How the attack works
Protected Users and authentication policies exist to make stolen credentials of privileged accounts useless. A member of Protected Users cannot authenticate with NTLM and cannot use DES or RC4 in Kerberos pre-authentication, so the NTLM hash of the account, which is also its RC4 key, opens nothing. An authentication policy, assigned to an account directly or through an authentication policy silo, lets the domain controller check the device a sign-in comes from and the service a ticket is requested for, and refuse both when the conditions of the policy are not met. NTLM cannot carry these conditions, so it is rejected for such accounts as well.
An attacker who has stolen a hash or a password of a protected account does not know about these limits until the attempt fails. The failure is the detection. A pass-the-hash attempt with the hash of a Protected Users member is refused and logged, a ticket request built on the RC4 key is refused and logged, and an administrator account used from a workstation outside its silo is refused and logged. The attacker gains nothing, and the defender learns that the credential is in the wrong hands and where it was tried. The same events also appear when a protected account is used with an old application, which is why the logs were made for troubleshooting.
The events are written by the domain controller to two operational logs under Applications and Services Logs, Microsoft, Windows, Authentication. ProtectedUserFailures-DomainController holds event 100 for an NTLM sign-in failure of a Protected Users member and event 104 for a failed Kerberos pre-authentication with DES or RC4. AuthenticationPolicyFailures-DomainController holds event 101 for an NTLM sign-in that failed because an authentication policy is configured, events 105 and 106 for a ticket-granting ticket or a service ticket that an enforced policy denied, and events 305 and 306 for the same findings of a policy in audit mode. Both logs are disabled by default and have to be enabled on every domain controller before any of these events exist.
What an attacker needs
- A stolen credential of an account that is a member of Protected Users or is assigned an authentication policy, typically its NTLM hash or its password.
- For the silo events, an attempt from a device or to a service that the policy of the account does not allow.
Impact
Where the protections are enforced the attempt fails, and the impact is that an attacker holds a credential of a privileged account and is working inside the network. The attacker can still use the credential in ways the protections do not cover, for example a password used from a permitted device. Where an authentication policy is only in audit mode, events 305 and 306 report an authentication that succeeded and would have been denied, so the account was used from a place it should never be used from.
Detection rules
Authentication Policy Violation Reported in Audit Mode
Detects events 305 and 306 in the AuthenticationPolicyFailures-DomainController log of a domain controller. They are the informational events of an authentication policy that is in audit mode. Event 305 says that a Kerberos ticket-granting ticket would be denied because the device did not meet the access control restrictions, and event 306 says the same for a service ticket and the user, the device or both. An audited policy does not alter the process, so the authentication went ahead. The events therefore report an account that was used from a device, or towards a service, outside what its policy allows, and that succeeded. Microsoft provides audit mode to find blocking issues before a policy is enforced, so most events during a rollout are devices that still have to be added. Event 305 displays the account, the device, the policy, the silo names and the ticket lifetime, and event 306 the device, the policy and the silo names. The log source maps the event ID only. The log is disabled by default and has to be enabled on every domain controller.
title: Authentication Policy Violation Reported in Audit Mode
id: e3ab8f74-2e9b-4bc0-b2e4-adfe7ce231d8
status: experimental
description: >-
Detects events 305 and 306 in the
AuthenticationPolicyFailures-DomainController log of a domain controller.
