9.3
CVSS score of the ESG authorization bypass, CVE-2026-116606
6
Barracuda product lines named for the command injection
17
exploitation attempts one firm saw from 9 to 11 Nov 2026
7
days from the advisory to the exploitation notice
The mail gateway is a server on your edge
An email security gateway is sold as a filter, and that is how most teams think of it. In practice it is a Linux server that sits on your internet edge, receives traffic from anyone who can send you mail, holds your mail routing and policy, and is trusted by everything behind it to have done its job. When an attacker owns it, they own a box your network was told to believe.
That is the lens for the two Barracuda flaws disclosed on 2 Nov 2026. Chained, CVE-2026-116606 and CVE-2026-111560 give an attacker root on the Barracuda Email Security Gateway (ESG). On 9 Nov Barracuda said it had seen exploitation affecting a small number of organizations, and on 10 Nov CISA added both to its Known Exploited Vulnerabilities catalog. The patches exist. Most appliances will have taken them automatically.
The part that decides who gets hurt is not the firmware version. It is one sentence in the advisory: exploitation requires the attacker to have access to the administrative interface. An appliance whose admin page answers only from an internal management network is a much harder target. One whose admin page answers from the internet is the target. On external tests we regularly find management interfaces of edge appliances reachable from networks they were never meant to serve, and this advisory is what that habit costs.
So this post makes one argument. Patch, yes. Then spend the rest of the week on placement, and on the honest question of whether any appliance that was reachable before the patch is still one you should trust.
What Barracuda disclosed and who is affected
Barracuda's advisory, published on 2 Nov 2026, describes two vulnerabilities in its own words: "a pre-authentication authorization bypass vulnerability and a command injection vulnerability." The first, CVE-2026-116606, affects only the ESG and carries a CVSS score of 9.3. The second, CVE-2026-111560, carries a CVSS score of 8.6 and is listed for six product lines: the ESG, the Load Balancer ADC, the Message Archiver, the WAF Control Center, the Web Application Firewall and the Web Security Gateway. Barracuda labels both High on its own scale. On the site's scale a 9.3 sits in critical territory, and we treat it that way.
Barracuda states the combined outcome plainly: "If exploited in combination, these vulnerabilities could allow an attacker to bypass authorization controls and obtain root access to the device." The advisory names no component for either flaw. SecurityWeek reports that an attacker with access to the Admin UI could make the appliance restore a crafted backup archive and modify its configuration, which is the clearest public description of the mechanism so far.
The advisory changed twice after publication, and both changes matter.
On 5 Nov Barracuda changed the authentication field for CVE-2026-116606 from "Yes, administrative access" to "No (Administrative UI Access Required)", and spelled it out: authentication is not required, but the attacker needs the Admin UI to be reachable. That turned a bug that needed a stolen admin password into one that needs only a route to a login page.
On 9 Nov Barracuda added: "We have identified exploitation activity affecting a small number of organizations." Seven days separate the advisory from that notice. Barracuda has not said who was hit, what the attacker did after root, or who is behind it, and it has published no indicators of compromise. SecurityWeek notes that neither Barracuda nor CISA has shared details of the exploitation.
-
Products //
ESG (both flaws); ADC, BMA, WCC, WAF, WSG (CVE-2026-111560 only) -
Flaws //
Pre-authentication authorization bypass; command injection -
CVSS //
9.3 and 8.6, per Barracuda -
Precondition //
Network access to the administrative interface -
ESG patches //
BNSF-40425 on 9.4.017 and later, BNSF-40430 before 9.4.017 -
Exploited //
Yes, per Barracuda and CISA KEV -
Indicators //
None published
There is one disagreement in the reporting worth knowing about. Barracuda's table still lists CVE-2026-111560 as requiring administrative access. BleepingComputer describes the second flaw as letting unauthenticated attackers inject OS commands that run as root. Read together, the safest interpretation is the vendor's chain: the bypass removes the need for credentials, and the injection then runs with the access the bypass provided. Either way the answer for a defender is the same, because the chain is what was exploited.
The patch table names a fix for every listed product: BNSF-40425 or BNSF-40430 for the ESG depending on firmware branch, BNADC-16586 for the Load Balancer ADC and the Web Application Firewall, BNMA-32468 for the Message Archiver, BNWF-60060 for the WAF Control Center and BNYF-24789 for the Web Security Gateway. Barracuda's wording on delivery is conditional: "If a patch has not been automatically applied, customers should manually apply the appropriate patch as soon as possible." It also warns that a device several firmware versions behind may need more than one update before the fix lands.
CISA added both CVEs to the KEV catalog on 10 Nov 2026 "based on evidence of active exploitation." SecurityWeek reports that federal agencies were given three days to patch under BOD 26-04. The same report cites the security firm Previdian, which saw 17 attempts against CVE-2026-116606 between 9 and 11 Nov from 15 IP addresses in three countries. Those are attempts, not confirmed compromises, but they tell you scanning for the bug started as soon as the exploitation notice went out.
