F5's November 18 security release includes six critical vulnerabilities in BIG-IP, its application delivery and access platform. All six are stack-based buffer overflows in the BIG-IP data plane, the part of the appliance that moves customer traffic. Each carries a CVSS v3.1 score of 9.8 and a CVSS v4.0 score of 9.3. According to the CVE record, an unauthenticated remote attacker who reaches a vulnerable configuration can crash the traffic-handling process, and F5 says remote code execution may be possible.

There are some limits. None of the six bugs affects every BIG-IP. Each one requires a particular configuration: either an iRule that uses a specific command, or an Access Policy Manager (APM) per-request policy that combines two particular agents. The Hacker News reports that none of F5's advisories says the flaws have been exploited. The harder part for defenders is the fix. For these issues F5 has shipped engineering hotfixes rather than scheduled maintenance releases, and it lists no workaround.

The six flaws, and what each one needs

The Hacker News, which reported the release on November 18, lists the six CVEs and the configuration each one depends on:

  • CVE-2026-123637: BIG-IP APM, when a per-request policy includes both the Client Information agent and Endpoint Security agents.
  • CVE-2026-121805: an iRule that uses the PEM::subscriber command on client traffic.
  • CVE-2026-120292: an HTTP profile plus an iRule that uses the MR::message route command.
  • CVE-2026-120285: an HTTP profile plus an iRule that uses the PSC::ip_address command.
  • CVE-2026-116953: an iRule that uses the GENERICMESSAGE::route command on client traffic.
  • CVE-2026-120722: an iRule that uses the DNS::scrape command on client traffic.

Some background on the jargon. An iRule is a script an administrator attaches to a BIG-IP virtual server so the box can inspect or rewrite traffic as it passes through. A virtual server is the listening address and port that clients connect to. In five of these bugs, the trigger is an iRule that calls the named command, and for two of them an HTTP profile must also be present. If none of your iRules uses those commands, those five CVEs do not apply to you. The APM bug is different because the trigger is an access policy, not a script.

According to The Hacker News, Sina Kheirkhah of watchTowr reported four of the six flaws, and F5 found the other two itself.

What "crash" means here

The NVD record for the APM bug, CVE-2026-123637, describes the failure mode. The Hacker News says all six flaws let an attacker crash the same component. When APM is provisioned and a virtual server has a per-request policy containing both the Client Information agent and Endpoint Security agents, "undisclosed requests" can terminate TMM (the Traffic Management Microkernel), the BIG-IP process that handles traffic. Traffic is disrupted while TMM restarts. If an attacker can repeat the request, that could keep an application or VPN gateway disrupted for as long as the requests continue, although the sources do not spell out that scenario.

The bigger risk is the possibility of code execution. A stack-based buffer overflow lets an attacker write past the end of a fixed-size memory buffer on the stack. Depending on the surrounding protections, that can sometimes be turned into running the attacker's own code. F5's language is cautious: remote code execution "may" be possible. The sources do not say it has been demonstrated.

The record also notes two limits. The flaw is in the data plane only, and the record says there is no control plane exposure (the management side used to administer the device). F5 also did not evaluate software versions that have reached End of Technical Support. If you run an older, unsupported branch, the advisories do not tell you whether you are affected, and you should not read their silence as a clean bill of health.

Beyond classic BIG-IP

The Hacker News reports that all five iRule flaws also affect BIG-IP Next for Kubernetes, F5's containerized offering. Fixed builds there are BNK-2.4.2-EHF1, BNK-2.3.4-EHF1 and BNK-2.2.5-EHF1. For BIG-IP CNF (Cloud-Native Network Functions), the outlet says CVE-2026-121805 and CVE-2026-120285 now affect it, with fixes in CNF-2.4.2-EHF1, CNF-2.3.4-EHF1 and CNF-2.2.5-EHF1. It also flags a discrepancy for CVE-2026-120722: F5's advisory lists CNF as not vulnerable, but the CVE record, updated on November 18, lists it as affected. Until F5 reconciles the two, CNF operators should treat that question as open and check with F5 directly.

Hotfixes, not releases

All six fixes ship as engineering hotfixes, and The Hacker News quotes F5 as warning that "engineering hotfixes do not undergo the extensive QA assessment of scheduled software releases." That leaves BIG-IP owners with an uncomfortable trade-off. They can install a less-tested build on a device that often carries production traffic for an entire organization, or they can remain exposed to a critical, unauthenticated, remotely triggerable bug with no listed workaround.