They are the informational events of an authentication policy that is in
audit mode. Event 305 says that a Kerberos ticket-granting ticket would
be denied because the device did not meet the access control
restrictions, and event 306 says the same for a service ticket and the
user, the device or both. An audited policy does not alter the process,
so the authentication went ahead. The events therefore report an account
that was used from a device, or towards a service, outside what its
policy allows, and that succeeded. Microsoft provides audit mode to find
blocking issues before a policy is enforced, so most events during a
rollout are devices that still have to be added. Event 305 displays the
account, the device, the policy, the silo names and the ticket lifetime,
and event 306 the device, the policy and the silo names. The log source
maps the event ID only. The log is disabled by default and has to be
enabled on every domain controller.
references:
- https://learn.microsoft.com/en-us/windows-server/security/credentials-protection-and-management/authentication-policies-and-authentication-policy-silos
- https://learn.microsoft.com/en-us/windows-server/identity/ad-ds/manage/how-to-configure-protected-accounts
- https://attack.mitre.org/techniques/T1550/002/
author: Forestall
date: 2026-11-18
tags:
- attack.lateral-movement
- attack.t1550.002
logsource:
product: windows
service: authentication-policy-failures
detection:
selection:
EventID:
- 305
- 306
condition: selection
falsepositives:
- The rollout of a new policy, during which administrators still sign in from devices that were not added to the silo. That is what audit mode is for. Resolve each device, then enforce the policy.
level: low
False positives
- The rollout of a new policy, during which administrators still sign in from devices that were not added to the silo. That is what audit mode is for. Resolve each device, then enforce the policy.
Kerberos Ticket Denied by an Authentication Policy
Detects events 105 and 106 in the AuthenticationPolicyFailures-DomainController log of a domain controller. Event 105 is logged when a Kerberos ticket-granting ticket was denied because the device did not meet the enforced access control restrictions of the authentication policy. It displays the account, the device, the policy, the silo names and the ticket lifetime. Event 106 is logged when a Kerberos service ticket was denied because the user, the device or both did not meet the enforced restrictions. It displays the device, the policy and the silo names. Both mean that an account an organization confined to certain hosts was used from somewhere else or towards a service it may not reach. With a stolen password or key that is what an attacker does who does not know about the silo. The log source maps the event ID only, so the account and the device are read in triage. The log is disabled by default and has to be enabled on every domain controller.
title: Kerberos Ticket Denied by an Authentication Policy
id: 893ece34-fc2b-4b5e-8c89-3b6a104323e6
status: experimental
description: >-
Detects events 105 and 106 in the
AuthenticationPolicyFailures-DomainController log of a domain controller.
Event 105 is logged when a Kerberos ticket-granting ticket was denied
because the device did not meet the enforced access control restrictions
of the authentication policy. It displays the account, the device, the
policy, the silo names and the ticket lifetime. Event 106 is logged when
a Kerberos service ticket was denied because the user, the device or
both did not meet the enforced restrictions. It displays the device, the
policy and the silo names. Both mean that an account an organization
confined to certain hosts was used from somewhere else or towards a
service it may not reach. With a stolen password or key that is what an
attacker does who does not know about the silo. The log source maps the
event ID only, so the account and the device are read in triage. The log
is disabled by default and has to be enabled on every domain controller.
references:
- https://learn.microsoft.com/en-us/windows-server/security/credentials-protection-and-management/authentication-policies-and-authentication-policy-silos
- https://learn.microsoft.com/en-us/windows-server/identity/ad-ds/manage/how-to-configure-protected-accounts
- https://attack.mitre.org/techniques/T1550/002/
author: Forestall
date: 2026-11-18
tags:
- attack.lateral-movement
- attack.t1550.002
logsource:
product: windows
service: authentication-policy-failures
detection:
selection:
EventID:
- 105
- 106
condition: selection
falsepositives:
- An administrator who signs in from a workstation that is not in the silo yet, and devices that do not support Kerberos armoring, whose requests fail when the policy requires it. Add the workstation to the silo if it belongs there.
level: high
False positives
- An administrator who signs in from a workstation that is not in the silo yet, and devices that do not support Kerberos armoring, whose requests fail when the policy requires it. Add the workstation to the silo if it belongs there.
