- Nov 13
Don't Let the 'Moderate' Score Fool You: The Outlook Bug That Leaks Your Password Hash, and How to Stop It
- IAMPathway Trainer
- Microsoft Entra ID
Your bookkeeper opens a message file attached to an email from a "client," decides it's junk and deletes it. No warning, no prompt, no crash.
In the second it took to open, classic Outlook quietly reached out to a server the attacker controls and handed over a scrambled form of her Windows credential: her NTLMv2 hash. In Microsoft's own words, the attacker "could use this to authenticate as the user."
That's CVE-2026-124194, fixed in Microsoft's November 10, 2026 Patch Tuesday. Microsoft says it has detected exploitation, and CISA added it to its Known Exploited Vulnerabilities (KEV) catalog the same day. Yet Microsoft rates it only Moderate, CVSS 6.5, the kind of label that gets pushed to "next month" in a release of more than a thousand fixes.
Don't push it. Here's what the bug does, why a leaked hash is as good as a stolen credential, and what a small business or MSP should do today, this week and this quarter.
Everything here was checked against Microsoft's Security Update Guide, Microsoft Learn, NIST's National Vulnerability Database and CISA on November 11, 2026. Microsoft hasn't published attack details or indicators of compromise, so this post doesn't guess at them.
What Microsoft actually says
Here's what the Security Update Guide entry says, with nothing added:
Title: Microsoft Outlook Spoofing Vulnerability.
Weakness: CWE-356, "Product UI does not Warn User of Unsafe Actions."
Severity: Moderate. CVSS 3.1 base score 6.5, temporal score 6.2.
Status: Exploitation Detected. Not publicly disclosed before the fix.
How it's triggered: "An unauthenticated attacker could send a specially crafted message file and convince a user to open it in an affected version of Microsoft 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."
Preview Pane: "No, the Preview Pane is not an attack vector." The user has to open the file.
The prize: "This vulnerability discloses a user's NTLMv2 hash to the attacker who could use this to authenticate as the user."
Affected products: Microsoft 365 Apps for Enterprise, Office LTSC 2021 and Office LTSC 2024, in both 32-bit and 64-bit editions.
What Microsoft hasn't said matters just as much. As SecurityWeek noted, the advisory "provides no information about the observed attacks." There's no named threat actor, no target list, no file names, no IP addresses or domains, and no detail on which network protocol the outbound connection uses. Anyone selling you "the IOCs for CVE-2026-124194" today is making them up.
Which Outlook is affected?
All three affected products are Windows desktop Office suites, so this is a classic Outlook for Windows problem. Rapid7's Patch Tuesday analysis puts it the same way: "All supported versions of classic Outlook receive patches."
Outlook on the web, the new Outlook for Windows and Outlook mobile aren't listed as affected in Microsoft's advisory. Microsoft doesn't explicitly say they're immune, though, so treat "not listed" as "not part of this fix," not as a guarantee about those apps.
Why a "Moderate" bug deserves a same-week patch
CVSS describes how bad a bug is in the abstract. It doesn't tell you whether anyone is using it. Look at the vector string Microsoft published:
CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:N/A:NIn plain English: reachable over the network (AV:N), easy to pull off (AC:L), no account needed (PR:N), but the user has to do something (UI:R). The only impact scored is high confidentiality loss (C:H). The leaked secret is what drives the score. Integrity and availability are scored "None" because the bug itself doesn't change or break anything.
That's where the score undersells the risk. It measures what the bug does to Outlook, not what an attacker does next with your credential: sign in to a file server, relay it to a certificate service, or crack it and try the password on your VPN.
Three signals beat the score here:
Microsoft says it's being exploited. Its exploitability field reads "Exploitation Detected."
CISA put it in the KEV catalog. It's listed there as "Microsoft Office Outlook Missing User Interaction for Potentially Unsafe Action Vulnerability," added November 10, 2026, with a remediation due date of November 24, 2026 for US federal civilian agencies, per the KEV fields shown on the NVD entry. That deadline doesn't bind a dental office in Indianapolis, but CISA explicitly encourages every organization to prioritize KEV entries.
