1. Resources
  2. Detection rules
  3. Kerberos Replay Attack

Kerberos Replay Attack detection rules

A captured Kerberos request is sent again to gain access as the original client. A domain controller that receives a request it has already seen answers KRB_AP_ERR_REPEAT and logs event 4649.

Platform
Active Directory
Severity
Medium
Log source
Windows Security event log
Last updated
Rule formats
Sigmaand queries for: SplunkMicrosoft SentinelIBM QRadar
What an attacker needs
A position on the network from which Kerberos traffic to a domain controller can be captured and sent again within the clock tolerance.
Impact
An accepted replay would give access as the original client without a password, and a real attempt shows that somebody can capture authentication traffic.
Response
Find and isolate the host or network device that sent the duplicate requests, and reset the password of the account whose requests were replayed.
Mitigation
Keep the Kerberos clock tolerance at the default of 5 minutes, and protect the network path between clients and domain controllers.
Event IDs
464947684769

How the attack works

Kerberos proves the identity of a client with two things that are sent together: a ticket, and an authenticator that holds a fresh timestamp encrypted with the session key of that ticket. An attacker who can capture this traffic, for example from a position between a client and a server, does not learn any key from it. A replay attack sends the captured request again unchanged, in the hope that the server accepts it as a new request from the original client.

The protocol defends against this with time. Microsoft's description of the Kerberos policy explains that timestamps are part of the protocol to prevent replay attacks, and that a timestamp counts as authentic only while the clocks of both sides differ by less than the configured tolerance, which is five minutes by default. Within that window the server has to remember what it has already seen. Domain controllers cache information from recently received tickets, and when the server name, client name, time and microsecond fields of an authenticator match an entry in the cache they answer with the error KRB_AP_ERR_REPEAT.

A domain controller records that answer as event 4649 (a replay attack was detected) when Audit Other Logon/Logoff Events is enabled. The event is evidence of a duplicate request, not of an attacker. Microsoft names a misconfigured network device that sends the same packets repeatedly as one possible cause, and recommends raising an alert and investigating the reason in either case. The event is written on domain controllers, so a replay against another server does not appear here. A second rule reads the same answer from the Kerberos audit events of the domain controllers, a failed ticket request (4768 or 4769) with result code 0x22, which Microsoft's monitoring recommendations for event 4768 name as a possible attack attempt.

What an attacker needs

  • A position on the network from which Kerberos traffic to a domain controller can be captured and sent again, such as a compromised host or device on the path.
  • A replay inside the clock tolerance of the domain, before the captured authenticator is too old to be accepted.

Impact

If a replayed request were accepted, the attacker would act as the original client towards that service without knowing a password or a key. The cache on the domain controller makes a plain replay of the same request fail, so the event usually marks an attempt that was blocked or a duplicate that was never an attack. Its value is in what a real replay implies. Somebody is able to capture authentication traffic of the domain, and that position also allows other attacks on the same traffic.

Detection rules

Kerberos Replay Detected by a Domain Controller

Detects event 4649 (a replay attack was detected), which a domain controller writes when it answers a Kerberos request with KRB_AP_ERR_REPEAT because the server name, client name, time and microsecond fields of the authenticator match a request it has recently received. The event means that the same request arrived twice. That is either a captured request sent again or, as Microsoft notes, a network device that sends the same packets repeatedly, and Microsoft recommends an alert and an investigation in both cases. The event page gives no example and no field names, so the rule matches on the event ID. Requires Audit Other Logon/Logoff Events on domain controllers.

LevelMediumStatusExperimentalLog sourceWindows Security event log
title: Kerberos Replay Detected by a Domain Controller
id: a9da7c2b-aed2-47f9-9829-fd3bf6e732ff
status: experimental
description: >-
  Detects event 4649 (a replay attack was detected), which a domain controller
  writes when it answers a Kerberos request with KRB_AP_ERR_REPEAT because the
  server name, client name, time and microsecond fields of the authenticator
  match a request it has recently received. The event means that the same
  request arrived twice. That is either a captured request sent again or, as
  Microsoft notes, a network device that sends the same packets repeatedly,
  and Microsoft recommends an alert and an investigation in both cases. The
  event page gives no example and no field names, so the rule matches on the
  event ID. Requires Audit Other Logon/Logoff Events on domain controllers.
references:
  - https://learn.microsoft.com/en-us/previous-versions/windows/it-pro/windows-10/security/threat-protection/auditing/event-4649
  - https://learn.microsoft.com/en-us/previous-versions/windows/it-pro/windows-10/security/threat-protection/auditing/audit-other-logonlogoff-events
  - https://attack.mitre.org/techniques/T1550/003/
author: Forestall
date: 2026-11-16
tags:
  - attack.lateral-movement
  - attack.t1550.003
logsource:
  product: windows
  service: security
detection:
  selection:
    EventID: 4649
  condition: selection
falsepositives:
  - A misconfigured network device between client and server that sends the same packets repeatedly, and other network configuration or routing problems, which Microsoft names as causes of this event. They usually affect many clients behind the same device and keep producing events. Fix the device, or exclude the affected clients by the fields your collector extracts from the event.
  - The rule has no further conditions, so every event 4649 matches. Expect to add exclusions after the first occurrences have been explained.
level: medium

False positives

  • A misconfigured network device between client and server that sends the same packets repeatedly, and other network configuration or routing problems, which Microsoft names as causes of this event. They usually affect many clients behind the same device and keep producing events. Fix the device, or exclude the affected clients by the fields your collector extracts from the event.
  • The rule has no further conditions, so every event 4649 matches. Expect to add exclusions after the first occurrences have been explained.