The CVE.org record for CVE-2026-123637, published by F5 as the CVE Numbering Authority, gives the affected APM branches as 21.1.0, 17.5.0 and 17.1.0. It lists these engineering hotfix builds as unaffected:

  • BIGIP-21.1.0.4.0.19.3-ENG
  • BIGIP-17.5.1.11.0.61.12-ENG
  • BIGIP-17.5.1.10.0.80.35-ENG
  • BIGIP-17.1.3.7.0.21.4-ENG
  • BIGIP-17.1.3.6.0.14.20-ENG

The NVD record points to F5 article K000163937 for the APM issue. For the five iRule-related CVEs, take the matching hotfix builds from F5's advisories for each CVE rather than assuming they match the APM list. The Hacker News adds that a system already on a custom engineering hotfix needs a new one from F5's support team.

Not the APM bug from September

These flaws are separate from CVE-2026-94127, the APM vulnerability F5 disclosed on September 22. The Hacker News says it was later exploited and that CISA added it to the Known Exploited Vulnerabilities catalog the same day. If your team patched that one, it does not cover this batch. That flaw is a heap-based buffer overflow in APM setups that serve as OAuth authorization servers, while these six are stack-based overflows that need different settings.

The rest of F5's November release

The six criticals were part of a larger release. The Hacker News counts 22 F5 advisories published the same day, 10 of them for NGINX. They include six lower-rated BIG-IP and BIG-IQ flaws. The first one listed is CVE-2026-125274 in iControl SOAP, rated 8.8. That one has conventional fixes: BIG-IP 21.1.0.3, 17.5.1.10 and 17.1.3.6, and BIG-IQ 8.4.2.3.

What It Means for Defenders

Who is affected: organizations whose BIG-IP configuration matches at least one trigger. For the APM bug, that means a per-request policy combining the Client Information agent and Endpoint Security agents, on the 21.1.0, 17.5.0 or 17.1.0 branches. For the other five, it means an iRule calling PEM::subscriber, MR::message route, PSC::ip_address, GENERICMESSAGE::route or DNS::scrape (with an HTTP profile for the MR and PSC commands); check each CVE's F5 advisory for affected branches and modules. BIG-IP Next for Kubernetes deployments are also affected, and CNF status is disputed for CVE-2026-120722 between F5's advisory and the CVE record.

What to check first:

  • List every iRule on every virtual server and search them for the five commands above. Include iRules attached to rarely used or legacy virtual servers.
  • If APM is provisioned, review your per-request policies for the Client Information plus Endpoint Security combination.
  • Confirm which software branch each device runs. Anything past End of Technical Support was not evaluated, so plan to upgrade it whatever these results show.

What to patch: F5 lists no workaround; the fix is the engineering hotfix from the relevant F5 advisory. For the APM flaw, the unaffected builds are BIGIP-21.1.0.4.0.19.3-ENG, BIGIP-17.5.1.11.0.61.12-ENG, BIGIP-17.5.1.10.0.80.35-ENG, BIGIP-17.1.3.7.0.21.4-ENG and BIGIP-17.1.3.6.0.14.20-ENG (see F5 article K000163937). For BIG-IP Next for Kubernetes, move to BNK-2.4.2-EHF1, BNK-2.3.4-EHF1 or BNK-2.2.5-EHF1 as appropriate. Because F5 says hotfixes get less QA, test on a standby unit of an HA pair or in a staging environment first if you can, and have a rollback plan ready.

Can you remove the trigger instead? This is our inference, not F5 guidance: since each flaw depends on a specific configuration, removing or rewriting an iRule that uses an affected command, or an APM per-request policy whose agent combination is not essential, should take that device out of scope for the matching CVE until you can schedule the hotfix. Confirm with F5 before relying on it, and make the change through normal change control, because iRules and access policies often carry logic that only becomes visible once it breaks.

Watch for the symptom: Because the observable effect is TMM terminating and restarting, unexpected TMM restarts or traffic interruptions on devices with matching configurations deserve investigation and should not be dismissed as routine instability.

Do not stop at the criticals: The same release fixed CVE-2026-125274 in iControl SOAP (8.8). Update to BIG-IP 21.1.0.3, 17.5.1.10 or 17.1.3.6, or BIG-IQ 8.4.2.3. If you run NGINX, review the ten NGINX advisories from the same day as well.

For CMMC and federal contractors: BIG-IP and APM can sit at the boundary of environments that handle controlled unclassified information. Record which devices you checked against each trigger, the result, and the hotfix or configuration change you applied. If a hotfix is deferred for QA reasons, document the risk acceptance and the compensating steps you took.

Sources