Cut-paper illustration of a firewall appliance with a guarded front door for SAML and MFA, and a smaller side door standing open with a plain key slipping through

Your firewall asked every VPN user for SAML, and SAML asked for MFA. On paper, that is a solid front door. On November 6, SonicWall admitted the front door had a side entrance: anyone holding valid credentials for a different login method could walk past SAML entirely, and the MFA behind it never ran.

This is CVE-2026-19868, and it is already being exploited. SonicWall's advisory, published November 6, says its PSIRT team "investigated a report indicating potential exploitation" and urges customers to upgrade immediately. The same day, CISA added the flaw to its Known Exploited Vulnerabilities catalog, with a federal remediation deadline of November 20. Canada's Cyber Centre issued its own advisory the same day repeating the exploitation statement.

The uncomfortable part is not the severity number. It is the shape of the bug: the control you enforced was real, and it still did not apply.

What actually broke

CVE-2026-19868 is an authentication-enforcement weakness (CWE-308) in the SonicOS SSL-VPN service. A user with valid credentials for an alternate authentication method can bypass SAML authentication controls. Successful exploitation may also bypass SAML-enforced MFA. SonicWall scores it 7.1 High. The attacker needs low-privilege valid credentials for that alternate method, which is why the score is not higher — but "needs valid credentials" is doing a lot of work here, because the whole point of a VPN is that lots of people hold valid credentials for it.

Two scope notes matter. First, the flaw only affects the SSL-VPN interface when SAML is enabled for SSL-VPN logins. If you never turned SAML on for the VPN, this specific CVE does not apply to you — though the rest of the bundle below still might. Second, there is no workaround. SonicWall lists it plainly: Workaround: false. The patch is the only fix.

No named actor, campaign, victim count, or public proof-of-concept had been reported as of November 6. CISA's KEV entry marks ransomware use as Unknown.

Step 1: inventory every SonicWall against the fixed builds

You cannot patch what you have not counted. Pull your firewall inventory — every site, every branch office, every virtual appliance — and check each one against this table.

Generation Affected Fixed in
Gen7 SonicOS 7.3.3-7015 and earlier 7.3.4-7013 or later
Gen8 SonicOS 8.2.2-8015 and earlier 8.2.3-8015 or later

Check the managed ones and the unmanaged ones. The firewall at the small branch that nobody has logged into in a year counts. So does the NSv virtual appliance someone spun up for a test and never tore down. If your inventory lives in a spreadsheet, now is a good week to find out whether the spreadsheet is lying.

Step 2: find out which SSL-VPN logins actually use SAML

This is the step most people skip, and it is the step this CVE is about. "SAML is enabled" and "every VPN login goes through SAML" are two different claims, and only the second one protects you.

Open the SSL-VPN authentication settings on each firewall and list every authentication method the service accepts: SAML, local users, LDAP, RADIUS, TACACS+, whatever is configured. For each method, note whether it routes through your identity provider or stands on its own. Any method that does not route through the identity provider is a second front door. Write them all down. The list is usually longer than people expect, because alternate methods get added for a migration, a vendor, or a one-off exception, and then nobody removes them.

Step 3: audit the VPN logs for sign-ins that skipped the identity provider

Patching stops future abuse. The logs tell you whether abuse already happened. Look at SSL-VPN sign-ins over the last few weeks and split them by authentication method. Any successful login that did not touch your identity provider deserves a look: who logged in, from where, at what hour, and what they reached afterward.

Pay attention to accounts that should not exist anymore. A service account created for a migration, a vendor account that was supposed to be disabled, a former employee whose local firewall account was never cleaned up — these are exactly the credentials this flaw is built for, because they are valid, forgotten, and not protected by your identity provider's MFA.

If your logs do not record the authentication method per sign-in, that is a finding in itself. Fix the logging before the next incident, not during it.

Step 4: lock down the alternate methods until the patch is in

Patching is the fix, and the patch window should be measured in days, not weeks — CISA's federal deadline is November 20, and this one is already exploited. But patching takes scheduling, maintenance windows, and change approval, and the flaw does not wait for your change board.

While the patch is pending, remove or disable every alternate authentication method on the SSL-VPN that you do not strictly need. If a method must stay, restrict it: limit it to specific users or groups, limit the source addresses that can reach it, and require MFA on it directly rather than relying on SAML's MFA to cover it. Until the firewall is patched, no login should reach the VPN without passing through the identity provider.

The same patch window closes six more flaws

Advisory SNWLID-2026-0014 bundles seven CVEs in one release. The exploited one is the SAML bypass, but the patch you are already scheduling also fixes these:

CVE What it is Score
CVE-2026-107154 Unauthenticated session validation flaw, management interface 8.8
CVE-2026-107155 Permissive cross-domain policy, management interface 8.8
CVE-2026-82308 Pre-auth heap buffer overflow, SSL-VPN (denial of service) 7.5
CVE-2026-77708 Pre-auth stack buffer overflow, TACACS+ client 5.9
CVE-2026-77709 Pre-auth heap buffer overflow, TACACS+ client 5.9
CVE-2026-19703 Weak password hashing, virtual appliances 5.2

The two 8.8s are unauthenticated flaws in the management interface. They are not reported as exploited, but they are the kind of thing that does not stay unexploited for long once the patch is public and the diff is readable. One maintenance window, seven CVEs. That is the rare patch where the scheduling argument makes itself.

The durable point

Most VPN hardening advice is about adding a stronger lock to the front door: turn on SAML, enforce MFA, require phishing-resistant factors. All of that is correct, and none of it mattered here, because the lock was on a door the attacker never had to use.

A second login method is a second front door. "SAML enforced" only means something if no other method can reach the same service. The next time you review a VPN, a web app, or anything with an authentication setting, do not just check what the primary method requires. List every method that can authenticate to it, and ask whether each one is held to the same standard. The answer is usually the audit.

Patch the firewalls. Then go count the doors.

Sources