Post

Active Directory Internals: Structure, Replication and Security

How Active Directory Domain Services is structured and operated: forests, domains, OUs, directory partitions, domain controllers, LDAP, the Global Catalog, replication, FSMO roles, Group Policy, SYSVOL, attack paths, telemetry and hardening.

Active Directory Internals: Structure, Replication and Security

Introduction

Active Directory Domain Services (AD DS) is often introduced as “the server that authenticates domain users.” Authentication is only one part of it. AD DS is a distributed, replicated directory and security system that stores identities, computers, groups, policies, service metadata, trust relationships and access-control information, then makes that data available to clients and applications through several protocols.

Three services sit at its core:

  • Directory services store and retrieve structured objects;
  • authentication proves the identity of users, computers and services;
  • authorization data supplies SIDs and group memberships from which Windows builds access tokens.
flowchart LR
    AD[Active Directory Domain Services] --> DS[Directory services]
    AD --> AU[Authentication]
    AD --> AZ[Authorization data]

    DS --> LDAP[LDAP queries and updates]
    DS --> REP[Multi-master replication]
    DS --> GPO[Group Policy and configuration]

    AU --> K[Kerberos]
    AU --> N[NTLM fallback and compatibility]

    AZ --> SID[User and group SIDs]
    SID --> TOK[Windows access token]

Figure 1 — AD DS is the identity control plane: directory data, authentication and authorization information are related, but they are not the same service.

This article follows the path from the logical hierarchy down to the database and replication engine, then back up through authentication, Group Policy, telemetry and defensive engineering. The Kerberos, NTLM and Windows identity articles cover those subsystems in greater depth.

The logical structure

Forest, tree, domain and OU

The logical hierarchy is easy to draw but easy to misunderstand:

flowchart TB
    F[Forest: example.com]
    F --> T1[Tree: example.com]
    F --> T2[Tree: fabrikam.net]

    T1 --> D1[Domain: example.com]
    T1 --> D2[Domain: eu.example.com]
    T1 --> D3[Domain: lab.example.com]
    T2 --> D4[Domain: fabrikam.net]

    D2 --> OU1[OU: Workstations]
    D2 --> OU2[OU: Servers]
    OU2 --> OU3[OU: Tier-0]

    S1[Site: Rome<br/>IP subnets] -. may contain DCs<br/>from several domains .-> D1
    S1 -.-> D2

Figure 2 — Forest, trees, domains and OUs form the logical hierarchy. Sites and subnets describe the physical network topology and cut across it.

Level What it represents What it does not represent
Forest One schema, one configuration partition, one Global Catalog namespace and a set of domains connected by automatic two-way transitive trusts Merely a DNS suffix or a collection of unrelated domains
Tree One or more domains in a contiguous DNS namespace, such as example.com and eu.example.com A separately administered database object with its own security policy
Domain A directory partition, DNS namespace, authentication scope and unit of administration and replication A strong security boundary against forest-level administrators
OU A container used for delegation and Group Policy scope inside one domain A security principal or an isolation boundary by itself
Site One or more well-connected IP subnets used for DC location, service affinity and replication topology A child of a domain or an authorization container

Microsoft describes the forest as the top-level logical and administrative boundary. Domains partition data and administration, but every domain shares the forest schema and configuration, and forest-level administrators can ultimately affect all of them.1

Why OUs exist

An organizational unit is a container for users, groups, computers and nested OUs. Its two primary design purposes are:

  1. delegation: its ACL can grant a help-desk group narrowly scoped rights, such as resetting passwords only inside one subtree;
  2. policy scope: GPOs can be linked to the OU so that settings apply to its user or computer objects.

OUs should follow administrative and policy requirements, not simply reproduce an organization chart. A department name may change frequently, while the delegation boundary between workstation administrators and server administrators is a durable security requirement. OU delegation reduces the need for Domain Admin membership, but does not isolate an OU from forest or domain service administrators.2

The directory object model

Objects, attributes and schema

AD DS stores typed objects. A user, computer, group, OU, site, GPO container and trust are instances of schema classes. The schema naming context defines which classes exist, which attributes each class may or must contain and how they are related.

