Most Active Directory compromises leave evidence on the domain controllers, but default audit settings either do not record it or bury it among millions of routine events. This guide is a domain controller monitoring reference: which audit subcategories and SACLs to enable, which event IDs matter (4768, 4769, 4771, 4776, 4662, 5136, 4728 and others), which fields inside them carry the signal, and how to turn them into detections for Kerberoasting, DCSync, privileged group changes, SID History and directory persistence. It assumes you already know how Windows event logging works, which the Windows Event Log Guide covers, and it goes deeper on the DC-specific telemetry than the ITDR Guide. It is for detection engineers, SOC analysts and AD administrators.
This is an independent guide from NHI Mgmt Group, based on Microsoft's public documentation as of November 2026. It is not sponsored or reviewed by Microsoft. Product features change, so confirm settings against Microsoft's current documentation.
Key takeaways
- Directory telemetry needs SACLs as well as audit policy. Events 4662 and 5136 are generated only when the object carries a matching SACL entry, so enabling the subcategory alone records almost nothing.
- Read the fields, not just the event ID. A 4769 is routine; a 4769 with ticket encryption type 0x17 (RC4) for a service account on an AES-only estate is a Kerberoasting signal.
- Some events justify an alert on a single occurrence. Microsoft rates 4719 audit policy changed, 4765 and 4766 SID History, 4794 DSRM password set and 4964 special groups logon as high criticality.
- Collect from every DC, plus the tier 0 servers around them. AD CS, AD FS and Entra Connect servers hold equivalent power and need their own events.
- Baseline before alerting. Group membership, Kerberos and LDAP events are noisy; know who normally creates accounts, which accounts request which tickets, and which hosts replicate.
What to collect: audit policy and SACLs
Configure advanced audit policy through the Default Domain Controllers Policy, and set "Audit: Force audit policy subcategory settings to override audit policy category settings" so legacy category settings do not overwrite it. Microsoft's audit policy recommendations give a stronger baseline of Success and Failure for the Account Logon, Account Management, Logon/Logoff and Policy Change rows below; the DS Access rows are needed for the 4662 and 5136 events that Microsoft Defender for Identity also requires:
| Category | Subcategory | Setting | Key events |
|---|---|---|---|
| Account Logon | Credential Validation | Success, Failure | 4776 NTLM credential validation |
| Account Logon | Kerberos Authentication Service | Success, Failure | 4768 TGT requested, 4771 pre-authentication failed |
| Account Logon | Kerberos Service Ticket Operations | Success, Failure | 4769 service ticket requested |
| Account Management | User, Computer and Security Group Management | Success, Failure | 4720, 4726, 4738, 4741, 4743, 4728, 4732, 4756 and related |
| DS Access | Directory Service Access | Success, Failure | 4662 operation performed on an object |
| DS Access | Directory Service Changes | Success | 5136 modified, 5137 created, 5139 moved, 5141 deleted |
| Logon/Logoff | Logon, Special Logon | Success, Failure | 4624, 4625, 4672, 4964 |
| Policy Change | Audit Policy Change, Authentication Policy Change | Success, Failure | 4719, 4713, 4739, 4706 |
Then add SACLs. Microsoft Defender for Identity's collection guidance, a useful vendor-neutral baseline, sets a Success audit entry for Everyone on the domain object for descendant User, Group, Computer, msDS-GroupManagedServiceAccount, msDS-ManagedServiceAccount and (with a Windows Server 2025 DC) msDS-DelegatedManagedServiceAccount objects, with write permissions only. Add entries on the domain root (domainDNS), AdminSDHolder, Group Policy containers and the configuration partition where AD CS templates live. Keep read auditing narrow: auditing all reads on all objects can overwhelm DCs.
For NTLM visibility (event 8004 in the NTLM operational log), set the three "Network security: Restrict NTLM" audit options on domain controllers: outgoing traffic to Audit all, NTLM authentication in this domain to Enable all, and incoming traffic to Enable auditing for all accounts. This also supports the NTLM Deprecation and Elimination Guide.
