Skip to content
NewCybercrime Report 2026 · Eight Chapters · A Post Is Not An Incident → Explore The ChaptersCybercrime Report 2026 → Explore
Security NewsNews Explainer

Mosquitto WebSocket RCE: Affected Brokers and Fixes

Mosquitto WebSocket RCE affects 2.1.0 to 2.1.2 brokers using the built-in listener. Check exposure, choose a fixed release, and verify the running build.

By
Brandefense Research Team
Reviewed by
Brandefense Analyst Team
Published
18 November 2026
Reading time
7 min read
Guarded broker and WebSocket gateway during a software update

Mosquitto WebSocket RCE affects Eclipse Mosquitto 2.1.0 through 2.1.2 when a broker uses its built-in WebSocket listener. Eclipse lists 2.1.3 as patched, and its November 17 release calls the issue critical. Operators should identify the running build and listener implementation before deciding which brokers need an urgent change.

The vulnerable path can corrupt memory while the broker reads an MQTT connection request, before MQTT authentication finishes. A research report demonstrates a broker crash without credentials and code execution when an anonymous session is available. This distinction matters even where a password is required.

The technical scope is recorded under CVE-2026-110486 in the Eclipse advisory. An external attack surface management view can help locate reachable brokers, while local records establish their builds.

Mosquitto WebSocket RCE: Which Brokers Are in Scope?

Version alone is insufficient. Eclipse’s advisory requires both a broker in the 2.1.0 through 2.1.2 range and a listener using the built-in WebSocket implementation. A standard MQTT listener is a different protocol path. The advisory names a build using libwebsockets as a mitigation for those versions.

Check the effective configuration for a listener with protocol websockets. Mosquitto’s configuration manual says mqtt is the default protocol, while websockets tunnels MQTT over WebSockets. Follow any include_dir settings because the main file may load additional listener definitions. Then verify the actual running binary and how that package was built.

Broker finding Decision
2.1.0 to 2.1.2, built-in WebSocket listener In Eclipse’s affected scope. Upgrade or apply a named mitigation.
2.1.0 to 2.1.2, libwebsockets build Eclipse lists this implementation as a mitigation. Confirm the running build uses it.
2.1.3 or later upstream build At or beyond Eclipse’s patched release. Verify the process after restart.
Version or WebSocket implementation unknown Keep in the review queue until both facts are verified.

The table reflects the upstream advisory, not a promise about every distributor’s package. A distributor may backport a fix while retaining an earlier version string. Obtain that maintainer’s security statement before marking a packaged broker fixed. Record reachability separately, since an internal listener can still serve untrusted clients.

How Does the WebSocket Reader Reach MQTT Memory?

The Eclipse vulnerability report traces the issue from the built-in WebSocket reader into the generic MQTT packet reader. Under the reported condition, the first reader can count WebSocket framing data as if it had delivered MQTT payload bytes. The second reader trusts that count and moves its position within a heap buffer.

That mismatch leaves memory treated as received data even though the WebSocket reader supplied no matching payload. The research report describes both disclosure of uninitialized heap contents and a counter underflow that permits an out-of-bounds write. Eclipse’s release note summarizes the flaw as incorrect handling of WebSocket frames with potential remote code execution.

Where the broker loses track
  1. WebSocket listenerReceives framed MQTT data
  2. Built-in readerReports progress without matching payload bytes
  3. MQTT packet readerAdvances its memory position using that count
  4. Broker heapDisclosure or memory corruption may follow

The flow is a high-level explanation of the report, not a packet recipe. The read and write effects have different preconditions in the researcher’s account. That is why a configuration review must distinguish pre-authentication crash risk from the reported code-execution chain.

What Changes When Anonymous Access Is Disabled?

The researcher reports that a malicious connection can reach the out-of-bounds write during CONNECT handling before MQTT authentication. On a broker requiring credentials, the report demonstrates memory corruption and a crash without an authenticated session. Disabling anonymous access therefore does not remove the denial-of-service exposure in the affected listener.

The report’s laboratory code-execution demonstration used an anonymous session on an unmodified 2.1.2 broker. Its disclosure path requires an established session, either anonymous or authenticated. The report expects a user with valid credentials to reach the same chain on a broker requiring login, but labels that case as not separately tested.

Eclipse’s advisory headline says unauthenticated remote code execution and its description says potential code execution. Read that headline together with the report’s stated preconditions. An operator should not equate a possible chain with proof that an unknown remote actor has executed code in their environment.

