Active Directory is the heart of almost every corporate network, so it's the heart of almost every internal pentest. This chapter is deliberately purple: it shows how a tester maps the domain and identifies the weaknesses that lead to Domain Admin, and for each one it gives the detection and the fix, because a finding without a remediation is worthless.
Everything here assumes an authorised internal engagement with a provided low-privileged domain account ("assumed breach"), the most common real-world starting point.
What makes AD attackable
AD is a web of trust: users, groups, computers, permissions (ACLs), Kerberos tickets and delegation. Attackers don't break AD's crypto; they follow misconfigurations — over-permissioned accounts, legacy settings and weak passwords — from a normal user to full control. Your job is to find the shortest such path and show the client how to cut it.
low-priv domain user
│ enumerate (who can do what)
▼
find a misconfig / weak account
│ (Kerberoast, delegation, ACL, admin reuse)
▼
escalate to a privileged account
│ repeat
▼
Domain Admin / DC → full controlEnumerate the domain
Standard authorised-pentest enumeration from a domain user:
# Users, groups, policy, computers — NetExec and Impacket
nxc ldap dc.corp.local -u user -p pass --users --groups
nxc smb 10.10.10.0/24 -u user -p pass # map hosts, signing, OS
impacket-GetADUsers -all corp.local/user:pass -dc-ip 10.10.10.10
# Password policy (so any spraying stays safe)
nxc smb dc.corp.local -u user -p pass --pass-polBloodHound: see the attack paths
BloodHound collects AD relationships and draws the graph from any node to "Domain Admins". It's the single most valuable AD assessment tool, for attackers and defenders alike.
# Collect with the Python/Rust collector from a domain account
bloodhound-python -u user -p pass -d corp.local -ns 10.10.10.10 -c All
# or the C# SharpHound collector on-host, then import the JSON into BloodHound CEThen ask BloodHound questions: "Shortest paths to Domain Admins", "Who can DCSync", "Kerberoastable accounts", "Computers where a given user is local admin". The graph turns a vague "AD is complex" into a concrete, named path you can put in the report.
The common weaknesses (and their fixes)
Each of these is a standard finding. Understand the concept, then read the detection and hardening you'll hand the client.
Kerberoasting
- Concept: any domain user can request a service ticket for an account that has a Service Principal Name (SPN). Part of that ticket is encrypted with the service account's password hash, so it can be taken offline and cracked. Service accounts often have weak, non-expiring passwords.
- Tester does: requests roastable tickets (
impacket-GetUserSPNs), cracks offline with hashcat (mode 13100). Reports which accounts cracked and how fast. - Detect: Event 4769 for service-ticket requests, especially with RC4 (
0x17) encryption, in bursts from one user. Honeypot SPN accounts. - Fix: 25+ character random passwords on service accounts, group Managed Service Accounts (gMSA), AES-only Kerberos, and remove unused SPNs.
AS-REP roasting
- Concept: accounts with "Kerberos pre-authentication" disabled let anyone request a crackable blob without logging in.
- Detect: Event 4768 with pre-auth not required. Fix: require pre-authentication on all accounts.
Weak and reused passwords
- Concept: password spraying one seasonal password finds a foothold; shared local-admin passwords enable lateral sweep.
- Fix: long passphrases, banned-password lists, MFA on remote access, and LAPS so every machine has a unique local-admin password.
Dangerous ACLs and delegation
- Concept: a normal user may hold rights (
GenericAll,WriteDACL,ForceChangePassword,AddMember) over a privileged object, or a computer/account may have unconstrained or constrained delegation that lets it impersonate others. BloodHound surfaces these as edges. - Detect: monitor changes to privileged group membership (4728/4732/4756) and to ACLs; alert on delegation changes. Fix: remove excess rights, apply the tiered admin model, mark admins "sensitive and cannot be delegated", and avoid unconstrained delegation.
Credential theft from memory and DCSync
- Concept: with local admin on a host, attackers read credentials cached in memory; with specific replication rights, an account can request password data from a DC (DCSync). These are how a foothold becomes domain-wide.
- Detect: credential-access tooling signatures, LSASS access alerts, and unexpected 4662 replication requests from non-DC accounts (DCSync). Fix: Credential Guard, tiered admin, restrict replication rights to DCs, and disable legacy protocols.
The assessment loop
enumerate ─► BloodHound path ─► weakest edge
▲ │
└────── new privileges ◄──────────┘You repeat enumerate → find weakest link → escalate until you reach the objective (often Domain Admin or a specific crown-jewel system), recording each step and its fix as you go.
Turning it into a report the client can act on
For each finding give: the path (BloodHound screenshot), the named accounts/objects involved, the impact (what it leads to), the detection (specific event IDs and tooling), and the fix (specific setting). A good AD section reads as a prioritised remediation plan, not a trophy list.
🧪 Lab
Use a deliberately vulnerable lab AD you control: GOAD (Game of Active Directory), vulnerable-AD, or a HackTheBox/TryHackMe AD lab. Never a production domain.
- From a provided domain user, enumerate users, groups and the password policy with
nxc/Impacket. - Collect with BloodHound and find the shortest path to Domain Admins.
- Identify the Kerberoastable accounts and, in the lab, crack one; write the detection and fix.
- Find one dangerous ACL edge in BloodHound and explain the risk and remediation.
- Write a one-page AD findings section with path, impact, detection and fix for two issues.
Checkpoint
- Attackers rarely break AD's cryptography — what do they follow instead?
- What does BloodHound compute, and why is it valuable to defenders too?
- In concept, why is Kerberoasting possible, and what are its detection and fix?
- Why are shared local-admin passwords dangerous, and what fixes them?
- What must every AD finding in your report include besides the attack path?