SecurityNews

Patchwork – Saturday 21 November

This edition focuses on software found in the stack of almost every MSP. PHP fixes seventeen vulnerabilities, seven of them critical, and one of them affects WordPress servers without needing a vulnerable plugin. SonicWall is urging customers to replace Gen 6 firewalls immediately, because no more patches will be issued, even for critical flaws. Ruby has also fixed a heap overflow that carries through to GitLab and other self-hosted applications, the NGINX vulnerabilities turn out to affect PAN-OS firewalls too, and there is an unpatched method of bypassing the Windows lock screen.

PHP fixes seventeen vulnerabilities, PHP-FPM flaw leaves WordPress servers vulnerable without a plugin

Vulnerability Seventeen vulnerabilities in PHP 8.2, 8.3, 8.4, 8.5 and the 8.6.0 release candidates, seven of them critical. The most serious: ‘phpAfterShell’ in the default file-based session handler, and request smuggling in PHP-FPM that allows PHP code to be executed as the worker
CVE number CVE-2026-126714 (phpAfterShell), CVE-2026-129289 (PHP-FPM), plus CVE-2026-115560, CVE-2026-129496 and CVE-2026-129481, among others
CVSS score 9.2 (CVSS 4.0) for CVE-2026-126714 and CVE-2026-129289
Actively exploited No, there are no indications of exploitation (as of 20 November 2026)
Patch available Yes, PHP 8.2.36, 8.3.37, 8.4.28, 8.5.13 and 8.6.0, released on 19 November 2026. Distributions may backport fixes without changing the version number

How can the vulnerability be exploited?

According to Cloudflare, phpAfterShell lies in the way PHP writes session files. If a write is interrupted, a session file can end up containing new data followed by old data. On the next request, PHP turns that mixture back into objects, and if an attacker had got hold of the old data, this can lead to code execution. The vulnerability was disclosed on 19 November by depthfirst, which credits Leo Lin and Asahi as the reporters. The PHP-FPM flaw works differently: according to Patchstack, the web server must reuse connections to PHP-FPM, via fastcgi_keep_conn on in NGINX or enablereuse=on in Apache mod_proxy_fcgi, and the application must not read the request body in full. Every request to the WordPress REST API meets that second condition, so on a WordPress server it depends only on the FastCGI setting.

What could the consequences be?

An attacker without an account can execute PHP code on the web server, with the privileges of the PHP process. On shared hosting servers, this affects all sites running in the same pool. A WordPress update does not help here, because the problem lies in PHP itself. BleepingComputer reports that several of the vulnerabilities can lead to code execution.

What should IT administrators do now?

Update PHP to the patched version, rebuild container images and restart PHP-FPM or Apache so that every worker loads the new binary. For distribution packages, check the distribution’s advisory rather than relying on the version number alone. If you cannot patch yet, disable connection reuse to PHP-FPM, check that session.upload_progress.enabled is not set to 1 anywhere, and keep PHP-FPM listeners away from untrusted networks. This reduces the risk, but it does not fix the vulnerabilities.

top ^

SonicWall: replace Gen 6 firewalls immediately, even critical vulnerabilities will no longer be patched

Vulnerability Vendor security advisory: all Gen 6 and Gen 6.5 firewalls running SonicOS 6.x, including the TZ300 to TZ600, SOHO and SOHO 250, NSA/NSa 2600 to 9650, SuperMassive 9200 to 9800, NSsp 12400/12800 and the virtual NSv appliances, are end-of-support and will no longer receive firmware or security updates
CVE number No CVE assigned
Actively exploited No, no new vulnerability has been reported. SonicWall does expect internet-facing Gen 6 firewalls to come under active attack
Patch available No, not even for critical vulnerabilities. Advice: replace with Gen 8 running SonicOS 8 and, until then, block management and SSL VPN access on the WAN side

What is going on?

In a product advisory dated 18 November, SonicWall urges customers to take Gen 6 firewalls out of production immediately and stop exposing them to the internet. According to The Hacker News, SonicWall says attackers are using automation and AI to hunt for unsupported edge devices, and that every Gen 6 firewall still in use should be regarded as a high risk. Support ended at different times depending on the model: for the TZ350 and TZ500 series, the SOHO 250 and the NSA 4600, 5600 and 6600 it was 1 October 2026, and for the TZ400, TZ600 and NSA 3600 it was 1 August 2026.

What could the consequences be?

SonicWall itself points to an earlier SonicOS vulnerability from 2024, for which supported firewalls received a fix and Gen 6 did not. Any further critical flaw in SSL VPN or the management portal will therefore remain open on these devices. Many SME customers are still running a TZ model from this generation, often with SSL VPN enabled for home workers.

What should IT administrators do now?

Take an inventory of all Gen 6 devices at your customers, including spares and branch-office firewalls, and start with the devices that are exposed to the internet or provide remote access. Settings from SonicOS 6.5 can be migrated to Gen 8 using the import function or the NSM Migration App. Until the devices are replaced: disable HTTPS, SSH and SNMP management on the WAN side, disable SSL VPN or restrict it to known addresses, enforce MFA on all management and remote accounts, reset local passwords and monitor the logs for unusual logins and configuration changes.

top ^

Ruby2Shell: heap overflow in Ruby forces updates for GitLab and other self-hosted apps

Vulnerability Heap-based buffer overflow in Ruby’s string case conversion, with single-byte encodings such as ISO-8859-1 and Windows-1252. Carries through to, among others, GitLab CE/EE before 19.4.10, 19.5 before 19.5.8 and 19.6 before 19.6.1, and to GitHub Enterprise Server
CVE number CVE-2026-132825
CVSS score 7.4 according to MITRE; the Ruby security team and GitLab rate it as critical
Actively exploited No, no exploitation has been reported (as of 20 November 2026)
Patch available Yes, Ruby 4.0.10, 3.4.13 and 3.3.14; GitLab 19.6.1, 19.5.8 and 19.4.10 (20 November 2026). GitLab.com and GitLab Dedicated have already been updated

How can the vulnerability be exploited?