NTLM Sign-In Refused Because of an Authentication Policy
Detects event 101 in the AuthenticationPolicyFailures-DomainController log of a domain controller. Microsoft documents it as the event for an NTLM sign-in failure that occurs because an authentication policy is configured. NTLM authentication fails for such an account because access control restrictions are required and cannot be applied to NTLM. The event displays the account, the device, the policy and the silo names. Accounts with an authentication policy are the ones an organization chose to confine to certain hosts, so an NTLM attempt with one of them is either a leftover application or the use of a stolen hash that the policy stopped. The log source maps the event ID only, so the account and the device are read in triage. The log is disabled by default and has to be enabled on every domain controller.
title: NTLM Sign-In Refused Because of an Authentication Policy
id: 7c2a1831-499a-430c-aca6-6c4731fd20bc
status: experimental
description: >-
Detects event 101 in the AuthenticationPolicyFailures-DomainController
log of a domain controller. Microsoft documents it as the event for an
NTLM sign-in failure that occurs because an authentication policy is
configured. NTLM authentication fails for such an account because access
control restrictions are required and cannot be applied to NTLM. The
event displays the account, the device, the policy and the silo names.
Accounts with an authentication policy are the ones an organization
chose to confine to certain hosts, so an NTLM attempt with one of them
is either a leftover application or the use of a stolen hash that the
policy stopped. The log source maps the event ID only, so the account
and the device are read in triage. The log is disabled by default and
has to be enabled on every domain controller.
references:
- https://learn.microsoft.com/en-us/windows-server/security/credentials-protection-and-management/authentication-policies-and-authentication-policy-silos
- https://learn.microsoft.com/en-us/windows-server/identity/ad-ds/manage/how-to-configure-protected-accounts
- https://attack.mitre.org/techniques/T1550/002/
author: Forestall
date: 2026-11-18
tags:
- attack.lateral-movement
- attack.t1550.002
logsource:
product: windows
service: authentication-policy-failures
detection:
selection:
EventID: 101
condition: selection
falsepositives:
- Applications and scripts that still use NTLM for an account that was placed under an authentication policy. They repeat from one host. Move them to Kerberos or give them an account without the policy.
level: high
False positives
- Applications and scripts that still use NTLM for an account that was placed under an authentication policy. They repeat from one host. Move them to Kerberos or give them an account without the policy.
Kerberos Pre-Authentication With DES or RC4 Refused for a Protected Users Member
Detects event 104 in the ProtectedUserFailures-DomainController log of a domain controller. Microsoft documents it for the case that DES or RC4 encryption types are used for Kerberos authentication and a sign-in failure occurs for a user in the Protected Users security group. The pre-authentication failed because DES and RC4 cannot be used when the account is a member of the group, while AES is accepted. The RC4 key of an account is its NTLM hash, so a ticket request built on RC4 for a protected account is what an attacker produces who holds the hash and not the password. A current Windows client that knows the password uses AES. The other source is an account that has no AES keys yet because its password was not changed since the domain supports them. The log source maps the event ID only, so the account is read in triage. An event with the same ID exists in the ProtectedUser-Client log on client computers and is not what this rule reads. The log is disabled by default and has to be enabled on every domain controller.
title: Kerberos Pre-Authentication With DES or RC4 Refused for a Protected Users Member
id: 3771a26f-0f7c-4034-8fa3-79358ae9213c
status: experimental
description: >-
Detects event 104 in the ProtectedUserFailures-DomainController log of a
domain controller. Microsoft documents it for the case that DES or RC4
encryption types are used for Kerberos authentication and a sign-in
failure occurs for a user in the Protected Users security group. The
pre-authentication failed because DES and RC4 cannot be used when the
account is a member of the group, while AES is accepted. The RC4 key of
an account is its NTLM hash, so a ticket request built on RC4 for a
protected account is what an attacker produces who holds the hash and
not the password. A current Windows client that knows the password uses
AES. The other source is an account that has no AES keys yet because its
password was not changed since the domain supports them. The log source
maps the event ID only, so the account is read in triage. An event with
the same ID exists in the ProtectedUser-Client log on client computers
and is not what this rule reads. The log is disabled by default and has
to be enabled on every domain controller.
