SonicWall published SNWLID-2026-0014 on 6 November, and the advisory bundles seven CVEs for SonicOS. The one I keep coming back to is CVE-2026-19868. SonicWall describes an authentication enforcement vulnerability in the SSLVPN service: a user with valid credentials for an alternate authentication method can bypass the SAML authentication controls, and a login that way can also bypass the MFA the SAML configuration was enforcing.1 The advisory lists no workaround. Fixed builds are 7.3.4-7013 for Gen7 platforms and 8.2.3-8015 for Gen8.1

CISA added it to the KEV catalog the same day it was published, with a remediation deadline of 20 November.2 I do not run SonicOS anywhere I own. The configuration mistake behind it can sit on any remote access path I do assess, which is why I am writing about it.

What the advisory says, and where it stops

The scope note is the most useful line in the advisory. CVE-2026-19868 affects the SSLVPN interface only when SAML is enabled for SSLVPN logins.1 That condition splits the exposed population in two, and it explains the bug: the appliance accepts two ways for a user to sign in, and the enforcement covers one of them.

The advisory names no specific alternate authentication method and describes no request sequence. That omission matters when you are deciding whether your configuration is one of the affected ones, because “alternate method” could mean a local account on the box, a directory source, or a second identity path an administrator turned on years ago and never removed. I would rather have the detail, and I am not going to invent it here.

SonicWall’s note on this CVE says its PSIRT investigated a report indicating potential exploitation and urges customers to upgrade as soon as possible.1 That wording leaves the extent open, and the CISA coordinator record that NVD carries lists exploitation as active.3

The bundle covers seven CVEs under one release, and this one is the only one on the KEV catalog. SonicWall rates CVE-2026-107154 at 8.8 for session validation under CWE-291 and CVE-2026-107155 at 8.8 for a permissive cross-domain policy, and the advisory reports no exploitation for either.1

Two scores, one field apart

SonicWall rates CVE-2026-19868 at 7.1, on a vector of AV:N/AC:H/PR:L/UI:N/S:C/C:H/I:L/A:N.1 NVD’s record displays a 7.5 under the CISA-ADP label, and that vector differs in one field: privileges required is none rather than low.3 NVD’s own line in the same table reads Base Score N/A, with the note that no NVD assessment is provided yet, so the 7.5 is the publisher’s number.3 The record I pulled lists its status as Undergoing Analysis, with a last modification date of 7 November, and it carries no CVSS 4.0 score.3

The disagreement is a definitional argument about what the bug requires of an attacker. SonicWall’s 7.1 gives privileges required as low, which treats a credential for another method as a privilege the attacker already needed. The CISA-ADP 7.5 gives none.13 Everything else in the two vectors matches: network reachable, high complexity, scope changed, high confidentiality impact, low integrity impact, no availability impact.

The gap matters in practice. A 7.1 and a 7.5 land in the same band of a severity sort, and both understate the problem on a box that terminates VPN sessions for staff. The KEV listing and the CWE carry more weight than the base score when I set the order, because the catalog attaches a 20 November remediation deadline to this one and a base score attaches nothing.2

CWE-308 puts the flaw in the policy

The classification is CWE-308, use of single-factor authentication.1 That places the flaw in the policy layer. Nothing in the advisory suggests the SAML implementation is broken. The service accepts a sign-in that never touches SAML, and the MFA setting in the configuration applied to the SAML route alone.

That is the pattern worth carrying to other gear. When a gateway can authenticate against more than one source, the MFA policy usually attaches to the path an administrator configured, while the service keeps honoring every method it supports. The control holds for the logins that take the configured path, and the login that skips it rarely appears in the documentation the next administrator inherits.

What I check on the remote access paths I do assess

I start with an inventory of authentication methods per service, taken from the appliance rather than from the policy document. The documented method is the one people meant to use. The method that matters is the one that can complete a login, and those two lists diverge in the environments this bug reaches.

The next question is where MFA is enforced. SAML-enforced MFA is a claim about the SAML route, and it holds only if that route is the only way in. When MFA sits at the identity provider, every other method the appliance accepts sits outside the policy, and the provider’s logs will never show those sessions.

Then I go looking for the break-glass account. The credential that exists so administrators can get in when the identity provider is unavailable is, by construction, an alternate authentication method on the same internet-facing interface. It is the most defensible account to keep and the easiest one to leave reachable from the wrong side.

The logs have to record which method authenticated the session, because a bypass that writes the same line as a normal login leaves nothing to hunt on. Where the field exists, a SAML-enabled deployment showing a local or directory login on the SSLVPN interface is the signal.

The last step is a test with a real credential. Reading the policy tells me what should happen. Logging in with an account that is not supposed to reach the service tells me what does.

The KEV deadline and what I relay

Deployment in my environment is automated, so my contribution to a deadline like this one is the order and the reasoning behind it. Internet-facing SSLVPN with SAML enabled goes first, ahead of anything internal, because the advisory lists no workaround and offers no fix short of the upgrade.1 Where the fixed build cannot land before the 20 November date, the interim move I would argue for is narrowing who can reach the interface at all, which is a smaller change than it sounds and a much smaller one than accepting a bypass on the VPN.2

A severity sort gets the order wrong here. The release is urgent because of the 7.1 on the KEV catalog. A queue built on base scores alone would put the two 8.8s ahead of it, and neither of those carries a KEV deadline or a report of potential exploitation.12

I keep the authentication inventory next to the policy for this reason. A SAML configuration describes the route it was written for, and it cannot describe a login path nobody added to the diagram.


  1. SonicWall. (2026, November 6). SonicWall SonicOS affected by multiple vulnerabilities (SNWLID-2026-0014). SonicWall PSIRT. https://psirt.global.sonicwall.com/vuln-detail/SNWLID-2026-0014 ↩︎ ↩︎ ↩︎ ↩︎ ↩︎ ↩︎ ↩︎ ↩︎ ↩︎ ↩︎

  2. Cybersecurity and Infrastructure Security Agency. (2026, November 6). Known exploited vulnerabilities catalog (catalog version 2026.11.06). https://www.cisa.gov/known-exploited-vulnerabilities-catalog ↩︎ ↩︎ ↩︎ ↩︎

  3. National Institute of Standards and Technology. (2026, November 6). CVE-2026-19868 detail. NIST National Vulnerability Database. https://nvd.nist.gov/vuln/detail/CVE-2026-19868 ↩︎ ↩︎ ↩︎ ↩︎ ↩︎