etcd is a distributed key-value store for the data of a distributed system. Prior to versions 3.5.33, 3.6.14, and 3.7.1, a network attacker who can reach an etcd TLS listener can open many TCP connections and never send a ClientHello. In client/pkg/transport/listenertls.go, each connection handled by tlsListener.acceptLoop spawns a goroutine that blocks indefinitely inside tls.Conn.Handshake() and remains tracked in the pending map. Unbounded goroutine and map growth can exhaust memory in the etcd process, causing loss of availability for the cluster and, when etcd backs Kubernetes, the control plane. This issue is fixed in versions 3.5.33, 3.6.14, and 3.7.1.

etcd is a distributed key-value store for the data of a distributed system. Prior to versions 3.5.33, 3.6.14, and 3.7.1, a user granted READ permission on a single exact key can use the Watch gRPC API with clientv3.WithFromKey() to receive watch events for every key lexicographically greater than or equal to the permitted key. In server/etcdserver/api/v3rpc/watch.go, the open-ended RangeEnd sentinel is rewritten before the RBAC permission check in server/auth/rangepermcache.go function isRangeOpPermitted, causing the request to be treated as an exact-key watch. Range/Get and DeleteRange requests are not affected, and the issue affects only clusters with authentication enabled. This issue is fixed in versions 3.5.33, 3.6.14, and 3.7.1.

Impact What kind of vulnerability is it? Who is impacted?

Multiple vulnerabilities allow unauthorized users to bypass authentication or authorization checks and call certain etcd functions in clusters that expose the gRPC API to untrusted or partially trusted clients.

In unpatched etcd clusters with etcd auth enabled, unauthorized users are able to:

- call MemberList and learn cluster topology, including member IDs and advertised endpoints - call Alarm, which can be abused for operational disruption or denial of service - use Lease APIs, interfering with TTL-based keys and lease ownership - trigger compaction, permanently removing historical revisions and disrupting watch, audit, and recovery workflows

Kubernetes does not rely on etcd’s built-in authentication and authorization. Instead, the API server handles authentication and authorization itself, so typical Kubernetes deployments are not affected.

Patches Has the problem been patched? What versions should users upgrade to?

These vulnerabilities are patched in the following versions:

etcd 3.6.9 etcd 3.5.28 etcd 3.4.42

Workarounds Is there a way for users to fix or remediate the vulnerability without upgrading?

If upgrading is not immediately possible, reduce exposure by treating the affected RPCs as unauthenticated in practice.

- restrict network access to etcd server ports so only trusted components can connect - require strong client identity at the transport layer, such as mTLS with tightly scoped client certificate distribution

Reporters Community efforts help keep etcd secure

The etcd community thanks Isaac David, bugbunny.ai, Asim Viladi Oglu Manizada, Alex Schapiro & Ahmed Allam from Strix security, Luke Francis, and @OLU-DEVX for reporting these vulnerabilities.

Dependency Between Reported Issues

These issues all originate from the same underlying flaw in the gRPC API layer.

They affect the same API surface and share a common root cause. In practice, the fix is implemented as a single, unified change at the API layer, which resolves all issues together.

Given this, we believe these issues are best treated as a single vulnerability and should be assigned a single CVE.

etcd is a distributed key-value store for the data of a distributed system. Prior to 3.5.32 and 3.6.13, when etcd is configured with --listen-client-http-urls to split HTTP and gRPC client endpoints onto separate listeners, the --client-crl-file Certificate Revocation List is not enforced on the gRPC listener, allowing a client with a revoked certificate to authenticate successfully over gRPC. This issue is fixed in versions 3.5.32 and 3.6.13.

A flaw was found in etcd, where etc-io could allow a remote attacker to gain elevated privileges on the system caused by a vulnerability in the debug function. By sending a specially crafted request, an attacker can gain elevated privileges.

Impact What kind of vulnerability is it? Who is impacted?

A vulnerability in etcd allows read access via PrevKv, or lease attachment in Put requests within transaction operations, to bypass RBAC authorization checks. An authenticated user without sufficient read or lease-related permissions may be able to access unauthorized data or attach leases by invoking transaction operations with these features enabled.