The trigger is ordinary. Opening an attachment from a plausible sender is something your users do dozens of times a day.
Rapid7 makes the same argument: the "relatively low CVSS v3.1 base score of 6.5 does not tell the full story," because the bug is useful both to commodity attackers harvesting access at scale and to targeted campaigns.
We've seen this movie before. In March 2023, CVE-2023-23397 also made Outlook leak a user's Net-NTLMv2 hash to an attacker-controlled location. That one was scored 9.8 Critical, because a crafted email triggered it automatically, before the user even opened it. The new bug needs a click, which is why its score is lower. The stolen item is the same.
Why a hash is a credential
Most people hear "hash" and think "safely scrambled." For NTLM, that's the wrong mental model.
NTLM is Windows' older challenge-response sign-in protocol. Microsoft's NTLM overview describes it as proving "that a user knows the password associated with an account." When Windows connects to a server that asks for NTLM, it can answer automatically with a response computed from your password, without asking you. Single sign-on is a feature. Here, it's the problem.
When Outlook connects out to an attacker's server, that server can ask for NTLM and capture the NTLMv2 response. At a high level, the attacker can then do one of two things. Both are described in Rapid7's analysis, and relay is spelled out in Microsoft's 2023 advisory for a similar Outlook bug (more on that above):
Relay it. The attacker forwards your live authentication to another service on your network, such as a file server, LDAP on a domain controller, Exchange or Active Directory Certificate Services, and signs in as you. Rapid7 notes this works where those services don't enforce safeguards like SMB or LDAP signing or Extended Protection for Authentication (EPA).
Crack it offline. The attacker guesses passwords on their own hardware until one produces the same response. A short or reused password falls fast. A long, unique one may never fall.
Microsoft's own policy documentation lists the threats bluntly: NTLM and NTLMv2 are "vulnerable to various malicious attacks, including SMB relay, man-in-the-middle attacks, and brute force attacks."
What about my cloud-only Microsoft 365 tenant?
The hash belongs to the Windows account signed in to that PC. With on-premises Active Directory, it's the worst case: a relayed or cracked domain credential opens file shares, internal apps and more. With password hash sync to Entra ID, a cracked password is probably the user's Microsoft 365 password too, so MFA is what stands between it and the mailbox. Cloud-only shops have less to relay to, but a cracked password is still compromised.
Step 1: Patch classic Outlook today
This fix ships in Office, not in the Windows monthly update. Microsoft's Security Update Guide lists no KB number for the Microsoft 365 Apps or LTSC entries. Instead it points to Microsoft's Office security update release notes, which list the November 10, 2026 builds by channel. The version and build must be at least:
Current Channel: Version 2610 (Build 20527.20196)
Monthly Enterprise Channel: Version 2610 (Build 20527.20194), Version 2609 (Build 20430.20264) or Version 2608 (Build 20326.20320)
Semi-Annual Enterprise Channel: Version 2610 (Build 20527.20192), Version 2609 (Build 20430.20262) or Version 2608 (Build 20326.20318)
Office LTSC 2024 (volume licensed): Version 2408 (Build 17932.21140), which MSRC lists as fixed build 16.0.17932.21140
Office LTSC 2021 (volume licensed): Version 2108 (Build 14334.21048), which MSRC lists as fixed build 16.0.14334.21048
The same release notes also list November builds for Office 2024 Retail and Office 2021 Retail (Build 20527.20196) and Office 2019 volume licensed (Build 10417.20236). MSRC's affected list for this CVE doesn't name those editions, but if they're in your fleet, install the November update anyway. There's no reason to sit a month behind on Office.
For a small business without central management
On each PC, open classic Outlook and go to File > Office Account.
Select Update Options > Update Now. If the button is missing or greyed out, someone (maybe your MSP) is managing updates centrally. Skip to the next section.
Restart Outlook when prompted, then return to File > Office Account and check the version and build under About Outlook.
Compare the build to the list above. If it's older, the update didn't land.
