Skip to content
RESEARCH INDEX BREACHROAD / INTELLIGENCE NOTE

Outlook CVE-2026-124194 leaks NTLMv2 without a warning

Microsoft confirms exploitation of CVE-2026-124194 in Outlook. A message can expose an NTLMv2 hash; learn how to patch, hunt and reduce relay risk.

PUBLIC RESEARCH
AUTHOR
/ CEO of Breachroad · OSCP · PNPT
PUBLISHED
10 November 2026
READING TIME
18 min read
TOPIC
Vulnerabilities and CVEs
Outlook CVE-2026-124194 leaks NTLMv2 without a warning

Microsoft published a fix for CVE-2026-124194 on 10 November 2026, addressing an Outlook flaw that can disclose an NTLMv2 hash. The vendor marks the vulnerability Exploitation Detected. An unauthenticated attacker can send a crafted message file and convince a user to open it; Outlook then connects to an attacker-controlled location without the usual warning for external content.

The primary source is the Microsoft Security Response Center advisory. A CVSS score of 6.5 and a Moderate/Important presentation depending on the view should not obscure the context: Microsoft confirms exploitation, the impact is credential-material disclosure, and subsequent risk depends on whether NTLM can be cracked or relayed in the victim environment.

The confirmed mechanism

MSRC provides AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:N/A:N. The attack is network-reachable and requires no prior privileges, but it does require user interaction. A crafted message file causes a connection to an external location without the normal warning, potentially exposing an NTLMv2 hash.

The Preview Pane is not an attack vector. Operationally, a user has to open the crafted item as described by the vendor. That does not make mail filtering sufficient. The file may arrive through other channels and can appear to be a legitimate export or attachment.

MSRC says the issue was not publicly disclosed before the update and that exploitation has been detected. Microsoft credits researchers including Benoit Sevens and notes the use of Anthropic/Claude tools in the research process. That is discovery attribution, not evidence that attackers used the same model.

Why an NTLMv2 hash matters

NTLMv2 challenge-response does not reveal a plaintext password, but captured material can be attacked offline when passwords are weak or relayed when the target service and configuration permit it. The viability of a next step depends on password length, signing, channel binding, protocol, network paths and server protections.

It would therefore be wrong to claim that opening a message automatically “compromises the domain.” It would be equally wrong to dismiss the leak because a hash is not a password. Proper assessment asks which identity responded, where it sent the material, which services accept NTLM, and whether they enforce anti-relay protection.

Immediate actions

Update every supported Outlook installation covered by the advisory, including off-network endpoints and VDI images. Confirm effective state through management telemetry, not merely package approval. A user who did not launch the client during the update window may remain affected.

In parallel, restrict outbound SMB and related protocols from user networks to the Internet. Only explicitly required destinations should be allowed at the perimeter. Egress control does not fix Outlook, but it can interrupt a common credential-leak path.

Then:

  1. search for public SMB/WebDAV connections and unusual authentication attempts;
  2. correlate them with Outlook and message-file activity;
  3. identify the user, host, external address and time;
  4. preserve the file and mail logs without opening it in a vulnerable client;
  5. assess subsequent account use and authentication to internal services;
  6. revoke sessions and reset credentials if evidence supports compromise;
  7. find other recipients of the same or related file.

Endpoint and network detection

EDR should correlate the mail client with a connection to an external host over a protocol Outlook does not normally require. Process name alone is insufficient because operating-system components may perform part of the activity. Network, DNS, file path, parent process, user and open time provide useful context.

At the firewall and proxy, find traffic to newly seen domains or addresses immediately after a message arrives. Identity logs should reveal failed and successful NTLM use from unusual hosts. In Windows environments, target-server authentication events and disagreement between the claimed workstation and observed source can be valuable.

We do not reproduce a PoC or a one-string signature. Such rules are often trivial to evade. Defenders should detect the behavioural class: a mail client initiates unexpected authentication towards an external resource.

Treating NTLM reduction as a programme

Disabling NTLM can materially reduce risk, but legacy applications, appliances and processes may depend on it. Begin with audit mode, identify owners, migrate to Kerberos or modern authentication and introduce blocks in stages. An exception without an expiry date becomes a permanent vulnerability.

Enforce SMB signing and Extended Protection where supported. These controls impede relay but need compatibility testing. Segment privileged identities: a domain administrator should not routinely read email on the workstation used for administration.

Our guide to reducing NTLM relay risk and practical Zero Trust architecture focuses on reducing dependencies, not blocking a single address.