Every object has multiple forms of identity:

Identifier Example Stability and use
Relative distinguished name (RDN) CN=Alice Identifies the object inside its parent container
Distinguished name (DN) CN=Alice,OU=Engineering,DC=example,DC=com Unique location in the forest; changes after rename or move
objectGUID 128-bit GUID Stable directory-object identity across rename and move
objectSid S-1-5-21-...-1107 Security identity for a user, computer or security group
UPN alice@example.com User logon name; administratively changeable
sAMAccountName alice Legacy-compatible account name, unique within its domain

Applications should not treat the DN as an immutable identifier. Moving Alice to another OU changes the DN but preserves the objectGUID. Security descriptors use SIDs because authorization must also survive ordinary renames and moves.3

Security descriptors are directory data

AD objects are securable objects. Their security descriptors contain an owner, DACL and SACL using the same fundamental access-control model as other Windows objects. The rights are directory-specific: read or write an attribute, create or delete a child class, modify permissions, change a password or invoke an extended right.

This is why an account can become effectively privileged without joining Domain Admins. An inherited ACE on an OU, control of a group, write access to a service account’s attributes or a replication extended right can be enough to cross a major security boundary. The visible group-membership list is only one part of the authorization graph.

Directory partitions and NTDS.dit

Naming contexts

The directory is divided into partitions, also called naming contexts (NCs). Each partition is a contiguous subtree with its own replication scope and metadata.4

flowchart TB
    DB[(NTDS.dit on one domain controller)]
    DB --> S[Schema NC<br/>forest-wide]
    DB --> C[Configuration NC<br/>forest-wide]
    DB --> D[Domain NC<br/>full local-domain replica]
    DB --> A[Application NCs<br/>selected replica set]
    DB --> GC[GC partial replicas<br/>other forest domains,<br/>if this DC is a GC]

Figure 3 — NTDS.dit is one local database instance containing several naming contexts. Replication is scoped per partition, not as a copy of the whole file.

Partition Main content Default replica scope
Schema Class and attribute definitions Every DC in the forest
Configuration Sites, services, partitions, replication topology and forest configuration Every DC in the forest
Domain Users, groups, computers, OUs and domain-local configuration Every writable DC in that domain
Application Application-specific directory data, commonly AD-integrated DNS zones Explicitly selected DCs

NTDS.dit is the local Extensible Storage Engine (ESE) database file. ESE transaction logs and a checkpoint support durability and crash recovery; operators should never manipulate these files as if they were ordinary text or copy-based configuration.5 The database stores far more than password material, but it contains credential-derived secrets and the entire authorization graph for its replicated partitions, making every writable DC a Tier-0 asset.

The Global Catalog

A Global Catalog (GC) is a role enabled on a domain controller. It stores:

  • full writable replicas of schema and configuration;
  • a full writable replica of its own domain partition;
  • a partial, read-only replica of every other domain in the forest.

The partial attribute set contains every object but only forest-search-relevant attributes. This lets a client search across domains without contacting a DC in each domain. The GC also participates in resolving UPNs and universal group membership during authentication. LDAP GC traffic normally uses TCP 3268, or 3269 with TLS; ordinary LDAP uses 389, or 636 for LDAPS.67

The phrase “Global Catalog is a partial replica of the forest” is therefore incomplete: it is partial for other domains, while retaining a full writable copy of the GC server’s local domain.

Domain controllers and service discovery

A domain controller is not only a host for NTDS.dit. It exposes a coordinated set of services:

Service Role
LDAP / LDAPS Query and modify directory objects
Kerberos KDC Issue TGTs and service tickets
Netlogon Secure channel, DC locator support and domain operations
DNS, commonly AD-integrated Publish and locate domains, DCs, KDCs and GCs
Directory Replication Service Exchange naming-context changes between DCs
SMB Publish SYSVOL and NETLOGON shares
Global Catalog, when enabled Forest-wide search and logon support
sequenceDiagram
    participant C as Domain client
    participant DNS as DNS
    participant DC as Selected domain controller
    participant DB as AD DS database
    participant SYS as SYSVOL

    C->>DNS: Query site-aware DC locator SRV records
    DNS-->>C: Candidate DCs and services
    C->>DC: LDAP ping / locator exchange
    DC-->>C: Site and DC capabilities
    C->>DC: Kerberos or NTLM authentication
    DC->>DB: Resolve account, groups and policy
    DC-->>C: Authentication result and authorization data
    C->>DC: LDAP queries and SMB policy retrieval
    DC->>SYS: Serve GPO template and logon scripts

Figure 4 — Clients discover a suitable DC through DNS and DC Locator, then use different protocols for authentication, directory access and policy.

