Once an attacker holds Domain Admin or the krbtgt secret, every account, ticket and machine in the domain has to be treated as untrusted until you prove otherwise. Generic ransomware runbooks stop at isolating hosts and restoring data, but a domain compromise needs a different sequence: preserve evidence from domain controllers, cut the attacker off the directory itself, reset krbtgt twice in the right order, rotate trust and service secrets, and decide early whether the domain can be cleaned or must be rebuilt. This template gives incident commanders, identity engineers and SOC leads a playbook they can adopt with [Organisation] placeholders. It covers detection to containment and eradication decisions in depth and hands the restore mechanics to the Identity Recovery Guide.
How to use this template
- Adopt, then rehearse: replace every [placeholder], attach your tier 0 inventory and contact sheet, and run the playbook in a tabletop before you need it. Untested krbtgt and trust resets are where real responses stall.
- Read with the parent plan: this playbook sits under the [Organisation] incident response plan. Severity, legal hold and notification rules come from there. See the Incident Response Plan Template.
- Use the decision points: sections marked Decision need a named approver and a logged rationale. Keep a decision log from the first hour.
- Assume out-of-band working: if the domain is compromised, so are domain-joined email, chat and ticketing. Plan a separate channel and clean admin workstations in advance.
- Scope: on-premises Active Directory Domain Services, including hybrid links to Microsoft Entra ID. Pure cloud tenant takeover is covered in the Identity Provider Tenant Compromise Playbook.
1. Scope and triggers
This playbook applies when [Organisation] has evidence, or a credible suspicion, that an attacker has obtained control over one or more domains in the [Forest name] forest. Invoke it on any of the following triggers:
- Replication abuse: directory replication requests (DCSync, ATT&CK T1003.006) from a host that is not a domain controller, usually seen as event 4662 on a DC with replication rights exercised by a user or computer account.
- NTDS.dit access: volume shadow copy creation, ntdsutil or copies of the NTDS database on a DC (T1003.003). The joint ASD ACSC, CISA, NSA, NCSC-UK, CCCS and NCSC-NZ guidance lists events 1102, 4103, 4104, 4656, 4663, 4688 and 8222 as relevant.
- Forged tickets: Golden Ticket (T1558.001) or Silver Ticket (T1558.002) indicators, such as Kerberos service tickets with no matching TGT request, implausible ticket lifetimes, or alerts from your identity threat detection tooling.
- Privileged group changes: unexpected additions to Domain Admins, Enterprise Admins, Schema Admins, Administrators or other protected groups, or changes to AdminSDHolder.
- Domain-wide changes: new or modified Group Policy Objects that push scheduled tasks or scripts (T1484.001), a rogue domain controller registration (T1207), or SID History added to accounts (T1134.005).
- External notice: a government, vendor or retainer partner tells you your domain credentials or hashes are circulating.
- Ransomware with domain reach: encryption or deployment through GPO or management tools across many hosts at once.
Out of scope: a single compromised standard user with no path to tier 0 (use the Identity Incident Response Playbook) and directory failures with no attacker (use the recovery guide).
2. Roles and contacts
| Role | Responsibilities | Named lead / deputy |
|---|---|---|
| Incident commander | Declares the incident, owns the decision log, approves containment, krbtgt reset and rebuild decisions | [Name] / [Name] |
| Identity lead (AD engineering) | Executes privileged resets, krbtgt and trust resets, GPO review, replication health checks | [Name] / [Name] |
| Forensics lead | Collects DC memory and disk images, preserves logs, maintains chain of custody | [Name] or [IR retainer] |
| SOC lead | Monitoring for re-entry, hunting for persistence, alert tuning during the response | [Name] |
| Infrastructure and cloud lead | Network isolation, Entra Connect and federation servers, backups, hypervisors hosting DCs | [Name] |
| Legal, privacy and communications | Privilege, regulator and customer notification, internal messaging | [Name] |
| Executive sponsor | Business downtime acceptance, budget, ransom policy | [Name] |
Contacts kept offline: [IR retainer hotline], [Microsoft or platform support case route], [cyber insurer], [national CSIRT or regulator], [out-of-band chat bridge]. Store a printed copy and a copy outside the domain.
3. Detection and triage
- Confirm the signal (target: [60] minutes). Pull the triggering events from the DC security logs or your SIEM, not from a domain-joined analyst workstation that may be watched. Record the source host, account and time.
- Identify the tier 0 accounts touched. List every account with replication rights, membership of protected groups, control of DCs, AD CS, Entra Connect, AD FS, backup systems and the hypervisors that host DCs. See the Tiered Administration and Enterprise Access Model Guide.
- Establish the earliest evidence. Look back for the first DCSync, first privileged group change and first new GPO. The attacker's dwell time sets how far back backups may be tainted.
- Scope across the forest. Check every domain and every trust (T1482 discovery often precedes cross-domain movement). A compromise of a child domain can reach the forest root through SID History and trust abuse.