According to SecurityWeek, when converting to uppercase Ruby reserves space for one byte, whereas some characters produce two bytes: a ‘ß’ becomes ‘SS’. As a result, Ruby writes one byte past the end of the buffer. According to Veria Labs, an unauthenticated attacker can trigger the flaw in the default configuration via Rails and classic Sinatra applications. watchTowr states that any Ruby application that changes the case of external text to uppercase in a single-byte encoding is vulnerable. The same Ruby release fixes six other memory errors.

What could the consequences be?

In certain scenarios, the flaw can lead to code execution. Yordan Ganchev of watchTowr expects exploitation in production to be difficult: it requires the right memory layout, an allocator other than glibc’s and a way around ASLR. The most likely outcome of an attack is a crash. Veria Labs intends to publish a proof of concept once the major vendors have patched, and Chatwoot, Mastodon, OpenProject and Discourse have already released updates.

What should IT administrators do now?

In its patch release, GitLab advises updating all self-managed installations immediately to 19.6.1, 19.5.8 or 19.4.10. The official Linux packages and container images already contain the patched Ruby; for an installation from source, you must rebuild Ruby with 3.3.14 or 3.4.13. The release contains no new migrations. In addition, check which other Ruby applications you host for customers and update the Ruby runtime on those servers as well.

top ^

NGINX vulnerabilities also affect PAN-OS and Panorama: first hotfixes now available

Vulnerability Use-after-free and integer underflow in the NGINX component of PAN-OS and Panorama (including virtual and M-Series), only if HTTP/2 is enabled. Affected: PAN-OS 12.2.0 up to and including 12.2.4-h*, 12.1.5 up to and including 12.1.10-h* and 11.2.8 up to and including 11.2.16-h*. Older 12.1 and 11.2 versions are not affected
CVE number CVE-2026-123633 and CVE-2026-123007
CVSS score 9.2 (CVSS 4.0); Palo Alto Networks rates the urgency as ‘Moderate’
Actively exploited No, Palo Alto Networks is not aware of any exploitation (20 November 2026)
Patch available Partly. 12.2.4-h3, 12.1.10-h6 and 11.2.16-h3; 12.1.7-h15, 11.2.10-h19 and 11.2.13-h6 are expected on 23 November. Temporary measure: disable HTTP/2

What is new?

The NGINX vulnerabilities featured in the Patchwork of Thursday 19 November. Now Palo Alto Networks reports in an advisory that two of those vulnerabilities also affect PAN-OS firewalls and Panorama. These are the same flaws in HTTP/2 handling, which in NGINX itself can lead to unauthenticated code execution. Palo Alto rates the attack as complex, but automatable and requiring no login or user interaction. Palo Alto patches the SaaS services itself.

What should IT administrators do now?

For every PAN-OS firewall and Panorama instance, check the version and whether HTTP/2 is enabled. If it is, update to 12.2.4-h3, 12.1.10-h6 or 11.2.16-h3, or schedule the hotfixes due on 23 November. If you cannot update yet, disable HTTP/2. Check the version table in the advisory itself, as Palo Alto is still updating it.

top ^

FastUnlock: unpatched method bypasses Windows lock screen via Narrator

Vulnerability Bypass of the Windows 11 lock screen via Narrator in the recovery procedure for a Microsoft account. An attacker with physical access starts programs under the restricted WSIAccount account, without signing in. There is no official list of affected versions
CVE number No CVE assigned
Actively exploited No, no attacks known. There is, however, a public demonstration, and ThreatLocker confirmed on 19 November that the method works
Patch available No, Microsoft has not yet released an update or advisory. Mitigation: BitLocker with a startup PIN

How can the vulnerability be exploited?

The researcher behind the names MSNightmare and Nightmare Eclipse, who has previously published several Windows zero-days, put a short description with a video online on 19 November. ThreatLocker tested the method the same day: via Narrator in the recovery steps for a Microsoft account, you can reach Windows settings from the lock screen and start programs. The attacker needs physical access, and there must be a Microsoft account on the device that is reachable via the recovery options.

What could the consequences be?

The attacker gets a restricted account, not administrator rights. According to The Hacker News, BitLocker in its common configuration, where the TPM unlocks the drive at startup, offers no protection: by the time the lock screen appears, the files are already readable. FastUnlock does not crack the encryption, it merely bypasses the screen in front of it. That makes a lost or stolen laptop a bigger risk than many customers realise.

What should IT administrators do now?

Consider a BitLocker startup PIN for laptops used outside the office, so that the drive does not unlock automatically. A PIN does not protect a device that is already at the lock screen, so advise users to shut down laptops or put them into sleep mode when leaving them unattended. Restrict physical access to shared and publicly accessible PCs and keep an eye on updates from Microsoft.

top ^

This article was originally published in Dutch on MSP Business. Read the original →