Skipper routesrv CVE-2026-54246
MEDIUMSeverity by source
AV:A/AC:L/PR:L/UI:N/S:U/C:H/I:N/A:N
In-cluster adjacency limits scope to AV:A; initial pod compromise required gives PR:L; pure confidentiality exposure with no write or disruption capability confirmed.
Primary rating from GitHub Advisory.
CVSS VectorGitHub Advisory
Lifecycle Timeline
2DescriptionGitHub Advisory
Description
The routesrv component exposes the full cluster route topology (Ingress/RouteGroup configurations, backend URLs, filter chains, OAuth/OIDC callback paths) and cache-cluster topology (Redis/Valkey shard addresses) over plain HTTP with zero authentication. Any pod in the Kubernetes cluster can reach routesrv via its predictable DNS name and retrieve sensitive cluster-wide routing and cache infrastructure data.
Vulnerable Code
routesrv/routesrv.go:87-99,114-137 - all handler registrations on the main mux:
mux.Handle("/routes", b) // eskipBytes.ServeHTTP - all route data
mux.Handle("/routes/{zone}", b) // zone-scoped route data
mux.Handle("/swarm/redis/shards", rh) // Redis cluster addresses
mux.Handle("/swarm/valkey/shards", vh) // Valkey cluster addressesroutesrv/eskipbytes.go:134-196 - eskipBytes.ServeHTTP:
func (e *eskipBytes) ServeHTTP(rw http.ResponseWriter, r *http.Request) {
// ... only checks GET/HEAD method, NO auth check
if r.Method != "GET" && r.Method != "HEAD" {
w.WriteHeader(http.StatusMethodNotAllowed)
return
}
// ... serves all route data immediately
}routesrv/redishandler.go:28-41 - RedisHandler.ServeHTTP:
func (rh *RedisHandler) ServeHTTP(w http.ResponseWriter, r *http.Request) {
if r.Method != "GET" {
w.WriteHeader(http.StatusMethodNotAllowed)
return
}
// ... serves Redis cluster addresses immediately, NO auth check
}routesrv/valkeyhandler.go:28-41 - ValkeyHandler.ServeHTTP:
func (vh *ValkeyHandler) ServeHTTP(w http.ResponseWriter, r *http.Request) {
if r.Method != "GET" {
w.WriteHeader(http.StatusMethodNotAllowed)
return
}
// ... serves Valkey cluster addresses immediately, NO auth check
}Attack Path
- Initial Compromise: Attacker compromises any pod in the Kubernetes cluster (via application CVE, supply-chain attack, malicious container image, etc.)
- Discovery: Attacker discovers routesrv via predictable Kubernetes DNS name:
skipper-ingress-routesrv.kube-system.svc.cluster.local:9090(documented atdocs/tutorials/operations.md:108,docs/tutorials/ratelimit.md:137,197) - Data Extraction without Auth:
GET http://<routesrv>:9090/routes→ All Ingress/RouteGroup configurations across ALL namespacesGET http://<routesrv>:9090/swarm/redis/shards→ Redis cache cluster node addressesGET http://<routesrv>:9090/swarm/valkey/shards→ Valkey cache cluster node addresses
- Subsequent Attacks: With cache cluster topology, attacker can perform direct cache-level attacks (ratelimit data manipulation, session data exfiltration)
Permission Boundary Analysis
The routesrv uses a ServiceAccount with cluster-wide RBAC to list Ingress (networking.k8s.io), RouteGroup (zalando.org), Endpoints, and Services across all namespaces (see clusterclient.go:648-653 fetchClusterState). The kube-apiserver requires proper ServiceAccount token + RBAC authorization for the Kubernetes API itself, but routesrv exposes the aggregated data over HTTP with zero authentication.
A compromised pod with limited RBAC (restricted to its own namespace) can bypass Kubernetes RBAC entirely by reading routesrv. This crosses the boundary from *"namespace-scoped Kubernetes workload with restricted RBAC"* to *"full cluster route topology across all namespaces"*.
No NetworkPolicy manifests exist in the deploy/ directory. The default Kubernetes flat network model allows any pod to reach any service, further widening the attack surface.
Exposed Data
| Endpoint | Data Exposed | Impact |
|---|---|---|
GET /routes | All ingress/routegroup backends: internal service URLs, filter chains (auth, rate limiting, OAuth, JWT, OPA policies), load balancer group membership | Cluster-wide reconnaissance, targeted backend attacks |
GET /routes/{zone} | Zone-scoped subset of above route data | Same, scoped |
GET /swarm/redis/shards | Redis cluster internal IP:port pairs | Direct cache-level attacks, ratelimit data manipulation |
GET /swarm/valkey/shards | Valkey cluster internal IP:port pairs | Same |
Additionally, the data-plane client (eskipfile/remote.go:190-219) also performs plain HTTP GET with no credentials - only an ETag header is sent - confirming that no auth capability exists in the architecture at all.
Mitigation
- Add authentication to all routesrv HTTP endpoints (basic auth, bearer token, mTLS, or shared secret) via flag
-route-server-filters="" - Deploy Kubernetes NetworkPolicies restricting ingress to routesrv to only the data-plane skipper pod selectors
- Consider using mutual TLS authentication between data-plane and control-plane components
NetworkPolicy does not remove the missing-auth condition
Restrictive NetworkPolicies are a valid mitigation, but they are not an application-layer authentication mechanism. The security-relevant defect remains that routesrv serves control-plane-derived data to unauthenticated callers whenever network reachability exists.
Impact framing
This report does not rely on claiming direct integrity or availability impact. The verified issue is a confidentiality-focused control-plane exposure: route definitions, backend topology, filter-chain details, and Redis/Valkey shard addresses become readable to any reachable in-cluster client.
