The Root Brief #1: SSHole, a patched-but-not-fixed Zyxel and the RMM that pulled its own plug
Eleven new entries in CISA's exploited catalogue in eight days, an OpenSSH bug exploited within hours of disclosure, a Zyxel zero-day hiding behind a June patch, and 33 public exploits that forced an RMM vendor to take its own cloud offline. What to patch first, and why.
Week of 6–12 November 2026. Triage, research and one tool, in about ten minutes.
This was the week the patch itself became the attack surface. A WordPress race condition that only exists while you update. A Zyxel switch that was "patched" in June and owned anyway. An OpenSSH bug exploited within hours of a fix, which means attackers diffed the commit before most of us had read the advisory. And a remote management vendor that responded to a public exploit dump the only way it could: by switching its cloud off.
Between 3 and 10 November, CISA added eleven vulnerabilities to its Known Exploited Vulnerabilities (KEV) catalogue, including the Outlook zero-day that arrived with Patch Tuesday. Here is how we'd spend your next patch window. (New here? Issue #0 explains our labels.)
The triage table
| Priority | Vulnerability | Why |
|---|---|---|
| PATCH NOW | Barracuda ESG chain, CVE-2026-116606 + CVE-2026-111560 | Pre-auth to root on an email gateway; exploited, in KEV, public PoC |
| PATCH NOW | Chrome / Chromium XML use-after-free, CVE-2026-125350 | Exploited in the wild; every Chromium browser in your fleet |
| PATCH NOW | Outlook NTLM leak, CVE-2026-124194 | Exploited zero-day; credential capture from a crafted file |
| PATCH NOW | Zyxel GS1900, CVE-2026-107441 | 996 switches already compromised; June firmware does not fix it |
| PATCH NOW | GPUStack auth bypass, CVE-2026-102895 | CVSS 9.8, unauthenticated admin on your AI inference cluster |
| PATCH NOW | FreeSWITCH, CVE-2026-116779 | CVSS 9.8 code injection via a SIP user field; exploited |
| PATCH NOW | N-central (Able2Shell, no CVEs yet) | 33 flaws with public exploit code; follow N-able's shutdown guidance |
| THIS WEEK | OpenSSH "SSHole", CVE-2026-123158 | Exploited, but only on musl systems without PAM. Alpine, look up |
| THIS WEEK | IBM Aspera Faspex, CVE-2026-123218 | Authorisation bypass to admin, 5.0.0–5.0.19 |
| THIS WEEK | Cisco BroadWorks, CVE-2026-20320 | Unauthenticated XXE in OCI-P; in KEV, fix has existed since RI.2026.07 |
| SCHEDULED | WordPress core, CVE-2026-119977 | In KEV, but the race is hard to win; auto-updates have mostly closed it |
| SCHEDULED | Openwave Webtop Transcoder, CVE-2026-111793 | In KEV, but authenticated and CVSS 4.3 |
The stories behind the table
SSHole: a bug with a very specific address
OpenSSH vulnerabilities make headlines because sshd is on everything. CVE-2026-123158, nicknamed SSHole, is the rare case where the blast radius is narrow, and the exploitation curve is anything but. OpenSSH shipped 10.6p2 on 5 November; exploitation attempts were seen from 6 November, and CISA added it to KEV on 9 November.
The conditions are precise. The bug lives in a password-authentication code path on musl-libc Linux builds without PAM: in practice, Alpine. Exploitation needs password logins enabled, an account whose password field is *, and at least one local account with a $5$ or $6$ hash. Debian, Ubuntu and RHEL (glibc) aren't named as affected, nor are Dropbear or OpenBSD.
Why it still matters: Alpine is the default base image for a huge share of containers, appliances and "tiny" jump boxes, and those are exactly the systems nobody remembers to patch. Find every Alpine host or container that runs sshd and accepts passwords, upgrade to Alpine's patched openssh package (10.6_p2 on edge, backports on 3.21–3.24), restart sshd, and turn off PasswordAuthentication while you're there. If you can't patch tonight, disabling password logins removes the exploitable path entirely.
