6.5
CVSS 3.1 base score, rated Moderate by Microsoft
1
flaw marked exploited among the 1,094 CVEs in the November release
24 Nov
CISA KEV due date for federal agencies, 14 days after listing
445
the outbound TCP port Microsoft says to block to the internet
A Moderate rating on a credential theft bug
Microsoft's November release formally lists 1,094 CVEs. One of them was being used in attacks before the patch existed, and it is not among the fixes rated Critical. It is CVE-2026-124194, an Outlook flaw Microsoft rates Moderate with a CVSS score of 6.5. A user opens a crafted message file, Outlook connects to a server the attacker controls without showing the usual external content warning, and Windows sends that server the user's NTLMv2 hash. CISA added it to the Known Exploited Vulnerabilities catalog on 10 Nov 2026, the day the fix shipped.
Our argument is about the gap between that score and that listing. The 6.5 rates what the bug does on its own: it discloses one value, after one click, and changes nothing on the machine. What the value is worth is decided somewhere else, by the network around the user. If a workstation can open an SMB connection to any address on the internet, the hash leaves the building. If servers inside still accept NTLM without signing, or the user's password is a season and a year, the hash becomes a login. If neither is true, the attacker has spent a zero-day to learn nothing.
So the same file is a nuisance in one estate and a domain account in the next, and the score cannot tell you which estate you run. Attackers picked this bug out of more than a thousand because credential theft does not need a high score to pay. A triage queue sorted by CVSS puts it on page three. We would put it first, and then spend the afternoon on the two settings that decide whether the next bug of this kind matters at all.
What is CVE-2026-124194 and who is affected
We keep to the advisory's words. Microsoft describes CVE-2026-124194 as a spoofing vulnerability: the product interface does not warn the user of unsafe actions in Microsoft Office Outlook, which allows an unauthorized attacker to perform spoofing over a network. An unauthenticated attacker sends a specially crafted message file and convinces a user to open it in an affected version of Outlook. Opening the file could cause Outlook to connect to an attacker-controlled network location without displaying the usual external content warning, potentially disclosing authentication information that could be used to impersonate the user. Microsoft's answer to the confidentiality question is direct: the vulnerability discloses a user's NTLMv2 hash to the attacker, who could use it to authenticate as the user.
Three details matter for triage. The Preview Pane is not an attack vector, so reading the email is not enough; the user has to open the attached file. No authentication is needed on the attacker's side. And Microsoft's exploitability assessment reads exploitation detected, with the flaw not publicly disclosed before the fix. CyberScoop, counting the release, reports that only one flaw is under active attack and none were publicly known beforehand. That one flaw is this one.
-
Affected //
Microsoft 365 Apps for Enterprise, Office LTSC 2021 and Office LTSC 2024, 32-bit and 64-bit -
Rated //
Moderate, CVSS 6.5 (3.1, Microsoft) -
Needs //
no credentials; the user must open a crafted message file; the Preview Pane is not a vector -
Discloses //
the user's NTLMv2 hash, to a server the attacker controls -
Fixed in //
Office LTSC 2021 build 16.0.14334.21048, Office LTSC 2024 build 16.0.17932.21140, and the 10 Nov 2026 security release for each Microsoft 365 Apps channel -
Exploited //
yes, per Microsoft and CISA KEV; listed 10 Nov 2026, due 24 Nov 2026 -
Public before the fix //
no, per Microsoft
The KEV entry gives federal agencies until 24 Nov 2026, 14 days from listing, and records ransomware use as unknown. Neither Microsoft nor CISA has said who is exploiting it, against whom, or how many victims there are, and we will not guess. The advisory lists no older Office versions. Microsoft assesses the products it still supports, so an older Outlook missing from the list has not been cleared; it has not been looked at.
How one opened file becomes a domain login
The chain below is built from the advisory and from how this class of leak is used once the hash is out. It describes what each stage achieves, with no exploit detail.
Figure 01
From one opened file to a working identity
- 01
Deliver //
A crafted message file reaches a user, attached to an ordinary looking email.
- 02
Open //
The user opens it. Outlook connects to the attacker's server and shows no external content warning.
- 03
Capture //
Windows authenticates to that server on the user's behalf. The attacker records the NTLMv2 hash.
- 04
Use //
The attacker cracks the hash offline to recover the password, or relays the live authentication to a service that accepts NTLM.
- 05
Arrive //
The attacker signs in as the user wherever that password or that relayed session is honoured.
Result //
One opened file became a working identity, and no malware ran on the laptop.
Stage three deserves a plain explanation, because it is why this class keeps returning. When a Windows program reaches for a file on a remote server, Windows offers to prove who the user is, automatically, so that file shares work without a password prompt. Inside a company that is a convenience. Pointed at a server on the internet it is a leak: the server receives a cryptographic response computed from the user's password. Outlook's external content warning exists to put a human decision in front of that reach. This bug removes the decision.