For MSPs and managed fleets
Inventory channels first. One tenant can have devices on Current Channel and others on Monthly Enterprise. Microsoft's update channel overview explains how each one is serviced.
Push or approve the November builds through whatever manages Office for each client: Intune, Configuration Manager, Cloud Update in the Microsoft 365 Apps admin center, or your RMM's Office patching.
Shorten restart deadlines. Office doesn't apply the fix until it restarts.
Report on build numbers, not "update succeeded," and flag anything below its channel's fixed build.
Chase the stragglers: shared front-desk PCs, the owner's home laptop, Remote Desktop hosts and golden images.
Until a device is patched, tell its users plainly: don't open message files or attachments you weren't expecting, even from a name you know, and forward anything odd to IT instead.
Step 2: Block outbound SMB to the internet this week
Microsoft hasn't said which protocol the outbound connection uses. But for years, the most common way Windows leaks NTLM to the internet has been SMB on TCP port 445, and Microsoft's own SMB security guidance says to block it regardless:
Block TCP port 445 outbound to the internet at your corporate firewall.
Microsoft adds that you're unlikely to need outbound 445 to the internet at all, unless you use a public cloud service like Azure Files over SMB, in which case it recommends a VPN restricted to the service's IP ranges.
This is defense in depth, not a substitute for the patch. If the leak uses another protocol, a port 445 block won't stop it. That's why Step 3 matters too.
At the office firewall
Add an outbound deny rule for TCP 445 from internal networks to any internet destination.
Before you apply it, check whether anyone uses Azure Files over SMB directly. If so, route that traffic through a VPN as Microsoft recommends, or allow only the required service IP ranges.
Log the rule's hits for a week. Any internal machine trying port 445 to the internet deserves a second look.
On laptops that leave the office
The office firewall does nothing for the sales rep on hotel Wi-Fi. Microsoft's guidance calls out an outbound template for guest and public networks as "important to enforce on mobile devices and home-based telecommuters," because it "reduces the odds of phishing attacks that send users to malicious servers to harvest credentials."
A simple starting point is a Windows Firewall rule that blocks outbound SMB on the Public profile. Deploy it through Intune, Group Policy or your RMM, and test on a pilot group first:
# Block outbound SMB (TCP 445) when on a Public network.
# Test on a pilot group first; Azure Files over SMB will break.
New-NetFirewallRule `
-DisplayName "Block outbound SMB on Public networks" `
-Direction Outbound `
-Protocol TCP `
-RemotePort 445 `
-Profile Public `
-Action BlockThe tradeoff: if a laptop wrongly marks the office network as "Public," users lose file shares there, so check network profiles on a few devices first. Microsoft's guidance links a support article with fuller templates, including allowlists for domain controllers and file servers.
On Windows 11 24H2 and Windows Server 2025: block NTLM on the SMB client
Newer Windows adds a stronger control. Per Microsoft Learn's SMB NTLM blocking page, the SMB client in Windows 11, version 24H2 and Windows Server 2025 can refuse to send NTLM on outbound SMB connections, which "prevents bad actors from tricking clients into sending NTLM requests to malicious servers." The Group Policy setting is Computer Configuration > Administrative Templates > Network > Lanman Workstation > Block NTLM (LM, NTLM, NTLMv2). The PowerShell equivalent is:
# Windows 11 24H2 / Server 2025 only.
# Destination SMB servers must support Kerberos.
Set-SmbClientConfiguration -BlockNTLM $trueThe tradeoff is real. Any SMB server that can't do Kerberos, such as a workgroup NAS, a non-domain-joined server or an old scanner share, stops working unless you add it to the Block NTLM Server Exception List policy. In a small office with one Synology box and no domain, that box is probably your exception. Audit first, then enforce.
Step 3: Shrink where NTLM can be used and relayed
Patching closes this bug, not the next one that leaks a hash. Microsoft already lists all versions of NTLM, including NTLMv2, as deprecated on its deprecated features page, and NTLMv1 has been removed starting in Windows 11, version 24H2 and Windows Server 2025. Microsoft's advice for applications is to call Negotiate, which tries Kerberos first and only falls back to NTLM when necessary.