references:
- https://learn.microsoft.com/en-us/windows-server/security/credentials-protection-and-management/protected-users-security-group
- https://learn.microsoft.com/en-us/windows-server/identity/ad-ds/manage/how-to-configure-protected-accounts
- https://attack.mitre.org/techniques/T1550/002/
author: Forestall
date: 2026-11-18
tags:
- attack.lateral-movement
- attack.t1550.002
logsource:
product: windows
service: protected-user-failures
detection:
selection:
EventID: 104
condition: selection
falsepositives:
- An account without AES keys, because its password was last changed before the domain supported AES, and clients or appliances that request RC4 only. Change the password of the account and update the client.
level: high
False positives
- An account without AES keys, because its password was last changed before the domain supported AES, and clients or appliances that request RC4 only. Change the password of the account and update the client.
NTLM Sign-In Refused for a Protected Users Member
Detects event 100 in the ProtectedUserFailures-DomainController log of a domain controller. Microsoft documents it as an error that is logged when NTLM authentication failed because the account was a member of the Protected Users security group. The event displays the account name and the device name. Members of Protected Users cannot authenticate with NTLM, so an attempt to do so comes from an application that was never moved to Kerberos or from someone who holds the NTLM hash of the account and tries to use it. For a privileged account the second case is a pass-the-hash attempt that the protection stopped. The log source maps the event ID only, so the account and the device are read in triage. The log is disabled by default and has to be enabled on every domain controller.
title: NTLM Sign-In Refused for a Protected Users Member
id: ad5348e2-4f1c-4122-b647-56ffbed50654
status: experimental
description: >-
Detects event 100 in the ProtectedUserFailures-DomainController log of a
domain controller. Microsoft documents it as an error that is logged
when NTLM authentication failed because the account was a member of the
Protected Users security group. The event displays the account name and
the device name. Members of Protected Users cannot authenticate with
NTLM, so an attempt to do so comes from an application that was never
moved to Kerberos or from someone who holds the NTLM hash of the account
and tries to use it. For a privileged account the second case is a
pass-the-hash attempt that the protection stopped. The log source maps
the event ID only, so the account and the device are read in triage. The
log is disabled by default and has to be enabled on every domain
controller.
references:
- https://learn.microsoft.com/en-us/windows-server/security/credentials-protection-and-management/protected-users-security-group
- https://learn.microsoft.com/en-us/windows-server/identity/ad-ds/manage/how-to-configure-protected-accounts
- https://attack.mitre.org/techniques/T1550/002/
author: Forestall
date: 2026-11-18
tags:
- attack.lateral-movement
- attack.t1550.002
logsource:
product: windows
service: protected-user-failures
detection:
selection:
EventID: 100
condition: selection
falsepositives:
- Applications, mapped drives and scripts that still authenticate with NTLM for an account that was added to Protected Users. They repeat from one host. Move them to Kerberos, because every remaining event then points to a stolen hash.
level: high
False positives
- Applications, mapped drives and scripts that still authenticate with NTLM for an account that was added to Protected Users. They repeat from one host. Move them to Kerberos, because every remaining event then points to a stolen hash.
Detection logic
Patterns that need more than one event, a baseline or knowledge of your own environment. Each is a hypothesis to build and tune in your SIEM, not a ready rule.
Protected account failure joined to its source in the Security log
The Authentication logs say that a protected account was refused, and the Security log says from where. Joined on the account and the time, the two answer the question an analyst asks first, and a source that is not an administrative workstation separates a stolen credential from an old application.
- Take event 100 or 101 (an NTLM sign-in of a protected account failed) from the Authentication logs of a domain controller and read the account name and the device name.