Domain controllers register DNS SRV records such as _ldap._tcp and _kerberos._tcp, including site-specific variants. Correct subnet-to-site mapping matters: it influences which DC a client selects and which replication path carries changes. Hard-coding a DC hostname in applications defeats this availability and locality model.8

LDAP is the access protocol, not the database itself. A bind establishes a security context, a search selects a base DN, scope and filter, and modify operations update attributes subject to ACLs and schema rules. LDAP signing, SASL sealing, TLS and channel binding solve different transport and relay problems; “use LDAPS” is not a complete security design.910

Replication: one directory, many writable copies

Multi-master does not mean conflict-free

Most directory attributes can be changed on any writable DC holding the partition. That DC creates an originating update, assigns local replication metadata and propagates the change to partners. Partners forward the change without turning it into a new originating write.

flowchart LR
    A[Administrator writes object on DC1] --> U[DC1 creates originating update<br/>with version, timestamp,<br/>originating DSA and USN]
    U --> C1[Inbound connection to DC2]
    U --> C2[Inbound connection to DC3]
    C1 --> V1[High-watermark and<br/>up-to-dateness checks]
    C2 --> V2[High-watermark and<br/>up-to-dateness checks]
    V1 --> D2[(DC2 replica)]
    V2 --> D3[(DC3 replica)]
    D2 -->|forward if needed| D4[(DC4 replica)]

Figure 5 — AD replication transfers attribute-level changes and metadata. It does not copy NTDS.dit from one DC to another for each update.

Update sequence numbers (USNs) are meaningful only within one DC database instance. Replication partners track high-watermarks and up-to-dateness vectors so they request only unseen changes and suppress loops. Version, originating time and originating DSA metadata help resolve concurrent writes. “Last writer wins” is a simplification: conflict resolution evaluates replication metadata, and linked attributes such as group membership have additional behavior.

The Knowledge Consistency Checker (KCC) runs on DCs and creates inbound connection objects. Within a site it favors fast convergence; between sites it uses site links, costs and schedules. Sites contain subnets, not domains: one site can host DCs from several domains and one domain can span many sites.11

Why FSMO roles still exist

Multi-master replication works for most updates, but a few operations cannot safely accept independent concurrent writers. AD DS assigns five Flexible Single Master Operation roles:12

Scope Role Single-master responsibility
Forest Schema Master Authorizes changes to the schema naming context
Forest Domain Naming Master Adds or removes domain and application partitions
Domain RID Master Allocates RID pools to DCs for unique security-principal SIDs
Domain PDC Emulator Preferred password-change replication, lockout handling, time hierarchy and several compatibility/administrative functions
Domain Infrastructure Master Maintains cross-domain object references; limited work in common modern GC/recycle-bin configurations

FSMO holders are not the only DCs that authenticate users or accept routine object writes. Losing one role holder affects its specific operation; it does not instantly turn every other DC into a read-only server. Roles should be transferred normally and seized only when the old owner will not return.

Group Policy and SYSVOL

A Group Policy Object is one logical object stored in two places:

  • the Group Policy Container (GPC) in the domain partition stores metadata, links, status and version information;
  • the Group Policy Template (GPT) under \\<domain>\SYSVOL\<domain>\Policies\{GUID} stores file-based policy, scripts and templates.13
flowchart LR
    GPO[One logical GPO<br/>identified by GUID]
    GPO --> GPC[Group Policy Container<br/>in AD domain partition]
    GPO --> GPT[Group Policy Template<br/>in SYSVOL]
    GPC --> ADR[AD replication]
    GPT --> DFSR[DFSR replication]
    ADR --> DC2[Other domain controller]
    DFSR --> DC2
    DC2 --> CL[Client reads metadata via LDAP<br/>and files via SMB]

Figure 6 — GPO consistency depends on two replication engines: AD replication for the GPC and DFS Replication for the SYSVOL template.

Modern domains use DFS Replication (DFSR) for SYSVOL. Because GPC and GPT replicate independently, transient version mismatches and replication failures can produce inconsistent policy processing. Clients evaluate applicable GPOs through the site, domain and OU hierarchy, then use security filtering, WMI filters, inheritance, enforcement and loopback rules as appropriate.14

SYSVOL is replicated only across DCs in its domain, not across the entire forest. It is also not just a generic shared folder: write access to policy files or logon scripts can become code execution on large numbers of systems. Its ACLs, DFSR health and change history belong in the security monitoring plan.