Zyxel: when "patched" is a lie your scanner tells you
This is the one to study. GreyNoise counted 996 compromised GS1900 switches in 48 countries, with activity going back to mid-August. The attacker took device configurations, hashed root passwords and network topology. Early reporting blamed CVE-2026-7273, fixed in June. Then Binarly showed that the exploit was hitting a different, unknown bug, CVE-2026-107441: an out-of-bounds write in the web management interface that leads to unauthenticated command execution and works on switches running the June firmware.
The lesson is uncomfortable: your vulnerability scanner would have marked those switches compliant. Version-based detection can't see a bug nobody has named yet. Two practical takeaways:
- Update to firmware 2.90 Patch 3 on all ten affected GS1900 models. Patch 2 and earlier are vulnerable.
- A management interface reachable from user VLANs is a finding by itself. 564 of the compromised switches still had factory default credentials. Put switch management on its own out-of-band network and you remove the attack surface whatever the next CVE turns out to be.
Barracuda ESG: config restore as an attack primitive
Barracuda's Email Security Gateway is back in the headlines with a two-bug chain that watchTowr took apart. CVE-2026-116606 (Mandiant rates it 9.3) lets an unauthenticated attacker restore a crafted configuration backup and rewrite the appliance's global settings. CVE-2026-111560 is an OS command injection through the timezone field that runs as root. Chain them and you're root on the box that reads all of your organisation's mail, without logging in.
Barracuda confirmed exploitation against "a small number of organisations" on 9 November, and both CVEs went into KEV on 10 November. Firmware 9.4.0.030 or later fixes the bypass. The command injection also affects ADC, BMA, WAF, WSG and WCC, but there it needs admin access first. If your ESG was reachable and unpatched at any point this month, patching isn't enough: review the appliance configuration for changes you didn't make, and treat it as a potential incident.
N-central and Able2Shell: the RMM kill switch
On 8–9 November a researcher published Able2Shell: a write-up and 26 Python proof-of-concept scripts covering 33 vulnerabilities in N-able N-central, grouped into four chains, three of which end in remote code execution. N-able took its hosted platform offline, patched and upgraded the servers it runs for customers, and told on-premises customers to shut their servers down until it confirms a fix. No CVE IDs had been assigned at the time of writing, and N-able reports scanning activity using the published code but no confirmed exploitation.
Think about what an RMM is: an authenticated, trusted channel that can run code on every endpoint a managed service provider looks after. Compromise it and you don't own one company, you own all of its customers. That's why ransomware crews have targeted N-central before; CISA already lists five earlier N-central flaws as exploited. If you're an MSP customer, ask your provider today which RMM they use and whether it's offline or patched. If you run N-central yourself, follow N-able's guidance and hunt the server for unfamiliar processes, new admin accounts and outbound connections from the period it was exposed.
Chrome's eighth zero-day of the year
CVE-2026-125350 is a high-severity use-after-free in Chrome's XML handling, reported on 6 November and patched on 9 November in Chrome 157.0.8103.50/.51. A crafted page gets code execution inside the renderer sandbox; real-world chains pair a bug like this with a sandbox escape. Google hasn't said who's using it or against whom, which usually means targeted operations rather than mass exploitation.
The operational trap is the relaunch: Chrome downloads the update and keeps running the vulnerable build until the browser restarts. Use your management tooling to force a relaunch deadline, and remember that Edge, Brave, Opera and every Electron or CEF-based desktop app embed Chromium on their own release schedules.
Outlook: the NTLM leak that keeps coming back
Microsoft's November Patch Tuesday was enormous: more than a thousand CVEs (counts range from 1,092 to 1,235 depending on how you count), second only to October's record release. Exactly one was exploited before the patch: CVE-2026-124194, an Outlook spoofing bug that's really a credential leak. Open a crafted file and Outlook quietly connects to an attacker-controlled location, handing over the user's NTLMv2 hash for offline cracking or relay.