- Check the hybrid edge. Review Entra Connect servers, the directory synchronisation account and any on-premises admin account with a cloud admin counterpart. In the Storm-0501 hybrid cloud attacks, Domain Admin access led to sync account credential theft and a cloud Global Administrator takeover.
- Classify severity. Use the table below and record the classification in the decision log.
| Evidence | What the attacker can do | Severity and path |
|---|---|---|
| Domain Admin session or password, no secret extraction seen | Act as admin until reset | High: privileged reset, hunt for extraction |
| DCSync or NTDS.dit copy confirmed | Every password hash, including krbtgt, is exposed; Golden Tickets possible | Critical: full credential reset, krbtgt double reset, rebuild assessment |
| Golden Ticket or Golden Certificate indicators | Forge identities without passwords | Critical: krbtgt reset and AD CS CA key review |
| Malicious GPO, rogue DC or schema change | Domain-wide code execution or directory tampering | Critical: isolate, consider rebuild |
| DPAPI domain backup key extracted | Decrypt user DPAPI secrets indefinitely | Critical: the joint guidance says Microsoft's only supported recovery is a new domain |
4. Containment
Containment in a domain compromise is a trade-off: move too slowly and the attacker deploys ransomware through GPO; move too noisily and they switch to persistence you have not found. Decision: the incident commander chooses between rapid containment (stop active harm now) and coordinated containment (observe briefly, then act on everything in one window). Active encryption or data theft always means rapid containment.
- Preserve first, in parallel. Capture memory and security, directory service and PowerShell logs from at least one affected DC and the attacker's pivot hosts before resets overwrite evidence. See the Memory Forensics Guide.
- Restrict access to DCs. Block inbound administration to DCs except from designated clean privileged access workstations; block DC outbound internet access entirely.
- Disable, then reset, attacker-used accounts. Disable accounts the attacker created; reset the passwords of compromised admin accounts twice where hash reuse is suspected; remove unexpected protected group members.
- Break hybrid propagation. Put Entra Connect into staging mode or stop the sync service, reset the sync account, and revoke sessions for cloud counterparts of on-premises admins.
- Freeze change. Disable or unlink suspicious GPOs and stop deployment tooling (SCCM, RMM, PsExec-style services) that the attacker could use to push payloads.
- Isolate the network. Segment DCs and tier 0 systems from user and server networks; for confirmed ransomware staging, cut site links or internet egress per the Ransomware Playbook Template.
- Protect backups. Take backup consoles offline from the domain and confirm immutable copies exist before anyone touches restore.
5. Eradication
Eradication removes every way back in. Order matters: reset krbtgt only after you control the DCs, or the attacker simply extracts the new key.
Decision: clean in place or rebuild
| Factor | Clean in place is viable | Rebuild (new forest or clean-room restore) is safer |
|---|---|---|
| Dwell time and visibility | Short, well-logged, timeline complete | Long, gaps in DC logging, unknown start |
| Persistence found | Limited to accounts and GPOs you can enumerate | Schema changes, rogue DC, DSRM or authentication package tampering (T1556) |
| Secrets lost | Admin passwords only | All hashes plus DPAPI backup key or AD CS CA private key |
| Size and dependencies | Small number of domains and apps | Large, but compromise breadth outweighs migration cost |
Eradication steps (clean in place)
- Confirm replication health across all writable DCs before any krbtgt change. CISA countermeasure CM0050 makes this a prerequisite.
- Reset krbtgt, first time. Use Microsoft's reset procedure or the New-KrbtgtKeys script, separately for writable DCs and each RODC krbtgt_<number> account. Let it replicate everywhere.
- Wait at least 10 hours, the default maximum user and service ticket lifetime, or longer if [Organisation] has changed those policies. Microsoft and CISA give the same interval.
- Reset krbtgt, second time. The account keeps a two-password history, so only the second reset invalidates tickets forged with the stolen key.
- Reset trust passwords for each inter-domain and external trust using the NetDom /resetOneSide method Microsoft documents for breach-related recovery.
- Reset every privileged and service account, then all user and computer accounts if NTDS.dit was taken. Rotate gMSA keys, AD FS signing certificates and the Entra Connect account.
- Remove persistence: SID History on accounts that do not need it, unexpected ACLs on the domain root and AdminSDHolder, rogue SPNs, malicious GPOs and scheduled tasks, and unknown certificate templates. See the Active Directory ACL and Delegation Hygiene Guide.
- Rebuild the pivot hosts rather than cleaning them, including any tier 0 server the attacker logged on to.
Rollback: the krbtgt reset cannot be undone, so the rollback plan is operational. CISA warns that every application relying on Kerberos tickets is affected and non-Windows clients may fail until old tickets expire. Schedule resets in a change window with application owners on call, and have a tested list of services to restart.
6. Recovery
- Restore path: if rebuilding, follow the clean-room forest recovery sequence in the Identity Recovery Guide, using a backup from before the earliest evidence and verifying its integrity before use.
- Return in rings: bring back DCs, then tier 0 services (AD CS, Entra Connect, backup), then tier 1 servers, then user devices, each with monitoring in place before reconnection.