Authentication and authorization

When a domain user signs in, the DC commonly acts as a Kerberos KDC. NTLM remains available for compatibility and fallback where policy allows it. LDAP may be used before, during and after authentication to resolve objects and configuration, but an LDAP bind is not synonymous with an interactive Windows logon.

flowchart LR
    U[Account credentials] --> DC[Domain controller]
    DC --> AUTH{Authentication protocol}
    AUTH --> K[Kerberos tickets]
    AUTH --> N[NTLM validation]
    DC --> D[Directory lookup]
    D --> S[User SID, group SIDs,<br/>claims and policy]
    K --> L[Local logon authority]
    N --> L
    S --> L
    L --> T[Access token]
    T --> R[Resource access checks]

Figure 7 — The DC authenticates the principal and supplies authorization data; the target Windows system builds a token and performs local access checks.

Domain Admins is not a directory protocol or storage component. It is a default privileged global security group with RID 512. By default it is nested into the local Administrators group on domain-joined computers, including DCs, which gives its members broad control. Microsoft recommends keeping day-to-day accounts out of it and reserving membership for exceptional build or recovery operations.15

Equally important, Domain Admin membership is not the only path to domain control. Enterprise Admins, delegated ACLs, GPO ownership, replication rights, service-control paths on DCs and control of a trusted identity system can all create equivalent or forest-wide impact.

Attack surface and abuse paths

Identity and group control

An attacker who can reset a privileged account, add a member to a privileged group, change sIDHistory, modify an SPN or alter delegation-related attributes can transform a directory write into later authentication or authorization. The dangerous permission may be inherited from a parent OU and may belong to a group nested several levels away.

This is why attack-path analysis should expand group nesting and effective object ACLs rather than list only the direct members of Domain Admins.

Kerberos and NTLM abuse

Weak service-account passwords, disabled Kerberos preauthentication, delegation mistakes and legacy NTLM exposure create credential and relay paths. The Kerberos attack section covers AS-REP roasting, Kerberoasting and forged-ticket concepts; the NTLM article explains why signing, Extended Protection and protocol reduction matter for relay resistance.

Replication rights and DCSync

Directory replication is an authorized API operation, not a request to read NTDS.dit as a file. An identity holding the relevant replication extended rights can ask a DC for password-derived data by imitating a replication partner. This is the basis of DCSync (MITRE ATT&CK T1003.006). The critical control is therefore not only “protect the database file,” but also “protect and monitor who may replicate domain secrets.”16

Legitimate DC replication originates from expected domain-controller accounts and hosts. A replication request from a workstation, member server or ordinary service identity is a high-priority anomaly even if the account was granted the rights months earlier.

GPO and SYSVOL control

Control of a broadly linked GPO can distribute local group changes, scripts, scheduled tasks, registry configuration or security-policy changes to every computer in scope. Control of only the GPC or only the GPT may still be dangerous and can also create a version mismatch useful for detection.

Audit GPO object ACLs, links and SYSVOL paths together. A change that looks small in LDAP may have a domain-wide execution consequence after the next policy refresh.

Domain controller and backup compromise

A writable DC holds enough information and authority to change the forest’s trust state. Offline database copies, system-state backups, hypervisor control, backup-agent credentials and Directory Services Restore Mode credentials must therefore be protected at the same tier as live DC administration. Microsoft states that after unrestricted privileged DC compromise, the forest cannot be considered trustworthy without recovery from known-good state.17

Trusts and forest boundaries

Trusts allow authentication across domains or forests; they do not merge the directories. Direction, transitivity, selective authentication, SID filtering and name resolution determine the actual path. A forest trust deliberately extends reachability across security boundaries, so it should be treated as a high-impact configuration object rather than a networking convenience.

Telemetry and detection

Core event sources

Source Events or data Defensive value
Security — logon 4624, 4625, 4648, 4672 Successful/failed logons, explicit credentials and privileged logon context
Security — Kerberos 4768, 4769, 4771 TGT/TGS requests, encryption types, service names and preauthentication failures
Security — NTLM 4776 plus NTLM Operational logs Credential validation and remaining NTLM dependencies
Security — account management 4720–4726, 4738, 4740–4743 User/computer creation, change, lockout and deletion
Security — group management 4728/4729, 4732/4733, 4756/4757 Members added to or removed from global, local and universal groups
Security — directory changes 5136–5139, 5141 Object modification, creation, undelete, move and deletion
Security — directory access 4662 Extended rights and sensitive object access when matching SACLs exist
Security — replication 4932, 4933 Naming-context synchronization begin/end when auditing is enabled
Policy and audit 4719, 4739, 1102 Audit-policy or domain-policy change and Security-log clearing
Directory Service log Replication, database, LDAP signing/channel binding and DC health events Control-plane health and protocol-hardening compatibility
DFS Replication log SYSVOL replication state and errors GPO file-plane consistency
DNS Server logs Dynamic updates and service-record changes DC location and AD-integrated DNS integrity
EDR / Sysmon on DCs Process, image, network and file activity Unexpected code execution or tooling on Tier-0 hosts