Stage four splits in two, and the split decides who should worry most. Cracking is an offline contest between the attacker's hardware and the user's password. An NTLMv2 response cannot be replayed later, but every guess can be checked against it at leisure, and a short or predictable password falls quickly. Our password hash strength audit guide shows how to measure that exposure for your own accounts without ever displaying a password, and the Hash Identifier will tell you what a captured line is and how fast it can be attacked.
Relaying is the live option: the attacker passes the user's authentication straight through to another service while it is still fresh. From the internet that needs a service of yours that is reachable from outside and still accepts NTLM. From a foothold inside, which is where we usually stand on an internal test, it needs any server that does not require signing. Microsoft's SMB documentation says signing confirms the sender's and receiver's identities and prevents relay attacks. Where signing is required, this route is closed.
Then stage five. A recovered password is tried against the VPN, the mail system and every single sign-on page the organisation exposes. If a second factor stands in front of all of them, the attacker has a password and a problem. If one forgotten portal takes a password alone, they are in. We covered how attackers go around the second factor instead of through it in our MFA piece, and it applies here without changes.
What stops a leaked NTLM hash from working
The patch closes this particular way of making Outlook reach out. It does not change what happens the next time any Windows program is persuaded to authenticate to a stranger, and there will be a next time. The controls below are the ones that make the whole class cheap to survive, and each comes from Microsoft's own guidance.
Start at the edge. Microsoft's guidance on securing SMB traffic says to block TCP port 445 outbound to the internet at the corporate firewall, and adds that it is unlikely you need to allow it at all unless a public cloud offering requires it, naming Azure Files and Office 365 as the main scenarios. The same page makes the point most estates miss: laptops that leave the building are not behind that firewall. It calls outbound rules for guest and public networks important to enforce on mobile devices and home workers, because enforcing them on laptops reduces the odds of phishing attacks that send users to malicious servers to harvest credentials. A perimeter rule protects the office. The host firewall rule protects the person opening email in a hotel.
Then stop the workstation offering NTLM to strangers. The policy named Network security: Restrict NTLM: Outgoing NTLM traffic to remote servers has an Audit all setting, which logs every outbound NTLM attempt, and a Deny all setting with a server exception list. Microsoft's own instruction is to audit first, review the log, then deny with exceptions. On Windows 11 version 24H2 and Windows Server 2025 there is a narrower switch as well: the SMB client can block NTLM for remote outbound connections, which Microsoft says prevents bad actors from tricking clients into sending NTLM requests to malicious servers.
Last, protect the accounts that matter most. Members of the Protected Users group cannot authenticate with NTLM at all against a domain controller, so a hash leaked from an administrator's mailbox session has nothing to relay to and no NTLM login to perform. Microsoft warns never to add service or computer accounts to the group, and members lose offline sign-in, so this is a control for privileged people, not for everyone.
Figure 02
The same opened file in two estates
WHERE THE HASH WORKS
-
TCP 445 open outbound from workstations
-
No host firewall rule on laptops off the network
-
Outgoing NTLM never audited or restricted
-
Servers accept NTLM without required signing
-
Administrators outside Protected Users
-
One portal that takes a password alone
One opened file is a login.
WHERE IT IS WORTHLESS
-
TCP 445 blocked outbound at the firewall
-
Public profile outbound SMB rules on every laptop
-
Outgoing NTLM denied, with a short exception list
-
SMB signing required on servers
-
Privileged people in Protected Users
-
A second factor in front of every sign-in
The attacker spent a zero-day and learned nothing.
How to detect an NTLM hash leaving the network
Microsoft has published no indicators for the attacks, so detection has to rest on the behaviour, and the behaviour is distinctive: a user workstation opening an SMB connection to a public address. Very few businesses have a legitimate reason for that. The Sigma rule below watches for it in Sysmon network connection events, Event ID 3, which Sysmon leaves disabled until you configure it.
One detail saves an afternoon of confusion. The rule does not filter on Outlook's process name. The Windows SMB client runs in the kernel on the program's behalf, so the outbound connection is commonly recorded against the System process and not against Outlook. Hunt on the destination, then work back to what the user had open at that moment.
Figure 03
A workstation opens SMB to a public address
Sigma
title: Workstation opens an SMB connection to a public address
status: experimental
description: An outbound connection on TCP 445 from a Windows endpoint to an
address outside private and link local ranges. Forced NTLM authentication
to an attacker's server looks like this. Add your approved cloud file
service ranges to the filter before alerting.
logsource:
product: windows
category: network_connection
detection:
selection:
Initiated: 'true'
DestinationPort: 445
filter_internal:
DestinationIp|cidr:
- '10.0.0.0/8'
- '172.16.0.0/12'
- '192.168.0.0/16'
- '127.0.0.0/8'
- '169.254.0.0/16'
- 'fc00::/7'
- 'fe80::/10'
- '::1/128'
condition: selection and not filter_internal
falsepositives:
- Azure Files or another approved file service reached over SMB
- Laptops on home networks that use public address space internally
level: high The second source costs nothing and most estates have never turned it on. With the outgoing NTLM policy set to Audit all, Windows records every NTLM authentication a device attempts to a remote server in the operational log under Applications and Services Logs, Microsoft, Windows, NTLM. A workstation offering NTLM to a name or address outside your organisation is the event you are looking for, and it fires whichever application was tricked.
