and original source.
Affected technology
Affected versions and fixes
Version information comes from OSV, NVD, and the CVE vendor record. Each source may describe a different package, release branch, or platform.
miren.dev/runtime · Go
OSV- All earlier versions · before 0.11.1 (exclusive)
Fixed releases / commits: 0.11.1
Advisory overview
### Summary The Miren Runtime RPC server (QUIC/HTTP3, default port `:8443`) requested TLS client certificates but never verified that a presented certificate chained to the cluster Certificate Authority. Because a client-certificate identity is granted a blanket RBAC bypass, any party able to reach the RPC listener could present a forged, self-signed certificate and be authenticated as a cluster **superuser** — no valid CA signature, credential, or prior authorization required. ### Impact This is a complete, unauthenticated authentication bypass against the cluster control plane. An attacker who can reach the RPC port can perform any RPC operation with full privileges: enumerate and manipulate entities, apps, sandboxes, disks, and other cluster resources. This yields full compromise of confidentiality, integrity, and availability of the cluster and everything it orchestrates. Exposure is network-reachability–gated only. Deployments that expose `:8443` beyond a trusted network (e.g. `0.0.0.0/0`) are remotely exploitable from the internet by anyone; deployments that restrict `:8443` to a trusted network are exploitable by any party with access to that network. Due to a lack of ability to detect if the RPC has been accessed unauthorized, Miren recommends that all applications are redeployed and any credentials reference in environment variables are rotated. A read-only script to help review standing access, recent changes, and the exact secrets to rotate is available at https://github.com/mirendev/runtime/tree/089582a132f7fc1e1d5b5c0aee1e0911535fed53/scripts/security/audit-exposure. ### Details Three independent factors combined to produce the bypass: 1. **The listener never verified the client-certificate chain.** When a cluster CA was configured, the RPC listener used `tls.RequestClientCert`, which requests a client certificate but does not verify it against the configured `ClientCAs`. `tls.ConnectionState.VerifiedChains` was therefore always empty, and no `VerifyPeerCertificate`/`VerifyConnection` hook re-checked the chain. 2. **The authenticator derived identity from an unverified certificate.** The RPC authenticators read `PeerCertificates[0].Subject.CommonName` directly as the caller identity, with method `cert`, without checking that the chain had been verified. 3. **Authorization granted a blanket bypass to certificate identities.** The authorizer returned "allowed" for any `cert`-method identity, skipping all RBAC checks. Go's TLS client only presents a certificate whose issuer Distinguished Name matches a CA advertised by the server, but that comparison is on the (public, non-secret) DN bytes only — it does not require a valid signature from the CA. An attacker can therefore mint a self-signed certificate whose **issuer DN is byte-identical to the cluster CA subject**, signed with an attacker-controlled key. The stock client presents it, and the vulnerable server accepts it and trusts its CommonName as a superuser identity. ### Patches Fixed by verifying the client-certificate chain before trusting it: - The listener now uses `tls.VerifyClientCertIfGiven` when a cluster CA is configured. A presented certificate must chain to the configured CA to complete the handshake, while certless callers (JWT/OIDC, which authenticate via the `Authorization` header) are still admitted. - As defense in depth, both RPC authenticators now require a non-empty `VerifiedChains` before deriving an identity from a client certificate. Users should upgrade to the patched release and restart the server so the new listener configuration takes effect. ### Workarounds Until the patch is deployed, restrict network access to the RPC port (`:8443`) so that it is not reachable from untrusted networks — remove any `0.0.0.0/0` ingress and limit it to trusted management networks. This is a containment measure only and does not remove the underlying vulnerability.
Also known as
CVE-2026-75521