User interaction does not mean low risk

Requiring a file to be opened can reduce the likelihood of fully unattended mass exploitation, but it does not remove social engineering. A lure can borrow current business context and appear to be a ticket, report or mailbox export. User education matters, but an employee should not be the final technical barrier.

Phishing and cybersecurity training should teach safe reporting rather than promise perfect recognition. The reporting button must lead to a process that isolates the artefact, searches other recipients and returns useful feedback.

Verifying the update

Check the installed build across Microsoft 365 Apps channels, LTSC and VDI variants present in the organisation. Then conduct a safe regression test against a controlled resource without capturing real credentials. Confirm that the client does not make the prohibited connection and that network and EDR layers would alert.

Where multiple Office releases exist, compliance reporting should show endpoint and user detail. An average “98% patched” does not reveal whether the remaining two per cent are executives, administrators or shared session hosts.

Microsoft facts and Breachroad analysis

Source facts: CVE-2026-124194 lets a remote unauthenticated attacker send a crafted message file; after interaction Outlook may connect without the normal external-content warning and disclose an NTLMv2 hash; Preview Pane is not a vector; Microsoft detected exploitation. Breachroad analysis: egress filtering, hunting, NTLM migration and the validation approach are defensive recommendations. We do not claim every disclosure leads to relay or domain compromise.

A 24-hour, 7-day and 30-day plan

Within 24 hours, patch Outlook, block unnecessary egress and start hunting. Within seven days, resolve anomalies, validate all update channels and test the reporting scenario. Within 30 days, inventory NTLM dependencies, define a reduction schedule and assess signing and Extended Protection.

For independent validation, an infrastructure penetration test can assess relay viability and segmentation without exposing production credentials. The best outcome is not only a patched Outlook client but an environment less capable of sending and accepting legacy authentication.

Mapping possible relay paths

Build a graph from user endpoints to services accepting NTLM: SMB, HTTP, LDAP and intermediaries. For each edge, record reachability, signing or Extended Protection, the account type able to respond and the consequence of authentication. This separates theoretical disclosure from a viable path.

Pay particular attention to privileged identities and administration hosts. If an administrator reads mail in the same profile, captured material has greater value. Separating PAWs, daily identities and mailboxes reduces many attack classes, not only this CVE.

Mail-layer protection

Filter message-file types according to business need, label external mail and detonate content in an environment with no real credentials. The sandbox must block authentication egress or it can leak a lab token and create a false safety result.

An extension rule is incomplete because files can be archived, linked or renamed. Correlate MIME, structure and behaviour. Preserve the original in an evidence repository limited to investigators.

Remote-work variants

An off-network laptop may send traffic directly through a home router, bypassing corporate egress. Check whether always-on VPN, host firewall and EDR provide equivalent control. A policy that only works in the office does not describe complete risk.

Offline endpoints need a deadline and access restriction until patched. Email reminders may not reach them. Conditional Access should use effective version and device posture, with expiring exceptions.

What successful hunting looks like

Results should state period, sources, queries, anomaly count and disposition. “We searched the SIEM” is insufficient. If the firewall did not log denied SMB or DNS retention is short, document that gap and apply greater caution to identity review.

For each suspicious event, join mail arrival, file opening, DNS resolution, egress, authentication and subsequent identity use. Only that timeline supports a decision on rotation and scope.

NTLM-reduction metrics

Report hosts and services using NTLM, privileged identities in those flows, signing enforcement, exceptions and deadlines. Track public SMB egress and emergency Office patch time too.

An identity and access security audit can sequence migration without abruptly breaking legacy systems. The objective is not a context-free “zero NTLM” score; it is removing any path where a user endpoint sends authentication material to an untrusted host and an internal service accepts it without protection.

A decision for the identity owner

After patching, select the three services generating the most NTLM traffic and assign their migration. Each needs an owner, target protocol, compatibility test, deadline and rollback. Exceptions should expire automatically; extension requires a new impact review.

Compare audit telemetry monthly. Falling usage without rising authentication failures demonstrates safe progress. If a dependency has no owner, restrict its segment and privileged-account access rather than leaving it indefinitely.

Keep the mail-client-to-external-authentication SOC rule after closing the CVE. It protects against future forced-connection techniques regardless of message format.

Test it quarterly with a controlled event and confirm visibility across process, DNS, network and identity. An alert that cannot identify host and user will not shorten response even when it technically fires.

SHARE / COPY