Advanced audit policy enables the categories, while SACLs determine which directory objects and attributes generate 4662 or change events. Event 5136, for example, commonly produces a value-deleted and value-added pair for one modification. A high-volume “audit everything” SACL can hide the changes that matter; target privileged groups, AdminSDHolder, domain roots, GPOs, authentication/delegation attributes and replication rights.1819

High-value correlations

Privileged membership change → privileged logon. Join group changes to the actor’s Logon ID, source host and the affected account’s next 4624, 4768 and 4672. Alert immediately when an unexpected identity modifies Domain Admins, Enterprise Admins, Schema Admins or a delegated administration group.

ACL change → sensitive operation. A 5136 change to a security descriptor may be preparation, not the final action. Follow it with 4662 access, password reset, group modification, GPO change or replication activity by the new trustee.

Replication from a non-DC. Correlate replication extended-right usage, network telemetry and DC role inventory. A successful request from an unexpected source should not be dismissed because the protocol itself is legitimate. Defender for Identity exposes a dedicated DCSync analytic for this pattern.20

GPO dual-store divergence. Correlate GPC changes in AD with writes under the matching SYSVOL GUID. Unexpected writers, a change to only one side or policy files launched from writable locations warrant investigation.

LDAP security drift. Inventory clients generating unsigned binds or missing channel-binding tokens before enforcement, then alert on new legacy clients after the migration window. Windows Server version and deployment age affect defaults; verify effective policy instead of assuming all DCs behave alike.9

DC process execution. Domain controllers should run a small, stable set of software. Interactive browsers, office applications, ad-hoc script hosts, unexpected remote-management tools or unsigned binaries on a DC deserve a much lower alert threshold than on a workstation.

Read-only inventory and health checks

The following commands inspect structure and replication without modifying the directory. They require the appropriate RSAT modules and read permissions.

1
2
3
4
5
6
7
8
Get-ADForest
Get-ADDomain
Get-ADDomainController -Filter * |
    Select-Object HostName, Site, IsGlobalCatalog, OperationMasterRoles

Get-ADRootDSE |
    Select-Object defaultNamingContext, configurationNamingContext,
                  schemaNamingContext, rootDomainNamingContext

Replication and DC health:

1
2
3
4
5
6
7
repadmin.exe /replsummary
repadmin.exe /showrepl * /csv
dcdiag.exe /e /c

Get-ADReplicationPartnerMetadata -Target * -Scope Forest |
    Select-Object Server, Partner, LastReplicationSuccess,
                  LastReplicationResult, ConsecutiveReplicationFailures

DC discovery and policy inventory:

1
2
3
nltest.exe /dsgetdc:example.com /force
Get-GPO -All |
    Select-Object DisplayName, Id, GpoStatus, CreationTime, ModificationTime

Treat these outputs as time-stamped evidence. A healthy result today does not prove that an unauthorized change was never present, and a replication error is not merely an availability problem: it can delay password resets, group removal, policy correction and incident containment.