How the Admin UI becomes root
The public detail is thin by design, and we keep it that way. Here is the route at the level of what each stage gives the attacker, built only from what the vendor and the two outlets above have said.
Figure 01
From a login page to root
- 01
Reach //
The Admin UI answers the attacker's network.
- 02
Bypass //
Admin actions run with no credentials.
- 03
Inject //
A command injection runs as root.
- 04
Own //
The gateway that inspects your mail is theirs.
Result //
One reachable login page decides the outcome.
Two points are worth drawing out.
First, the bypass works on a function that exists for good reasons. Restoring a configuration backup is what an administrator does after a hardware swap or a bad change. The reported abuse is making the appliance accept that function from someone who never logged in. Bugs in restore and import paths keep turning up in management planes because those paths are built to accept large, structured, trusted input and then act on it with high privilege. We wrote about the same pattern a few days ago in the Check Point gateway chain, where a management service handed root to anyone who could reach it.
Second, root on a mail gateway is a position, not an endpoint. The appliance sees inbound mail before your users do. It holds mail routing, relay settings, policy, and whatever credentials it uses to talk to your directory and your mail servers. Barracuda has not said what attackers did after root this time, and we will not guess at it. What we can say is what the position offers to anyone who holds it, which is why the response in the next section is more than a patch check.
Why a patched ESG is not proof of a clean one
Barracuda ESG owners have been here before, and the last time is the best guide to this one.
In 2023 a flaw in the same product, CVE-2023-2868, was exploited in an espionage campaign that BleepingComputer, citing Mandiant, attributes to the group tracked as UNC4841. Barracuda's own notice records the earliest evidence of exploitation in October 2022, the flaw identified on 19 May 2023, and a patch applied to all ESG appliances worldwide on 20 May 2023. Less than three weeks later the guidance changed. In June 2023 Barracuda wrote that "Compromised ESG appliances must be immediately replaced regardless of patch version level." By the vendor's own timeline, attackers had been on some boxes for months before anyone knew, and the patch closed the door without removing what had come in through it. BleepingComputer, citing Mandiant, reports that almost a third of the appliances hacked in that campaign belonged to government agencies.
Nothing in the 2026 advisory says this campaign looks like that one. There is no attribution, no indicator set and no replacement guidance. The lesson we take is narrower and holds anyway: on this appliance, the patch answers whether the bug can be used from now on. It does not answer whether it was used already.
Three facts in the current advisory make that question harder than it should be.
Patches arrive automatically only if automatic patching is on. Barracuda's advice to customers who turned it off is to turn it back on. An appliance with updates disabled is exactly the one that was most likely to still be vulnerable on 9 Nov, and the one least likely to have anyone watching it.
Firmware that is several versions behind may need more than one update before the patch applies. That is a familiar trap: an administrator applies an update, sees success, and does not notice the box is still one step short.
And there are no published indicators. A team cannot hunt for a known bad file or a known bad address list from the vendor, because there is not one yet. The 15 source addresses Previdian saw are a lead for your own logs, not a blocklist anyone has published to us.
Field Effect's advice for an appliance you suspect is to review admin access logs and configuration restore activity, and to replace credentials and keys the appliance could reach. We agree, and we would put the decision in a simple form. If the Admin UI of an ESG answered from the internet at any time between 2 Nov and the moment you confirmed the patch, that appliance is in the review pile, not the done pile.
How to detect exposure and exploitation
Detection here splits into two jobs: proving the precondition is gone, and looking for signs it was used while it was present.
The first job is the higher return. Every listed Barracuda appliance needs one answer: does its administrative interface accept connections from anywhere outside your management network? Check it from the outside, not from the firewall rule base, because rules drift and NAT hides things. Our external recon walkthrough and the exposed admin endpoint checks are the method we use to answer that question for a whole domain rather than one appliance. If you are not sure which appliance fronts your mail, the DNS records tool shows where your MX records point.
The second job uses the logs you already have at the network edge. A firewall or flow log that records connections to the appliance's admin ports is enough to ask the question the advisory implies: who reached this page who should not have? The rule below is a starting point in Sigma. It flags any connection to the admin interface of a listed appliance from a source outside your trusted management ranges. Replace the placeholder addresses and ports with your own appliances and the ports your admin interfaces actually use.