Event ID reference for domain controllers
| Event | Meaning | Fields that carry the signal | Typical use |
|---|---|---|---|
| 4768 | Kerberos TGT requested | Pre-Authentication Type (0 means no pre-authentication), Ticket Encryption Type, Client Address, Result Code (0x6 unknown user, 0x12 revoked) | AS-REP roasting exposure, enumeration, logons from unexpected sources |
| 4769 | Kerberos service ticket requested | Service Name, Ticket Encryption Type (0x17 RC4), Failure Code, Transited Services | Kerberoasting, unusual service access, delegation abuse |
| 4771 | Kerberos pre-authentication failed | Failure Code 0x18 (wrong password), Client Address | Password guessing and spraying against Kerberos |
| 4776 | DC attempted to validate NTLM credentials | Workstation name, error code | NTLM spraying, legacy authentication use |
| 4662 | Operation performed on an AD object | Object Type, Object Name, Accesses (Control Access, WRITE_DAC, WRITE_OWNER), Properties GUIDs | DCSync replication rights use, ACL tampering, sensitive reads |
| 5136 | Directory object modified | DN, Class, LDAP Display Name, Operation Type (Value Added), Correlation ID | GPO, AdminSDHolder, delegation and SPN changes |
| 4728, 4732, 4756 | Member added to global, local or universal security group | Target group, member, subject | Privileged group changes |
| 4720, 4726, 4738 | User created, deleted, changed | Changed attributes, subject | Account persistence, UAC flag changes |
| 4741, 4743 | Computer account created, deleted | Subject, new account name | Rogue computer accounts used for delegation or relay attacks |
| 4765, 4766 | SID History added, or add attempt failed | Target account, SID | SID History injection (high criticality) |
| 4794 | Attempt to set DSRM administrator password | Subject | DC backdoor persistence (high criticality) |
| 4719, 1102 | Audit policy changed; audit log cleared | Subject, changed subcategory | Defence impairment (4719 high criticality) |
| 4706, 4713, 4739 | Trust created; Kerberos policy changed; domain policy changed | Subject, new values | Trust abuse and policy weakening |
| 4964 | Special groups assigned to a new logon | Account, logon ID | Domain admins logging on outside tier 0 (requires configured special groups) |
| 7045 | Service installed (System log) | Service name, image path | Remote execution on DCs |
Microsoft's events-to-monitor appendix rates 4719, 4765, 4766, 4794 and 4964 as high criticality, meaning a single occurrence should be investigated, and notes that medium and low events matter only when unexpected or above baseline. Since the January 2025 security update, Windows Server 2016 and later add fields to 4768 and 4769, including msDS-SupportedEncryptionTypes, available keys and ticket hashes, which make RC4 analysis easier; update parsers accordingly.
Turning events into detections
Kerberos abuse
- Kerberoasting (T1558.003): alert on 4769 where Ticket Encryption Type is 0x17 for a service account that supports AES, and on one account requesting tickets for many distinct service names in a short window. Microsoft's guidance is to alert on any type other than 0x11 or 0x12 on modern systems. See the Kerberoasting and AS-REP Roasting Defence Guide.
- AS-REP roasting exposure (T1558.004): any 4768 with Pre-Authentication Type 0 means an account has "Do not require Kerberos preauthentication" set. Treat each as a posture finding.
- Forged tickets (T1558.001, T1558.002): forged tickets often lack a matching 4768 for the same logon, or carry unusual lifetimes or encryption types. Correlate 4769 and 4624 with prior 4768 per account and host. The Golden Ticket, Silver Ticket and DCSync Guide covers these patterns.
- Honey accounts: a decoy account with an SPN and no legitimate use turns any 4769 for it into a high-fidelity alert.
Replication and credential theft
- DCSync (T1003.006): alert on 4662 where Properties contains the DS-Replication-Get-Changes (
1131f6aa-9c07-11d1-f79f-00c04fc2dcd2) or DS-Replication-Get-Changes-All (1131f6ad-9c07-11d1-f79f-00c04fc2dcd2) extended rights and the subject is not a domain controller computer account or an approved sync account such as the Entra Connect directory synchronisation account. - NTDS extraction (T1003.003): watch for shadow copy creation, ntdsutil use and service installs (7045) on DCs from endpoint telemetry; the security log alone will not show it.
- Rogue DC (T1207): new server objects or nTDSDSA objects in the configuration partition (5137) and replication from hosts that are not DCs.