Resources
routesrv/routesrv.go:87-99- handler registration (zero auth)routesrv/eskipbytes.go:134-196- route data handler (no auth)routesrv/redishandler.go:28-41- Redis shard handler (no auth)routesrv/valkeyhandler.go:28-41- Valkey shard handler (no auth)dataclients/kubernetes/clusterclient.go:648-653-fetchClusterState()- shows cluster-wide RBACeskipfile/remote.go:190-219- data-plane client also has no auth capabilitydocs/tutorials/operations.md:108,docs/tutorials/ratelimit.md:137,197- documented routesrv DNS name
AnalysisAI
Unauthenticated exposure of full Kubernetes cluster routing topology and cache-cluster shard addresses in Zalando Skipper's routesrv component (versions prior to 0.27.13) allows any in-cluster pod to retrieve Ingress/RouteGroup configurations, OAuth/OIDC filter chains, backend URLs, and Redis/Valkey shard addresses via plain HTTP GET requests to predictable endpoints. An attacker who has established any pod foothold within the cluster can bypass Kubernetes RBAC entirely - escalating from namespace-scoped access to cluster-wide routing intelligence - enabling targeted follow-on attacks against cache infrastructure and internal backends. No public exploit code is confirmed and this CVE does not appear in the CISA KEV catalog, but the GHSA advisory provides a complete, step-by-step attack path with specific endpoint enumeration that substantially lowers the practical exploitation bar.
Technical ContextAI
This vulnerability is classified as CWE-306 (Missing Authentication for Critical Function) and affects the routesrv control-plane component of Zalando Skipper, a Kubernetes-native HTTP router and reverse proxy (Go package github.com/zalando/skipper, CPE pkg:go/github.com_zalando_skipper). In high-availability Skipper deployments, data-plane skipper instances poll routesrv for routing configuration rather than each querying kube-apiserver directly. The component registers four HTTP handlers - /routes, /routes/{zone}, /swarm/redis/shards, and /swarm/valkey/shards - on a plain HTTP mux (routesrv/routesrv.go:87-99). Each handler validates only the HTTP method (GET/HEAD) and immediately serves aggregated control-plane data with no authentication check, as confirmed in routesrv/eskipbytes.go:134-196, routesrv/redishandler.go:28-41, and routesrv/valkeyhandler.go:28-41. The service operates with cluster-wide RBAC permissions (via a dedicated ServiceAccount) to list Ingress, RouteGroup, Endpoints, and Services across all namespaces. Its DNS name (skipper-ingress-routesrv.kube-system.svc.cluster.local:9090) is documented in official Skipper tutorials, making discovery trivial. The data-plane client (eskipfile/remote.go:190-219) also sends no credentials, confirming authentication was never part of the architecture.
RemediationAI
Vendor-released patch: v0.27.13, available at https://github.com/zalando/skipper/releases/tag/v0.27.13. This release introduces support for arbitrary filter chains on the routesrv call path via the -route-server-filters flag, enabling operators to configure authentication mechanisms such as basic auth, bearer token validation, or mTLS between data-plane and control-plane components. Upgrade to v0.27.13 or later as the primary remediation. As a compensating control prior to patching, deploy Kubernetes NetworkPolicies that restrict ingress to the routesrv service on port 9090 exclusively to labeled data-plane skipper pod selectors; note that this reduces blast radius but does not eliminate the missing-authentication defect - if any pod matching the allowed selector is compromised, the unauthenticated endpoints remain reachable. Additionally, consider enabling mutual TLS between skipper data-plane instances and routesrv as a defense-in-depth measure once v0.27.13 filter chain support is configured. Do not rely on obscurity of the service DNS name; it is documented in official tutorials and discoverable by any in-cluster process.
Privilege escalation to root in the LiteSpeed User-End cPanel Plugin (versions 2.3 through before 2.4.5) lets attackers
UAF in Redis 8.2.1 via crafted Lua scripts by authenticated users. EPSS 12.4%. Patch available.
It was discovered, that redis, a persistent key-value database, due to a packaging issue, is prone to a (Debian-specific
Memory Corruption was discovered in the cmsgpack library in the Lua subsystem in Redis before 3.2.12, 4.x before 4.0.10,
Redis is an open source, in-memory database that persists on disk. Versions 8.2.1 and below allow an authenticated user
Redis before 2.8.21 and 3.x before 3.0.2 allows remote attackers to execute arbitrary Lua bytecode via the eval command.
A buffer overflow in Redis 3.2.x prior to 3.2.4 causes arbitrary code execution when a crafted command is sent. Rated cr
Code injection in OneUptime monitoring via custom JS monitor using vm module. PoC and patch available.
In applications using jfinal 4.9.08 and below, there is a deserialization vulnerability when using redis,may be vulnerab
An issue was found in Apache Airflow versions 1.10.10 and below. Rated critical severity (CVSS 9.8), this vulnerability
An Integer Overflow issue was discovered in the struct library in the Lua subsystem in Redis before 3.2.12, 4.x before 4
goanother Another Redis Desktop Manager =<1.6.1 is vulnerable to Cross Site Scripting (XSS) via src/components/Setting.v
Same technique Authentication Bypass
View allVendor StatusVendor
SUSE
Severity: Moderate| Product | Status |
|---|---|
| SUSE Linux Enterprise Server 16.1 | Affected |
| SUSE Linux Enterprise Server for SAP applications 16.1 | Affected |
| SUSE Linux Enterprise Module for Package Hub 15 SP5 | Affected |
| SUSE Linux Enterprise Module for Package Hub 15 SP6 | Affected |
| openSUSE Leap 15.5 | Affected |
| openSUSE Leap 15.6 | Affected |
Share
External POC / Exploit Code
Leaving vuln.today
GHSA-5587-2x54-jj6h