Figure 02
Barracuda admin interface reached from outside the management network
Sigma
title: Barracuda appliance admin interface reached from an untrusted source
status: experimental
description: Connections to the admin interface of Barracuda ESG, ADC, BMA, WCC, WAF or WSG appliances from outside trusted management ranges. Related to CVE-2026-116606 and CVE-2026-111560.
logsource:
category: firewall
detection:
appliance:
dst_ip:
- 203.0.113.10
- 203.0.113.11
admin_port:
dst_port:
- 8443
trusted:
src_ip|cidr:
- 10.20.0.0/24
condition: appliance and admin_port and not trusted
falsepositives:
- Administrators connecting from a new jump host or VPN range
level: high Run it back over the period from 2 Nov 2026 to today, not only forward. Any hit from an address you cannot explain moves that appliance into the review described in the previous section. Compare any external sources against the attempts reported publicly, and keep in mind that an absence of matches proves little if the log did not record the admin port at all.
On the appliance itself, the reported mechanism points at configuration restore. A restore that no administrator remembers, or a configuration change with no ticket behind it, during the exposure window is the event to chase. We have not seen Barracuda document the exact log entry a malicious restore leaves, so treat this as a question for your administrators and for Barracuda support rather than a field to search on blind.
How we test a security appliance on an engagement
On a sanctioned external test, security appliances get the same treatment as any other internet facing server, because that is what they are. The appliance's job title earns it no trust.
We start by finding every management surface the client owns, not just the ones on the asset list. Mail gateways, web application firewalls and load balancers are frequently managed by a different team from the servers behind them, and their admin pages are often reachable because someone needed to work on them from home once. The question for each one is the same: who can reach this login page, and should they?
Then we look at what the appliance trusts and what trusts it. A mail gateway that relays to internal mail servers, binds to the directory, or sends logs to a central collector holds credentials and network paths that matter more than the appliance itself. On an assumed breach test we ask what an attacker with root on that box could reach next, and we show the client that path without walking it in production.
We do not exploit vendor appliance bugs on a client's production edge to prove a point. A confirmed precondition, a version and patch check, and a clear statement of what root on that box would expose is enough for a client to act, and it carries none of the risk of breaking the mail flow. Where a client wants proof, we agree it in the rules of engagement and use a lab or a maintenance window.
The finding we write up most often is not the CVE. It is the reachable admin interface, because it outlives every patch cycle and it is the reason the next appliance advisory will matter to that client too. The same theme ran through the SonicWall SSL-VPN and Citrix NetScaler advisories this month: the edge appliance is a server, and its management plane is the soft spot.
What is still coming
Several questions are open, and each will change what owners need to do.
Barracuda has not published indicators, attribution or a description of what the attackers did after root. If the vendor follows its own 2023 pattern, an update with indicators or a change in guidance would come as an amendment to the advisory, so watch that page rather than the news.
The authentication field for CVE-2026-111560 still says administrative access is required, while one outlet reports otherwise. If Barracuda revises it the way it revised CVE-2026-116606 on 5 Nov, the five other product lines move from "needs a bypass first" to "exposed on their own", and the patch urgency for the Load Balancer ADC, the Message Archiver, the WAF Control Center, the Web Application Firewall and the Web Security Gateway goes up with it.
The CVSS version behind the 9.3 and 8.6 scores is not stated on the advisory page we read. We have quoted the scores as Barracuda gives them and will not convert them.
And scanning will continue. A bug with a public exploitation notice, a short list of products and a login page as its only precondition is the kind of thing automated campaigns pick up. The 17 attempts in three days are early numbers.
What to do today
-
Take every Barracuda admin interface off the internet. Restrict it to a trusted management network, which is Barracuda's own recommendation, and confirm it from outside, not from the rule base.
-
Confirm the patch on every listed appliance. Check the patch ID for the ESG (BNSF-40425 or BNSF-40430) and for the ADC, BMA, WCC, WAF and WSG, and turn automatic patching back on wherever it is off.
-
Watch for appliances several firmware versions behind. They may need more than one update before the fix applies, so verify the result rather than the action.
-
Review any appliance that was reachable before it was patched. Admin access logs, configuration restores and unexplained changes since 2 Nov 2026; if anything is unexplained, rotate the credentials and keys it holds and talk to Barracuda support before you trust it.
-
Run the Sigma starting point above back to 2 Nov. Connections to admin ports from outside your management ranges are the event this chain depends on.
-
Add appliance management planes to your external test scope. The next advisory will have the same precondition, and the answer should already be known.
SOURCES
- Barracuda advisory // both CVEs, CVSS scores, affected products, patch IDs, the Admin UI precondition and exploitation against a small number of organizations
- CISA // adds CVE-2026-116606 and CVE-2026-111560 to the KEV catalog on 10 Nov 2026 on evidence of active exploitation
- SecurityWeek // crafted backup archive restore through the Admin UI, and 17 attempts from 15 IP addresses seen by Previdian
- BleepingComputer // independent report of exploitation to root and the 2023 ESG campaign
- Barracuda ESG 2023 notice // the 2023 guidance to replace compromised appliances regardless of patch level