Privilege and persistence changes
- Privileged groups: any 4728, 4732 or 4756 for Domain Admins, Enterprise Admins, Schema Admins, Administrators, Account Operators, Backup Operators or DnsAdmins outside an approved change. ATT&CK maps this to T1098 Account Manipulation.
- AdminSDHolder and ACLs: 5136 on
CN=AdminSDHolder,CN=Systemor 4662 with WRITE_DAC or WRITE_OWNER on domainDNS. See the AdminSDHolder, SDProp and Protected Groups Guide. - Delegation and SPNs: 5136 where LDAP Display Name is
msDS-AllowedToActOnBehalfOfOtherIdentity,msDS-AllowedToDelegateTo,servicePrincipalNameoruserAccountControlon privileged or service accounts. - Group Policy (T1484.001): 5136 on groupPolicyContainer objects, especially
gPCFileSysPathandgPCMachineExtensionNames, linked to tier 0 OUs. - SID History (T1134.005): any 4765 or 4766 outside a documented migration.
- Machine accounts: 4741 by a non-admin user, which indicates the default ms-DS-MachineAccountQuota is being used, a common step in relay and delegation attacks.
Password attacks
- Spraying (T1110.003): many 4771 with 0x18 or 4776 failures across many accounts from one source, with few failures per account. See the MITRE ATT&CK T1110 Brute Force Guide.
- Enumeration: bursts of 4768 with result 0x6 (unknown user).
Beyond the DCs: tier 0 servers and identity platforms
| System | Events to collect | Why |
|---|---|---|
| AD CS issuing CAs | 4870, 4882, 4885, 4887, 4888, 4890, 4896 | Certificate issuance and CA configuration changes; certificate abuse (T1649) gives domain persistence. See the AD CS Security Guide. |
| AD FS servers | 1202, 1203, 4624, 4625 | Federation authentication success and failure; token signing abuse |
| Entra Connect servers | 4624 plus endpoint telemetry | The sync account can reset hybrid passwords; see the Microsoft Entra Connect and Cloud Sync Hardening Guide |
| PAM and backup servers | Logon, service install and process events | They hold credentials or copies of NTDS |
Where identity fits: service and machine accounts
Service accounts, gMSAs and computer accounts generate most DC events and are the accounts attackers most want. Monitoring works far better when these identities are known: maintain a list of service accounts with their expected hosts and services, alert on interactive logons (4624 logon types 2 and 10) by service accounts, and on 4769 requests for service accounts from hosts outside their baseline. Moving service accounts to gMSA removes most Kerberoasting exposure because their passwords are long and rotated automatically. See the Group Managed Service Accounts (gMSA) Guide and the Service Account Security Guide.
Decision guidance: SIEM rules, ITDR or both
- Small estates (a handful of DCs, no dedicated detection team): collect the events in this guide into the SIEM you already have and enable only the single-event high-criticality alerts plus privileged group changes. That set is low-noise and catches the most damaging actions.
- Mid-sized estates: add behavioural Kerberos and spraying rules once baselines exist, and decide whether an ITDR product's packaged detections are cheaper than writing and testing your own.
- Large or multi-forest estates: run both. ITDR sensors on DCs see network-level Kerberos and LDAP behaviour that the security log does not, while raw events in the SIEM give you correlation with endpoint and cloud data and an independent record if a sensor is disabled. See the ITDR Buyer's Guide.
- Retention: keep DC security events for at least the period your incident response needs to trace a long intrusion; local logs roll over in hours on busy DCs, so central retention is what counts.
Rolling out DC monitoring
- Inventory: list every writable and read-only DC, AD CS, AD FS, Entra Connect and PAM server. Missing one DC means missing a share of authentications.
- Size the volume: enable the subcategories on one DC for a week and measure events per second. Kerberos events dominate; plan SIEM licensing and retention before rollout.
- Enable audit policy by GPO: apply to the Domain Controllers OU, confirm with
auditpol /get /category:*on each DC, and check that the security log maximum size holds at least several days locally. - Add SACLs in stages: start with write auditing on the domain object descendants, AdminSDHolder and GPO containers, then add replication extended rights. Test on a non-production forest first; roll back by removing the audit entries, which does not affect access.