Hardening

Control Security effect
Treat the forest and every writable DC as Tier 0 Aligns host, hypervisor, backup and administrative controls with forest-wide impact
Use separate privileged identities and hardened privileged access workstations Keeps administrative credentials away from browsing, email and ordinary endpoints
Keep standing membership in Domain Admins, Enterprise Admins and Schema Admins near zero Reduces reusable forest-control credentials and makes exceptional use visible
Delegate narrow OU and object rights through purpose-built groups Replaces broad administrative roles with attributable, reviewable permissions
Review effective ACLs, inheritance and replication rights continuously Finds hidden control paths that group-membership reviews miss
Use gMSAs for services and Windows LAPS for local administrators Removes long-lived shared passwords and limits credential reuse
Protect privileged users with suitable authentication policies, silos and the Protected Users group after compatibility testing Restricts where valuable credentials can be used and reduces legacy authentication exposure21
Prefer Kerberos and reduce NTLM after auditing dependencies Shrinks relay and password-derived authentication paths
Enforce LDAP signing and plan TLS/channel-binding enforcement Protects directory sessions against tampering and applicable relay paths; test clients before rejecting them
Secure GPO creation, linking and SYSVOL write access Prevents policy infrastructure from becoming fleet-wide execution
Maintain accurate sites and subnets and monitor replication health Preserves local authentication, predictable convergence and containment changes
Minimize software, services and interactive use on DCs Reduces code-execution and credential-theft surface on the identity control plane
Centralize DC Security, Directory Service, DFSR, DNS and EDR telemetry Preserves evidence if one DC or its local logs are compromised
Maintain tested system-state backups and an isolated forest-recovery plan Makes destructive or integrity-loss incidents recoverable
Protect DSRM, backup, hypervisor and synchronization credentials as Tier 0 Closes indirect paths to the same directory secrets and authority

Microsoft’s current privileged-access guidance emphasizes explicit validation, least privilege, dedicated administrative devices and continuous monitoring. The important architectural rule is simple: never administer a more trusted system from a less trusted host.22

Forest recovery must be designed before compromise. A valid plan identifies which known-good DC is restored first in each domain, isolates recovered DCs from unsafe replication, resets privileged and krbtgt credentials as appropriate, cleans metadata and validates DNS, SYSVOL and replication before normal connectivity returns.23

Summary

AD DS is a distributed database and security control plane. A forest shares a schema, configuration and trust fabric; domains partition directory data and administration; OUs scope delegation and Group Policy; sites and subnets model the physical network. Each DC stores several naming contexts inside its local NTDS.dit, and a Global Catalog adds partial replicas of objects from other domains.

Replication moves attribute-level changes and metadata between DCs according to a KCC-generated, site-aware topology. Most writes are multi-master, while five FSMO roles serialize the operations that cannot safely accept concurrent writers. Group Policy adds a second replicated plane: the GPC in AD and the GPT in SYSVOL must converge for consistent client behavior.

From a security perspective, privileged groups are only the visible part of the graph. Object ACLs, delegation attributes, GPO control, replication rights, trusts, backups and the DC platform can all yield domain or forest control. Defenders need authentication events, directory-change auditing, replication context, SYSVOL telemetry and endpoint visibility on DCs, joined by identity, object GUID, DN, source host and time. Hardening succeeds when the organization protects AD not as “a login server,” but as the root of trust for every system that depends on it.

References

  1. Microsoft Learn, Understanding the Active Directory Logical Model. ↩︎

  2. Microsoft Learn, Creating an Organizational Unit Design. ↩︎

  3. Microsoft Learn, Object Names and Identities. ↩︎

  4. Microsoft Learn, Naming Contexts and Directory Partitions. ↩︎

  5. Microsoft Learn, Extensible Storage Engine Files. ↩︎

  6. Microsoft Learn, Active Directory Replication Concepts — Global catalog server. ↩︎

  7. Microsoft Learn, Service overview and network port requirements. ↩︎

  8. Microsoft Learn, Optimize domain controller location. ↩︎

  9. Microsoft Learn, LDAP signing overview for Active Directory. ↩︎ ↩︎2

  10. Microsoft Learn, LDAP channel binding in Active Directory. ↩︎

  11. Microsoft Learn, Active Directory Replication Concepts. ↩︎

  12. Microsoft Learn, Flexible Single Master Operations roles in AD DS. ↩︎

  13. Microsoft Learn, Group Policy Storage. ↩︎

  14. Microsoft Learn, Group Policy processing for Windows. ↩︎

  15. Microsoft Learn, Securing Domain Admins groups in Active Directory. ↩︎

  16. MITRE ATT&CK, T1003.006 — DCSync. ↩︎

  17. Microsoft Learn, Securing domain controllers against attack. ↩︎

  18. Microsoft Learn, Advanced Audit Policy Configuration settings. ↩︎

  19. Microsoft Learn, 5136(S): A directory service object was modified. ↩︎

  20. Microsoft Learn, Microsoft Defender for Identity security alerts. ↩︎

  21. Microsoft Learn, Authentication Policies and Authentication Policy Silos. ↩︎

  22. Microsoft Learn, Privileged access. ↩︎

  23. Microsoft Learn, Determine how to recover an Active Directory forest. ↩︎

This post is licensed under CC BY 4.0 by the author.