- On the same domain controller, within 1 minute, find event 4776 with the same account in
TargetUserName. Read the source computer inWorkstationand the result inStatus. - For event 104 (pre-authentication with DES or RC4 failed), find event 4771 with the same account in
TargetUserNameand read the client address inIpAddress. - Compare the workstation or address with your list of administrative workstations for that account, and report every source that is not on it.
Events10010110447764771
LimitsIt needs both Authentication logs enabled, the Security log of the same domain controller and your list of administrative workstations. The workstation name in event 4776 is supplied by the client and can be empty or false. The 1 minute is a starting point. Applications that still use NTLM for a protected account repeat from one known host and are excluded by that host.
Triage
- Read the account and the device in the event. Events 100 and 101 show the account name and the device name, and events 105, 106, 305 and 306 show the device, the policy and the silo.
- Find the source in the Security log of the same domain controller at the same time. An NTLM failure appears as event 4776 with the workstation, and a failed pre-authentication as event 4771 with the client address.
- Decide whether the source is a host the account is meant to use. A privileged account that is tried from a user workstation or a server of a lower tier is a stolen credential until shown otherwise.
- Look for an application as the cause, such as a scheduled job, a mapped drive or a management tool that still uses NTLM for that account. It repeats from the same host at regular intervals.
- Check what else was tried from the same source, with this account and with others, in the hours around the event.
Response
Once the attack is confirmed:
- Reset the password of the account so that the stolen hash or password stops working, and end its sessions.
- Isolate the host the attempt came from and investigate how the credential reached it.
- Search both Authentication logs and the Security log of every domain controller for other protected accounts that were tried from the same source.
- Review where the account signed in during the last weeks to find the host on which its credential was exposed.
- If the events came from a policy in audit mode, treat the authentication as successful and review what the account did from that device.
Mitigation
- Put privileged accounts in Protected Users after testing, and assign them authentication policies that limit sign-in to the administrative workstations of their tier.
- Enforce authentication policies once the audit events are understood. A policy in audit mode reports and does not block.
- Enable the ProtectedUserFailures-DomainController and AuthenticationPolicyFailures-DomainController logs on every domain controller and collect them, because they are disabled by default.
- Remove the dependencies of privileged accounts on NTLM and RC4, so that every remaining failure in these logs is a finding and not an application.
- Change the passwords of protected accounts after an authentication policy is configured, as Microsoft advises, to invalidate credentials that were cached before.
Frequently asked questions
Why is a failed sign-in of a protected account worth an alert?
The protections refuse exactly what stolen credentials are used for. NTLM with the hash of a Protected Users member, a ticket request built on its RC4 key, or a privileged account used from a device outside its silo. A legitimate user of such an account does not produce these failures once old applications are cleaned up, so each one is a credential in the wrong place.
Are these logs enabled by default?
No. Microsoft states that the Protected User and Authentication Policy logs under Applications and Services Logs, Microsoft, Windows, Authentication are disabled by default. Each log has to be enabled on every domain controller, and collected from there, before the rules on this page can match anything.
What is the difference between events 105 and 305?
Event 105 is written when an enforced authentication policy denied a ticket-granting ticket because the device did not meet the access control restrictions. Event 305 is the informational event of a policy in audit mode. It reports that the ticket would be denied, and the authentication goes ahead. Events 106 and 306 relate in the same way for service tickets.
Do the rules show where the attempt came from?
Only through the event text. The log sources map the event ID alone, so the rules cannot filter on the account or the device. The device name is shown in the event, and the Security log of the same domain controller holds the matching failure with the workstation or the client address (events 4776 and 4771).
References
- MITRE ATT&CK T1550.002: Use Alternate Authentication Material: Pass the Hash
- Protected Users security group (events 100 and 104)
- Authentication Policies and Authentication Policy Silos (events 101, 105, 106, 305, 306)
- Guidance about how to configure protected accounts
- Event 4776: The computer attempted to validate the credentials for an account
- Event 4771: Kerberos pre-authentication failed