Executive brief
kcp front-proxy is a Kubernetes proxy used in multi-tenant deployments to route and authenticate requests between clients and cluster shards. An authenticated user can inject forged identity headers to falsely claim system administrator privileges and access any tenant's data across the cluster, breaking the core isolation that protects each customer's workloads and resources.
Technical details
The vulnerability is a request-header spoofing and privilege escalation flaw in kcp's front-proxy authentication chain. The front-proxy authenticates clients and forwards their identity to downstream shards via Kubernetes request-header authentication headers (X-Remote-User, X-Remote-Group, X-Remote-Extra-*). The root cause is that the proxy appends its verified identity headers instead of replacing any client-supplied copies, allowing an authenticated attacker to inject forged headers (e.g., X-Remote-Group: system:masters) that the shard trusts as verified identity assertions. An attacker with any valid credential (certificate, OIDC token, or service account) can assert cluster administrator privileges, bypass authorization policies, forge delegated-identity warrants, and access any workspace's secrets and RBAC configuration. The fix (v0.31.4, v0.32.2) unconditionally strips all inbound X-Remote-* headers before authentication to prevent client-supplied values from reaching the shard.
Affected products
- kcp kcp before 0.31.4, before 0.32.2
Timeline
- 2026-09-18: disclosed
- 2026-09-18: patched: v0.31.4 and v0.32.2 released