Back to all articles

Teku, Besu and eMudhra: weak TLS CA keys put some Ethereum validator connections at risk

Share

A certificate authority does not hold an Ethereum validator's private keys. But when an authority's own signing keys can be reconstructed, an attacker in the network path may impersonate HTTPS services used by some validators. That is the risk behind the eMudhra certificate incident addressed by Besu 26.10.2, released 26 October 2026, and Teku 26.10.3, released 27 October.

The consequences depend on the architecture. An impersonated external endpoint could disrupt a validator's duties or, in some circumstances, cause a proposed block to allocate execution-layer fees to an attacker-controlled address. This does not mean staked ETH deposits or withdrawal keys have been stolen. Teku says a default mainnet node without external HTTPS URLs is not affected by this scenario, and its official security advisory reports no exploitation.

The broader lesson: updating a blockchain client does not necessarily fix its TLS dependencies if Java and its certificate trust store are managed separately.

Eight recoverable intermediate CA private keys

On 17 October 2026, a security researcher reported that the private keys associated with eight eMudhra intermediate certificate authorities could be derived from their public certificates because the keys had been generated incorrectly. Independent research identifies six TLS-issuing intermediates and two code-signing intermediates.

The company acknowledged a script error in the key-generation ceremony and revoked the intermediates on 18 October. It says its root private keys, infrastructure and customer data were not breached and that none of the affected intermediates had issued end-entity certificates. These are eMudhra's reported investigation findings, not the conclusions of a completed independent review.

A cryptographically recoverable key does not imply an attacker hacked the CA's private-key storage. Nevertheless, anyone who reconstructs an affected key may be able to issue a fraudulent server certificate. An application still trusting that certification chain may then accept the wrong server as authentic.

Google documented on 20 October that Chrome had blocked the affected keys on 17 October and was arranging removal of a relevant root. Java applications do not necessarily use the same trust lists, certificate-transparency enforcement or revocation mechanisms as web browsers.

Why Teku and Besu were mentioned

Teku is an Ethereum consensus client, while Besu is an execution client. Validator operations may depend on other systems such as a remote beacon node, external signing service, builder or relay, checkpoint-sync service, or other HTTPS endpoint.

The attack scenario requires all of the following:

  1. The client actually connects to an external service over HTTPS.
  2. Its TLS environment still accepts the affected certificate chain.
  3. An attacker can intercept or redirect the network connection to a server they control.
  4. That impersonated service can influence validator operations or sensitive decisions.

Simply running Teku, Besu or an older JDK does not prove that a validator was compromised. Ordinary Ethereum peer-to-peer traffic and connections that do not rely on this HTTPS trust path are not automatically exposed by the eMudhra incident.

Could staking rewards be diverted?

Potentially under specific conditions; not automatically. Teku's security notice discusses remote signers, remote beacon nodes, relay connections and checkpoint synchronization. A forged HTTPS endpoint might return misleading information or requests. The consequences depend on the endpoint's role and independent validation.

A financial example is the execution-layer fee recipient used for block proposals. If an affected dependency can influence the fee-recipient address or proposal data, an attacker could under certain conditions cause those fees to be directed elsewhere. Remote-signing setups also require strict validation of what is being signed and who is allowed to request signatures.

These are not the same as an automatic transfer of a validator's 32 ETH stake, withdrawal credentials or all consensus rewards. There is no public evidence in the cited notices that the incident extracted validators' private keys or resulted in confirmed stolen staking funds.

What Teku 26.10.3 and Besu 26.10.2 actually fix

Their official release notes distinguish Docker images from separately installed Java runtimes:

  • Teku 26.10.3, 27 October: its Docker images no longer trust emSign ECC Root CA - G3, emSign Root CA - G1, emSign Root CA - G2 or AAA Certificate Services. Teku recommends the update.
  • Besu 26.10.2, 26 October: removes the same four roots from its Docker images, explicitly warning that standalone tar/zip installations using a separate JDK are not automatically fixed.