You don't have to turn NTLM off in one weekend. You do have to start. Here's a sequence that works for small environments.
1. Make relay useless where it hurts most
Relay needs a target that accepts it. Remove the easy ones:
SMB signing. Signing defeats SMB relay. Microsoft Learn's SMB signing page says Windows 11 24H2 Enterprise, Pro and Education require outbound and inbound signing by default, Windows Server 2025 requires outbound signing, and Windows 11 Home requires neither. Older servers need signing turned on deliberately. Check your file servers.
Exchange Server, if you still run one on-premises. Enable Extended Protection. Microsoft's Exchange guidance says it's on by default from Exchange Server 2019 CU14, but upgraded or older servers may not have it.
Active Directory Certificate Services. If you run AD CS with web enrollment, follow Microsoft's KB5005413 to enable EPA and require HTTPS. A relayed credential that turns into a certificate is a long-lived foothold.
2. Put your most valuable accounts in Protected Users
For CVE-2023-23397, Microsoft's own mitigation guidance suggested adding high-value accounts to the Protected Users group, "which prevents the use of NTLM as an authentication mechanism." Microsoft Learn's Protected Users page lists the conditions:
It needs a Windows Server 2012 R2 or later domain functional level.
Never add service or computer accounts.
Test before adding Domain Admins or Enterprise Admins. Members lose offline sign-in and can't bypass the restrictions, so you can lock admins out.
3. Audit outgoing NTLM, then restrict it
The policy Network security: Restrict NTLM: Outgoing NTLM traffic to remote servers (under Computer Configuration > Windows Settings > Security Settings > Local Policies > Security Options) has three useful values: Allow all, Audit all and Deny all. Microsoft's policy documentation gives the safe order:
Set Audit all on a pilot group. No restart is needed.
Review the events in Applications and Services Logs > Microsoft > Windows > NTLM to see which servers still receive NTLM.
Fix what you can, such as connecting by DNS name instead of IP address and upgrading apps that only speak NTLM.
Add the leftovers to Network security: Restrict NTLM: Add remote server exceptions for NTLM authentication.
Move to Deny all for the pilot, then expand.
Microsoft warns that denying outgoing NTLM "may affect compatibility with client computers, services, and applications." That's why the audit step isn't optional.
What this costs
None of the Windows controls here needs an extra license; they're built into Windows and Active Directory. The cost is time and testing around old NAS boxes, printers and apps that only speak NTLM. Advanced hunting (next section) depends on your Defender plan.
Step 4: Check for signs of trouble, honestly
Microsoft published no IOCs for this vulnerability. So instead of matching a list of indicators, look for the behavior the fix is meant to stop: devices making SMB connections to the internet. In a healthy small business, that should be close to zero.
If you have Microsoft Defender advanced hunting (it comes with Defender for Endpoint and the Defender XDR portal; check Microsoft's Defender for Business overview for what your plan includes), this query uses columns documented in the DeviceNetworkEvents table reference:
// Outbound SMB to public IPs over the last 14 days.
// A hygiene hunt, not an IOC match for CVE-2026-124194.
DeviceNetworkEvents
| where Timestamp > ago(14d)
| where RemotePort == 445
| where RemoteIPType == "Public"
| summarize Connections = count(),
FirstSeen = min(Timestamp),
LastSeen = max(Timestamp),
Processes = make_set(InitiatingProcessFileName, 10)
by DeviceName, RemoteIP, RemoteUrl
| order by LastSeen descThe query hasn't been tested in a live tenant, so try it in a test workspace first. Expect some noise from Azure Files; anything else, especially a workstation reaching an unfamiliar public address, is worth investigating. Windows often attributes SMB connections to a system process rather than to the app that caused them, so don't filter on outlook.exe alone and assume you're clean.
No Defender? Your firewall logs answer the same question. Search for denied or allowed outbound TCP 445 to public addresses, especially after you add the block rule in Step 2.
If you find a hit, or a user admits opening a strange file
Patch that device immediately and confirm the build.