Kerberos Ticket Request Rejected as a Replay

Detects a failed Kerberos ticket request (4768 or 4769) with result code 0x22, KRB_AP_ERR_REPEAT, which Microsoft describes as a request that is a replay: the KDC has seen the same authenticator before. Microsoft's monitoring recommendations for event 4768 name this code as a possible sign of an attack attempt. The events come from the Kerberos audit subcategories of the domain controllers, so this rule also works where Audit Other Logon/Logoff Events, which writes event 4649, is not enabled. Requires failure auditing in Audit Kerberos Authentication Service and Audit Kerberos Service Ticket Operations.

LevelMediumStatusExperimentalLog sourceWindows Security event log
title: Kerberos Ticket Request Rejected as a Replay
id: bfea5399-62b3-4c03-8496-1f57d98872ce
status: experimental
description: >-
  Detects a failed Kerberos ticket request (4768 or 4769) with result code
  0x22, KRB_AP_ERR_REPEAT, which Microsoft describes as a request that is a
  replay: the KDC has seen the same authenticator before. Microsoft's
  monitoring recommendations for event 4768 name this code as a possible sign
  of an attack attempt. The events come from the Kerberos audit subcategories
  of the domain controllers, so this rule also works where Audit Other
  Logon/Logoff Events, which writes event 4649, is not enabled. Requires
  failure auditing in Audit Kerberos Authentication Service and Audit Kerberos
  Service Ticket Operations.
references:
  - https://learn.microsoft.com/en-us/previous-versions/windows/it-pro/windows-10/security/threat-protection/auditing/event-4768
  - https://learn.microsoft.com/en-us/previous-versions/windows/it-pro/windows-10/security/threat-protection/auditing/event-4769
  - https://learn.microsoft.com/en-us/previous-versions/windows/it-pro/windows-10/security/threat-protection/auditing/event-4649
  - https://attack.mitre.org/techniques/T1550/003/
author: Forestall
date: 2026-11-18
tags:
  - attack.lateral-movement
  - attack.t1550.003
logsource:
  product: windows
  service: security
detection:
  selection:
    EventID:
      - 4768
      - 4769
    Status: '0x22'
  condition: selection
falsepositives:
  - Network devices that send the same packets more than once, which Microsoft names as a cause of replay detections. Many events for many clients behind the same device point to it.
level: medium

False positives

  • Network devices that send the same packets more than once, which Microsoft names as a cause of replay detections. Many events for many clients behind the same device point to it.

Triage

  • Read the account whose credentials were replayed, the workstation name and the request type in the event. One account on one device points to that device or to its network path.
  • Count the events. A steady stream that involves many clients behind the same network device points to a device that duplicates packets. A few events for a privileged account are the case to take seriously.
  • Ask the network team about recent changes, retransmissions and duplicated packets on the path between the client and the domain controller.
  • Compare the workstation named in the event with the device the account normally uses, and check that device for signs of compromise.
  • Look at what the account did around the same time, such as logons from new addresses and ticket requests for sensitive services (4624, 4768, 4769).

Response

Once the attack is confirmed:

  1. Find the system that sent the duplicate requests and isolate it. In a real replay that is a host or a network device on the path that the attacker controls.
  2. Reset the password of the account whose requests were replayed, and for a privileged account review everything it accessed since the first event.
  3. Assume that other authentication traffic on the same path was captured as well. Review the accounts that authenticated over it and watch for further 4649 events and for logons from the attacker's system.
  4. Remove the attacker's position by rebuilding the compromised host or correcting the network device, and confirm afterwards that the events have stopped.

Mitigation

  • Keep the Kerberos policy setting "Maximum tolerance for computer clock synchronization" at 5 minutes, the default and the value Microsoft recommends, and keep clocks synchronized so that nobody needs to raise it.
  • Protect the network path between clients and domain controllers. Segment the network, keep the devices on the path patched and under change control, and limit which hosts can see authentication traffic of other systems.
  • Require signing and channel binding on the services of domain controllers, such as LDAP signing, LDAP channel binding and SMB signing, so that a position on the path gives less.
  • Enable Audit Other Logon/Logoff Events on all domain controllers. Microsoft recommends success auditing there to track possible Kerberos replay attacks.

Frequently asked questions

Does event 4649 mean that an attacker got in?

No. The event is written when a domain controller rejects a request because it has already seen the same authenticator. It records an attempt that failed, or a duplicated packet that was never an attack. Microsoft names a misconfigured network device as a possible cause and still recommends investigating why the event was generated. The concern is what a real replay implies, namely that somebody can capture authentication traffic.

Why does a replay against a file server not produce this event?

Microsoft documents event 4649 as an event of domain controllers, written when a domain controller sends the KRB_AP_ERR_REPEAT response. A request replayed to another server is handled by that server and does not appear in the Security log of a domain controller. The rule therefore covers the Kerberos requests that domain controllers receive themselves.

Which fields of event 4649 can be used to filter the rule?

Microsoft documents the text of the event, with the account whose credentials were replayed, the workstation name, the request type and the process, but gives no example of the event and no field names for the collected event data. The rule therefore matches on the event ID alone. Once you have seen the event in your own environment, add exclusions on the fields your collector extracts, for example for a device that is known to duplicate packets.

References

We respect your privacy

We use cookies to keep this site secure and working properly. With your permission, we also use optional cookies to understand usage and improve the experience. Cookie Policy

You can change your choice at any time.