The CVSS score is a modest 6.5, which is why it's easy to underrate. Don't. Patch it, and use it as the push to block outbound SMB (TCP 445) at the perimeter, add privileged users to the Protected Users group, and keep moving towards turning NTLM off. Every NTLM leak bug becomes harmless once the hash has nowhere to go.
The rest of the KEV list, quickly
- GPUStack (CVE-2026-102895): an authentication bypass to admin, CVSS 9.8, exploited against a 2.1.1 deployment. AI infrastructure is getting the same treatment web apps got fifteen years ago: deployed fast, exposed to the internet, and attacked straight away. Upgrade to 2.2.0 and put a login in front of the management plane.
- FreeSWITCH (CVE-2026-116779): the SIP handler splits the user field on
+and lets callers reach internal API commands. CVSS 9.8, fixed in 1.11.3. If you run VoIP, it's internet-facing by design. - IBM Aspera Faspex (CVE-2026-123218): crafted messages bypass authorisation to admin on 5.0.0–5.0.19. Fixed in 5.0.19 PL1. File-transfer appliances are a favourite extortion target.
- Cisco BroadWorks (CVE-2026-20320): unauthenticated XXE through the OCI-P service, fixed in RI.2026.07 and reportedly exploited in a broad campaign. A KEV entry for an old fix tells you how many providers haven't applied it.
- WordPress core (CVE-2026-119977): a time-of-check/time-of-use race in full-package core updates, 4.7 through 7.1.3, fixed in 7.1.4 with backports. Wordfence saw attackers probing staged installer paths under
wp-content/upgrade. Confirm you're on the fixed release for your branch and look for unfamiliar PHP files there.
Research worth your time
- Binarly / CSO: the Zyxel zero-day hiding behind an old fix: a textbook case of exploitation being misattributed to a known CVE, and why "patched" isn't the same as "safe".
- The Hacker News: the Barracuda ESG pre-auth chain: how a configuration-restore feature turns into an unauthenticated write primitive. Worth reading for anyone who builds backup and restore features.
- Cloud Security Alliance: Able2Shell and the RMM kill switch: a clear look at what happens when a supply-chain tool becomes the target, and at what MSP customers should ask their providers.
- CSA: when patching is the attack window: the WordPress update race, and a useful reminder that TOCTOU bugs hide in maintenance code nobody threat-models.
- Help Net Security: SSHole exploited hours after disclosure: the timeline from fix to exploitation, compressed to hours.
One tool: ssh-audit
Given SSHole, this week's tool is ssh-audit: an open-source scanner that connects to an SSH server (or client) and reports the software version, key exchange, host key and cipher algorithms, with clear advice on what to remove. Run it across your estate to find the forgotten sshd instances, the ones still allowing password authentication, and the weak algorithms that have survived three OS upgrades. It's fast, it doesn't need credentials, and its output reads well in a report. As always, only scan systems you own or are authorised to test.
The pattern this week
Four of this week's stories share a thread: the trusted maintenance path was the weak point. A switch's management interface, an email gateway's config restore, an RMM's remote control, a CMS's own updater. These features exist to have power over the system, and they get the least scrutiny in design reviews and pentests. If you're scoping an assessment this quarter, put the admin and maintenance planes in scope explicitly.
That's issue #1. If it saved you an hour, forward it to whoever owns your patch queue, and tell us in the community what we got wrong. We'll be back next week.
Sources: CISA KEV additions, 3 Nov, 9 Nov and 10 Nov; BleepingComputer on N-central; BleepingComputer on Zyxel; BleepingComputer on Chrome; Tenable on Patch Tuesday; SecurityWeek on SSHole; SecurityWeek on BroadWorks and Openwave. Details were correct at the time of writing; check vendor advisories for the exact fixed version on your platform.
Related research
The Root Brief #0: How we triage vulnerabilities
Before the first weekly issue, here is exactly how we decide what goes in, how we prioritise it, and what each label means.