Two limits, stated plainly. The advisory says attacker-controlled network location and does not name a protocol. SMB on port 445 is the usual carrier for this class and the one the rule watches; if your proxy logs show a workstation performing NTLM authentication to an outside web server, treat it the same way. And a firewall that already blocks 445 outbound will show the attempt as a denied connection. That is still worth an alert, because it means a user opened the file.
If you find a hit from before the patch, treat the user's password as disclosed: reset it, review sign-ins for that account from the date of the connection, and check whether the account can reach anything that accepts a password alone. Our write up on tokens that survive a password reset covers what a reset does not end.
How we test NTLM exposure on an engagement
We do not need a zero-day to ask the question this bug asks. On an internal test the first hour includes it by default, and our Active Directory recon guide lists the checks: which hosts accept SMB without required signing, which name resolution fallbacks are answering, what the password policy allows. Those are the facts that decide whether a captured authentication is useful, whoever captures it.
From the outside we test the egress. With the client's agreement we check from a standard workstation, and from a laptop on a guest network, whether an SMB connection to a server we control on the internet completes. The laptop off the network is the case most often left uncovered, because the perimeter rule does not travel with it. That single result tells the client more about their exposure to this advisory than the patch report does.
Then we measure what a captured hash would be worth. We audit password strength against an agreed wordlist and report by account, never by password. We map where NTLM is still accepted, especially on anything reachable from the internet. And we list the privileged people who are not in Protected Users, with the reason each was left out. When the people are in scope, an approved exercise measures how many open an attachment at all; our guide to a phishing simulation with an approved pretext shows how to run one without collecting a single password.
The report that comes out of this does not say Moderate. It says how many steps separate an opened attachment from a privileged session in this environment, and which one control removes the most steps for the least work. Most of the time that control is the outbound rule.
What is still coming
Expect imitation. The flaw was not public before the fix, and the patch now shows where to look. A bug that needs one opened file and no code execution is cheap to rebuild and cheap to send at scale, and anyone slow to update is the audience. Microsoft's advisory credits two finders, which tells us the bug was within reach of more than one party before it was fixed.
Expect the details of the exploitation to stay thin. Microsoft has said exploitation was detected and nothing more. The behavioural detection above does not depend on indicators, so do not wait for them.
Expect the class to outlive this patch. Microsoft's deprecation list says all versions of NTLM are deprecated and no longer under active feature development, and that NTLMv1 is removed starting in Windows 11 version 24H2 and Windows Server 2025. NTLMv2 still works, calls are meant to move to Negotiate with Kerberos first, and an estate that has not audited what still falls back to NTLM cannot turn it off. Until it can, any Windows feature that can be pointed at a remote path is a candidate for the next advisory that reads like this one. The outgoing NTLM audit log is how you find out what still depends on it. Turning it on this week serves the detection now and the retirement later.
What to do today
-
Update Outlook everywhere and confirm the running build. Office LTSC 2021 to build 16.0.14334.21048, Office LTSC 2024 to 16.0.17932.21140, and Microsoft 365 Apps to the 10 Nov 2026 security release for the channel. Restart Outlook and check the running build, not the deployment report; a downloaded update does nothing until then.
-
Block TCP port 445 outbound to the internet at the firewall. Microsoft's guidance is that you are unlikely to need it. Allow only the named cloud file services you use.
-
Push the public profile outbound SMB rule to every laptop. The perimeter rule does nothing for a user on home or hotel Wi-Fi, and that is where this file will be opened.
-
Set outgoing NTLM to Audit all, read the log for two weeks, then deny with exceptions. On Windows 11 version 24H2 and Windows Server 2025, turn on SMB client NTLM blocking as well.
-
Put privileged people in Protected Users and require SMB signing on servers. The first removes NTLM for the accounts that matter most; the second closes relay. Keep service and computer accounts out of the group.
-
Hunt back for workstations that connected out on port 445. The Sigma rule above is a starting point. Reset the password of any user behind a hit and review their sign-ins from that date.
SOURCES
- Microsoft Security Response Center // the advisory for CVE-2026-124194: Outlook spoofing, CVSS 6.5, Moderate, exploitation detected, not publicly disclosed, Preview Pane not a vector, the NTLMv2 hash disclosure, affected products and fixed builds
- CISA // the 10 Nov 2026 alert adding CVE-2026-124194 to the Known Exploited Vulnerabilities catalog on evidence of active exploitation
- CyberScoop // the release formally lists 1,094 CVEs, only one flaw is under active attack and none were publicly known before the release
- Microsoft Learn // block TCP port 445 outbound to the internet at the corporate firewall, and why outbound SMB rules on laptops matter
- Microsoft Learn // members of the Protected Users group cannot authenticate with NTLM, and service and computer accounts should never be added