- Forward and parse: use Windows Event Forwarding or an agent, and parse the XML fields (TicketEncryptionType, PreAuthType, Properties, AttributeLDAPDisplayName) rather than the rendered message.
- Baseline for two to four weeks: record normal group change owners, ticket requesters, replication partners and service account source hosts.
- Enable detections in tiers: single-event high-criticality alerts first (4719, 4765, 4794, DCSync from non-DCs, privileged group adds), then behavioural rules once baselines exist.
- Test: validate each rule with authorised, controlled test activity in a lab or with a purple team, and record expected alert IDs.
Measuring coverage
- Percentage of DCs and tier 0 servers forwarding all required events, checked daily (target 100 percent).
- Gap in minutes between event time on the DC and arrival in the SIEM.
- Number of service accounts still issuing RC4 service tickets, trending to zero.
- Accounts with pre-authentication disabled (4768 type 0), trending to zero.
- Privileged group changes matched to an approved change record (target 100 percent).
- Detections tested in the last quarter as a share of the DC rule set.
Framework mapping
| Area | NIST SP 800-53 Rev. 5 | NIST CSF 2.0 | MITRE |
|---|---|---|---|
| Audit policy and SACLs | AU-2, AU-12 | PR.PS-04 | D3FEND D3-DAM Domain Account Monitoring |
| Review and correlation | AU-6, SI-4 | DE.CM-03, DE.AE-03 | T1558, T1003.006, T1110.003 |
| Log protection and retention | AU-9, AU-11 | PR.PS-04 | Defence impairment (TA0112) |
| Privileged changes | AC-2, AU-6 | DE.CM-03 | T1098, T1484.001, T1134.005 |
CIS Controls v8.1 Control 8 Audit Log Management covers collection and retention; see the CIS Control 8 Audit Log Management Guide.
Common failure patterns
- Subcategory on, SACL missing: Directory Service Changes is enabled but no SACLs exist, so 5136 never fires for AdminSDHolder or GPO edits.
- Legacy policy overrides: category-level audit settings in another GPO silently overwrite subcategory settings because the force option is not set.
- Some DCs not forwarding: a DC in a branch site or a new DC built after the forwarding rollout sends nothing, and attackers authenticate there.
- Parsing the message text: detections built on rendered strings break when Microsoft adds fields, as it did to 4768 and 4769 in 2025.
- Alerting on everything: every 4728 pages the SOC, so analysts learn to ignore group change alerts, including the one that matters.
- Not acting after a hash dump: hashes, including krbtgt, can resurface long after an intrusion. Detection is wasted if the response does not rotate every exposed account, krbtgt twice.
Practitioner checklist
- Apply the stronger advanced audit baseline to the Domain Controllers OU and set the force subcategory option.
- Add write SACLs on domain object descendants, AdminSDHolder, GPO containers and AD CS templates.
- Enable NTLM auditing on DCs to collect event 8004.
- Forward security and System logs from every DC, AD CS, AD FS and Entra Connect server, and alert when one stops.
- Parse XML fields such as TicketEncryptionType, PreAuthType, Properties and AttributeLDAPDisplayName.
- Alert on single occurrences of 4719, 4765, 4766, 4794 and DCSync rights used by non-DC accounts.
- Alert on privileged group membership changes without a change record.
- Track RC4 service tickets and accounts without pre-authentication as posture metrics.
- Deploy honey service accounts with SPNs and alert on any ticket request for them.
- Test each detection with authorised activity at least quarterly.
Standards and references
- Microsoft Learn: Appendix L Events to Monitor
- Microsoft Learn: Audit Policy Recommendations
- Microsoft Learn: 4769 A Kerberos service ticket was requested
- Microsoft Learn: 4662 An operation was performed on an object
- Microsoft Learn: 5136 A directory service object was modified
- Microsoft Learn: Event collection with Microsoft Defender for Identity
- MITRE ATT&CK: Steal or Forge Kerberos Tickets (T1558)
- NIST: SP 800-53 Rev. 5
Related NHI Mgmt Group resources: ITDR Guide · Windows Event Log Guide · Active Directory and Entra ID Hardening Guide · MITRE ATT&CK T1558 Steal or Forge Kerberos Tickets Guide · Active Directory Domain Compromise Response Playbook