- Re-establish hybrid sync carefully: preview changes in staging mode before exporting to Entra ID, so a tainted on-premises state does not overwrite clean cloud objects.
- Raise the bar before reopening: enforce tiered administration, remove standing Domain Admin membership, and add canary accounts. See the Active Directory and Entra ID Hardening Guide.
- Enhanced monitoring: keep a 30 to 90 day watch on the events in the table below, tuned for the indicators found.
| Event ID (log) | What to watch for after recovery |
|---|---|
| 4662 (Security, DC) | Replication rights used by an account that is not a DC |
| 4769 / 4768 (Security, DC) | Service tickets for high value SPNs, RC4 ticket requests, TGT anomalies |
| 4728, 4732, 4756 (Security, DC) | Members added to privileged security groups |
| 5136 (Security, DC) | Directory object changes to GPOs, AdminSDHolder, domain root ACLs |
| 4738 / 4724 (Security, DC) | Account changes and password resets on tier 0 accounts |
| 1102 (Security), 4104 (PowerShell) | Log clearing and script block activity on DCs |
7. Communication and notification
- Internal: use the out-of-band channel only. Tell staff to expect password resets and service interruptions; give a single source of truth for status.
- Legal and regulatory: a domain compromise usually means access to personal data stores, so legal must assess duties under [GDPR Article 33 / sector regulator / contract] within the parent plan's timelines. Record when [Organisation] became aware.
- Partners and trusts: notify organisations with forest trusts or federation with you, because their trust passwords and federated tokens are part of your eradication.
- Insurer and retainer: notify per policy terms before engaging new third parties.
- Do not: announce containment before krbtgt has been reset twice and persistence removed; early "all clear" messages are a common source of re-compromise.
8. Post-incident review
- Timeline: initial access, first privileged access, first secret extraction, detection, containment, each krbtgt reset and return to service.
- Root cause: the path to tier 0, usually a reusable admin credential, an over-privileged service account, Kerberoastable SPN accounts or a misconfigured AD CS template. See the Kerberoasting and AS-REP Roasting Defence Guide.
- Control gaps: map each gap to an owner and a date. Note that stolen domain hashes keep circulating long after the incident, as the Cisco AD credentials leak showed when hashes from 2022 resurfaced in 2025.
- Playbook update: revise this template, the contact sheet and the tier 0 inventory within [30] days.
Implementation guidance
- Build the tier 0 inventory first: DCs, AD CS, Entra Connect, AD FS, backup and hypervisor platforms, plus every account and group that controls them.
- Pre-stage clean tooling: privileged access workstations, offline copies of reset scripts, and an out-of-band collaboration space.
- Rehearse the krbtgt reset in a lab and then in production on a routine schedule; the joint guidance recommends changing the krbtgt password every 12 months as well as after suspected compromise.
- Run a tabletop using the Storm-0501 pattern, including the clean versus rebuild decision.
- Measure readiness with the metrics below and review quarterly.
| Metric | Target example |
|---|---|
| Time from DCSync or NTDS alert to incident declaration | Under [1] hour |
| Time from declaration to first krbtgt reset | Under [24] hours once DCs are controlled |
| Tier 0 accounts with standing privilege | Falling to [agreed minimum] |
| Days since last krbtgt rotation | Under 365 |
| Privileged and service accounts reset after an incident | 100 percent, verified by pwdLastSet |
Framework mapping
| Framework | Reference | Playbook section |
|---|---|---|
| NIST CSF 2.0 | RS.MA-02, RS.AN-03, RS.MI-01, RS.MI-02, RC.RP-01, RC.RP-03, RS.CO-02 | Triage, containment, eradication, recovery, communication |
| NIST SP 800-53 Rev. 5 | IR-4, IR-5, IR-6, IR-8, AC-2, AC-6, IA-5, CP-10 | Whole playbook |
| CIS Controls v8.1 | Control 5 (Account Management), Control 6 (Access Control Management), Control 17 (Incident Response Management) | Containment and preparation |
| MITRE ATT&CK v20 | T1003.006, T1003.003, T1558.001, T1558.002, T1484.001, T1134.005, T1207, T1556, T1078.002 | Triggers and eradication |
Standards and references
- ASD ACSC, CISA, NSA and partners: Detecting and Mitigating Active Directory Compromises
- CISA: Reset Kerberos Ticket Granting Ticket (KRBTGT) Password Twice (CM0050)
- Microsoft Learn: AD Forest Recovery, Reset the krbtgt password
- Microsoft Learn: AD Forest Recovery, Reset a trust password
- NIST: SP 800-61 Rev. 3, Incident Response Recommendations and Considerations for Cybersecurity Risk Management
- MITRE ATT&CK: OS Credential Dumping: DCSync
- MITRE ATT&CK: Steal or Forge Kerberos Tickets: Golden Ticket
Related NHI Mgmt Group resources: Golden Ticket, Silver Ticket and DCSync Guide · Identity Recovery Guide · Top 10 Active Directory Security Risks · BloodHound and Active Directory Attack Path Guide