Reset the user's password to a long, unique one. A leaked hash is only as useful as the password behind it, and a reset makes captured material stale.
Revoke their sessions in Microsoft 365, and review recent sign-ins and mailbox rules. Our guide to revoking sessions when an attacker is still in Microsoft 365 walks through it.
Check sign-ins for the same account across Entra ID and on-premises logs for activity the user doesn't recognize. Our sign-in log hunting guide shows where to look.
If the user is an admin, escalate. Assume their on-premises rights may have been used and bring in incident response help.
A practical rollout order for a small tenant or MSP
Here's the order we'd follow for a 10 to 200-seat business, or for each client in an MSP book:
Today: confirm which Office channels and editions are in use, and push the November Office update.
Today: send users a two-line warning not to open unexpected message files or attachments until IT confirms their machine is updated.
Within 48 hours: report on Office build numbers and chase every device that's below its channel's fixed build.
This week: add the outbound TCP 445 deny rule at every office firewall, after checking for Azure Files.
This week: deploy the Public-profile outbound SMB block to laptops through Intune, Group Policy or your RMM, starting with a pilot.
This week: run the hygiene hunt above, or search firewall logs, for outbound 445 to the internet.
This month: confirm SMB signing on file servers, Extended Protection on any on-premises Exchange, and EPA on AD CS web enrollment.
This month: add IT admin accounts to Protected Users after testing.
This quarter: set Audit all for outgoing NTLM on a pilot group, fix what the audit shows, and plan SMB client NTLM blocking for Windows 11 24H2 devices.
Every month after: treat anything on CISA's KEV list as urgent, whatever its CVSS score.
For MSPs, steps 4, 5 and 9 make a good standard baseline, with a documented exception list per client (the NAS, the scanner, the accounting app). And keep MFA strong on the cloud side, while remembering its limits, as we covered in MFA can't save you when attackers log in as your apps.
Key takeaways
CVE-2026-124194 is rated Moderate (CVSS 6.5) but is actively exploited and on CISA's KEV list. Exploitation status beats the score.
Opening a crafted message file in classic Outlook can leak the user's NTLMv2 hash to an attacker-controlled location, with no external-content warning. The Preview Pane alone doesn't trigger it.
A leaked NTLM hash is a usable credential. It can be relayed to sign in as the user, or cracked offline to recover the password.
Patch through Office, not Windows Update. Fixed builds include 16.0.14334.21048 (LTSC 2021), 16.0.17932.21140 (LTSC 2024) and the November 10 builds for each Microsoft 365 Apps channel.
Microsoft published no attack details or IOCs. Hunt for the behavior (SMB to the internet), not indicators nobody has.
Block outbound TCP 445 to the internet at the firewall and on roaming laptops.
Shrink NTLM over time with signing, Extended Protection, Protected Users and NTLM auditing. Microsoft has deprecated NTLM, so this pays off for every future hash-leak bug.
Official sources
Microsoft Security Update Guide: CVE-2026-124194, Microsoft Outlook Spoofing Vulnerability (released November 10, 2026)
CISA: CISA Adds Six Known Exploited Vulnerabilities to Catalog (November 10, 2026)
Microsoft Learn: Release notes for Microsoft Office security updates
Microsoft Learn: Update history for Microsoft 365 Apps (listed by date)
Microsoft Learn: Overview of update channels for Microsoft 365 Apps
Microsoft Learn: Network security: Restrict NTLM: Outgoing NTLM traffic to remote servers
Microsoft Learn: Configure Windows Extended Protection in Exchange Server
Microsoft Learn: DeviceNetworkEvents table in the advanced hunting schema
Also referenced: Rapid7, Patch Tuesday - November 2026 (November 10, 2026), and SecurityWeek, Exploited Outlook Zero-Day Among 1,094 Flaws Patched by Microsoft (November 10, 2026).
About IAMPathway
IAMPathway publishes practical guides on Identity & Access Management: Microsoft Entra ID, governance, identity threats, Application Governance and AI Governance. Browse all topics or read the glossary.