Some removals are precautionary and vendor policies are not identical. Eclipse Adoptium explains that its updated Temurin trust store continues to trust emSign Root CA - G1, unlike the particular OpenJDK guidance and the affected client Docker image changes. A check that flags G1 alone as proof of an unsafe updated Temurin runtime could therefore be misleading.

Two CVE references in the ecosystem should not be conflated. CVE-2026-111943 appears in the Teku release note in connection with the wider response in Go; CVE-2026-114820 tracks Java root trust-store remediation after an earlier Java blocking measure proved insufficient. Neither means that Ethereum's consensus rules were exploited.

Consult the maintained OpenJDK guidance, Oracle's security alert and Adoptium's release guidance instead of relying on stale version tables.

An operator checklist

1. Inventory sensitive HTTPS endpoints

Inspect the actual Teku and Besu configuration. Identify remote signers, remote beacon nodes, checkpoint sync URLs, relay or builder endpoints, proxies, and any other HTTPS dependencies. Record their hostnames, certificate expectations, network paths, authorization controls and responsible teams. A private network changes the attacker opportunities, but does not make verification unnecessary.

2. Replace running Docker images

Deploy Teku 26.10.3 or Besu 26.10.2, or a later verified release with equivalent protections, where affected official Docker images are used. Confirm the image digest of the running container. Pulling a new tag without recreating the workload does not replace an already running image. Use a restart plan appropriate for validator uptime and duties.

3. Inspect every separately managed Java runtime

Standalone Teku/Besu installations, other Java software and signing tools may use different JREs or custom trust stores. Identify the actual Java binary used by each service and its associated cacerts store. Apply the vendor's guidance and test the result. Updating operating-system CA packages does not always update Java trust stores.

Adoptium lists fixed October Temurin builds including 25.0.5+8, 21.0.13+9 and 17.0.21+9, with rollout depending on platform and distribution channel. Validate actual trust behavior using documented test chains and current advisories rather than relying on version strings alone. Do not assume the unsupported Java 26 branch will receive a lasting update.

4. Harden the trust boundary

Restrict access to sensitive APIs and signers. Where supported, use private links, mutual authentication, precise service identity verification and controls independent from the TLS connection for signing permissions. If manual trust-store changes are necessary, back up the store and test dependencies before removing roots: excessive removal can break legitimate TLS connections.

5. Investigate only with corroborating evidence

Review time windows, network and proxy records, signer configuration, block-proposal records and execution-fee recipient addresses. Correlate suspicious changes with actual software versions and traffic. The presence of an HTTPS endpoint is not proof of interception, but the lack of public exploitation reports is not a reason to leave exposed dependencies unpatched.

Implications for Snow-Fall and GLOV Consulting

For Snow-Fall, this is a useful example of validator infrastructure risk: signer isolation, dependency visibility, image provenance, upgrade planning and operational continuity. It does not imply that Snow-Fall runs Teku or Besu or has been affected.

Our guides to operating validators over time and Web3 infrastructure diversification explore the same operational principle: several redundant nodes may still share a common failure domain if they depend on identical Java runtimes, trust stores or external services.

For GLOV Consulting, the case underscores the need to manage indirect dependencies and trust boundaries. A third-party certificate authority may affect the security of financially sensitive infrastructure without any newly discovered defect in the application's business logic.

Evidence gaps and uncertainties

As of 30 October 2026, the cited security notices do not confirm exploitation against Ethereum validators. The number of vulnerable real-world configurations and the inventory of customized Java trust stores are unknown. Independent review of the broader eMudhra incident is ongoing.

This is not evidence of an Ethereum consensus compromise, an intrusion into Besu or Teku, or theft of validator private keys. Real exposure depends on external HTTPS connections and the exact trust chain. Remediation must be checked for each runtime.

GLOV Solutions editorial analysis — information checked on 30 October 2026. No incident affecting GLOV Solutions or Snow-Fall is alleged.

Official sources and independent technical research

Related articles

Back to all articles