Which Release Should an Operator Deploy?

Eclipse lists 2.1.3 as the patched version for this WebSocket flaw. Its November 17 release also notes that 2.1.4 fixes a legacy SHA-512 password regression introduced in 2.1.3. Teams using those passwords should account for that regression during upgrade planning and choose a build containing both corrections.

Keep the patch decision tied to the deployed artifact. Check the distributor’s release and patch status if the broker comes from an operating-system package or appliance. Restart the relevant broker process, confirm its active version, and check that the intended listener remains available. A changed package on disk is not evidence that the running process changed.

Eclipse offers two temporary risk reductions for affected versions: disable WebSocket support or build with libwebsockets. Either option changes the deployment and may affect clients. Test compatibility and document the owner, deadline, and compensating control. Track the change through a patch management process.

What Should the Broker Owner Check First?

Begin with an inventory of every running Mosquitto instance, including containers and replicas. For each one, record the binary version, package source, WebSocket build implementation, listener configuration, authentication settings, and network reachability. Local build records establish the implementation, while network inventories establish reachability.

Prioritize reachable built-in WebSocket listeners running 2.1.0 through 2.1.2. Include password-protected instances because the report’s crash path occurs before MQTT authentication. Check the effective configuration rather than assuming a default file represents every process. Mosquitto’s manual permits included configuration files and multiple listeners.

Mosquitto response sequence
  • Inventory instancesIdentify running versions, packages, listeners, and build implementation.
  • Rank exposurePut reachable built-in WebSocket listeners in the affected range first.
  • Patch or mitigateDeploy a fixed build, disable WebSocket support, or use the documented alternative implementation.
  • Verify operationRestart and check the running version, listener availability, and client connectivity.
  • Review healthExamine broker crashes, restarts, and session records for anomalies.

These checks are defensive triage, not a signature for this CVE. A crash or unusual session needs investigation alongside maintenance records and other causes. Preserve relevant logs if an incident review begins, but do not infer compromise from a single service restart.

What Does the Available Evidence Say About Exploitation?

The Eclipse report describes a controlled demonstration on a test broker. It is evidence of technical feasibility, not a confirmed intrusion into a deployed organization. The vendor advisory addresses technical impact and mitigations. Keep those statements separate from the risk posed by a reachable vulnerable listener.

Brandefense Blackhole counted 16 dark-chat posts mentioning Mosquitto between August 20 and November 18, 2026. The unit is a post about the product, not a post necessarily naming this CVE. This period also starts before the November 17 disclosure. The count cannot establish affected brokers, incidents, or active exploitation.

For an operational decision, patch priority follows the verified version and listener conditions, not discussion volume. The vulnerability management workflow can hold unknown implementation cases until the package owner supplies evidence. Recheck that inventory after rollout because replica drift can leave an affected broker running.

What Remains to Verify After the Change?

Ask the service owner to show the running build and the listener configuration from each broker instance. Compare the result with the upstream fixed version or the package maintainer’s backport notice. If WebSocket support was disabled, confirm affected clients moved to an approved path. If libwebsockets was used, retain the build evidence.

Finally, compare broker availability and authentication behavior with the pre-change baseline. The 2.1.4 release note makes password compatibility a concrete check for teams using legacy SHA-512 hashes. Record any rollout exception and its due date. Those steps close the patch decision without claiming that an incident occurred.

Frequently Asked Questions

What should go into the Mosquitto upgrade ticket?

Record each running broker version, WebSocket listener, build implementation, package source, reachability, and authentication setting. Add the chosen fixed build or mitigation and a post-restart verification result.

Which post-upgrade check catches the 2.1.3 regression?

Eclipse says 2.1.3 could reject legacy SHA-512 passwords. Test representative authentication on 2.1.4 or another build that includes the regression fix.

Can dark-chat volume establish Mosquitto exploitation?

No. Brandefense Blackhole counted 16 Mosquitto mentions from August 20 to November 18, 2026. Those are product discussion posts, not verified incidents or CVE-specific attacks.

Sources

  1. Eclipse Mosquitto advisory GHSA-6j8m-r32v-4h38
  2. Eclipse Mosquitto 2.1.4 and 2.0.23 release note
  3. Eclipse vulnerability report 521
  4. Eclipse mosquitto.conf manual
Share this articleLinkedInXEmail
Brandefense logo

Stay Ahead of Digital Threats.

Take control of your digital security with an exclusive demo of our AI-driven threat intelligence platform.

Open The Full Demo Form →