DACL abuse

Theory

Access privileges for resources in Active Directory Domain Services are usually granted through the use of an Access Control Entry (ACE). Access Control Entries describe the allowed and denied permissions for a principal (e.g. user, computer account) in Active Directory against a securable object (user, group, computer, container, organizational unit (OU), GPO and so on)

DACLs (Active Directory Discretionary Access Control Lists) are lists made of ACEs (Access Control Entries) that identify the users and groups that are allowed or denied access on an object. SACLs (Systems Access Control Lists) define the audit and monitoring rules over a securable object.

When misconfigured, ACEs can be abused to operate lateral movement or privilege escalation within an AD domain.

Practice

If an object's (called objectA) DACL features an ACE stating that another object (called objectB) has a specific right (e.g. GenericAll) over it (i.e. over objectA), attackers need to be in control of objectB to take control of objectA. The following abuses can only be carried out when running commands as the user mentioned in the ACE (objectB) (see impersonation techniques).

Recon

DACL abuse potential paths can be identified by BloodHound from UNIX-like (using the Python ingestor bloodhound.py) and Windows (using the SharpHound ingestor) systems.

Other tools like, Get-DomainObjectAcl and Add-DomainObjectAcl from Powersploit's Powerview, Get-Acl and Set-Acl official Powershell cmdlets, or Impacket's dacledit.py script (Python) can be used in order to manually inspect an object's DACL. At the time of writing, the Pull Request (#1291) offering that dacledit is still being reviewed and in active development. It has the following command-line arguments.

See this page for more details

Abuse

In order to navigate the notes, testers can use the mindmap below.

All of the aforementioned attacks (red blocks) are detailed in the child notes, except:

SPN-jacking: very specific scenario, requires lots of access: see ADDS > Movement > Kerberos > SPN-jacking
Shadow Credentials: see ADDS > Movement > Kerberos > Shadow Credentials
Kerberos RBCD: see ADDS > Movement > Kerberos > Kerberos Delegations > RBCD
GPO abuses: see ADDS > Movement > GPOs
DCSync : see ADDS > Movement > Credential > Dumping > DCSync

Self-attacks

User and computers objects can conduct a Kerberos RCD attack on themselves.
Computer objects can conduct a Shadow Credentials attack on themselves.

ACE inheritance

If attacker can write an ACE (WriteDacl) for a container or organisational unit (OU), if inheritance flags are added (0x01+ 0x02) to the ACE, and inheritance is enabled for an object in that container/OU, the ACE will be applied to it. Impacket's dacledit (Python) can be used with the -inheritance flag for that purpose (PR#1291).

With enough permissions (GenericAll, GenericWrite) over a disabled object, it is possible to enable it again (e.g. set-aduser "user" -enabled 1)

BloodHound edges

BloodHound has the ability to map abuse paths, with some that rely on DACL abuse. The following edges are not includes in the mindmap above:

AddKeyCredentialLink, a write permission on an object's Key-Credential-Link attribute, for Shadow Credentials attacks
WriteSPN, a write permission on an object's Service-Principal-Name attribute, for targeted Kerberoasting and SPN jacking attacks
AddSelf, similar to AddMember. While AddMember is WriteProperty access right on the target's Member attribute, AddSelf is a Self access right on the target's Member attribute, allowing the attacker to add itself to the target group, instead of adding arbitrary principals.
AddAllowedToAct, a write permission on an object's msDS-Allowed-To-Act-On-Behalf-Of-Other-Identity attribute, for Kerberos RBCD attacks
SyncLAPSPassword, both DS-GetChanges and DS-GetChangesInFilteredSet, for synchronizing LAPS password domain-wise
WriteAccountRestrictions, which refers to the User-Account-Restrictions property set, which contains enough permissions to modify the msDS-Allowed-To-Act-On-Behalf-Of-Other-Identity attribute of the target objects, for Kerberos RBCD attacks