Kubernetes does not rely on etcd’s built-in authentication and authorization. Instead, the API server handles authentication and authorization itself, so typical Kubernetes deployments are not affected.

Patches Has the problem been patched? What versions should users upgrade to?

This vulnerability is patched in the following versions: - etcd 3.6.11 - etcd 3.5.30 - etcd 3.4.44

Workarounds Is there a way for users to fix or remediate the vulnerability without upgrading?

If upgrading is not immediately possible, reduce exposure by treating the affected RPCs as unauthenticated in practice.

- restrict network access to etcd server ports so only trusted components can connect - require strong client identity at the transport layer, such as mTLS with tightly scoped client certificate distribution

Reporters

Samy Ghannad (@SamyGhannad on Github) reported that read access via PrevKv in a Put request within etcd transactions bypassed RBAC authorization checks. Benjamin Wang (@ahrtr ) further analyzed that lease attachment in a Put request within etcd transactions also bypassed RBAC authorization checks

Impact What kind of vulnerability is it? Who is impacted?

An authenticated user with RBAC restricted permissions on key ranges can use nested transactions to bypass all key-level authorization. This allows any authenticated user with direct access to etcd to effectively ignore all key range restrictions, accessing the entire etcd data store. Kubernetes does not rely on etcd’s built-in authentication and authorization. Instead, the API server handles authentication and authorization itself, so typical Kubernetes deployments are not affected.

Patches Has the problem been patched? What versions should users upgrade to?

This vulnerability is patched in the following versions:

etcd 3.6.9 etcd 3.5.28 etcd 3.4.42

Workarounds Is there a way for users to fix or remediate the vulnerability without upgrading?

If upgrading is not immediately possible, reduce exposure by treating the affected RPCs as unauthenticated in practice.

restrict network access to etcd server ports so only trusted components can connect require strong client identity at the transport layer, such as mTLS with tightly scoped client certificate distribution

Reporters Our community helps keep etcd secure

SIG-Etcd thanks community members Luke Francis and Battulga Byambaa for reporting this vulnerability.

etcd is a distributed key-value store for the data of a distributed system. Prior to versions 3.4.26 and 3.5.9, the LeaseTimeToLive API allows access to key names (not value) associated to a lease when Keys parameter is true, even a user doesn't have read permission to the keys. The impact is limited to a cluster which enables auth (RBAC). Versions 3.4.26 and 3.5.9 fix this issue. There are no known workarounds.

DISPUTED Etcd v3.5.4 allows remote attackers to cause a denial of service via function PageWriter.write in pagewriter.go. NOTE: the vendor's position is that this is not a vulnerability.

etcd versions 3.2.x before 3.2.26 and 3.3.x before 3.3.11 are vulnerable to an improper authentication issue when role-based access control (RBAC) is used and client-cert-auth is enabled. If an etcd client server TLS certificate contains a Common Name (CN) which matches a valid RBAC username, a remote attacker may authenticate as that user with any valid (trusted) client certificate in a REST API request to the gRPC-gateway.

A flaw was found in etcd, where it is possible to have an entry index greater than the number of entries in the ReadAll method in wal/wal.go. This can cause issues when WAL entries are being read during consensus, as an arbitrary etcd consensus participant can go down from a runtime panic when reading the entry. The highest threat from this vulnerability is to system availability.

Vulnerability type Access Controls

Detail etcd creates certain directory paths (etcd data directory and the directory path when provided to automatically generate self-signed certificates for TLS connections with clients) with restricted access permissions (700) by using the os.MkdirAll. This function does not perform any permission checks when a given directory path exists already. Specific Go Package Affected github.com/etcd-io/etcd/pkg/fileutil Workarounds Make sure these directories have the desired permit (700).

References Find out more on this vulnerability in the security audit report

For more information If you have any questions or comments about this advisory: Contact the etcd security committee

Vulnerability type Data Validation

Detail The size of a record is stored in the length field of a WAL file and no additional validation is done on this data. Therefore, it is possible to forge an extremely large frame size that can unintentionally panic at the expense of any RAFT participant trying to decode the WAL.

Specific Go Packages Affected github.com/etcd-io/etcd/wal

References Find out more on this vulnerability in the security audit report

For more information If you have any questions or comments about this advisory: Contact the etcd security committee