National Cyber Warfare Foundation (NCWF)login

Mapping Active Directory attack surface: a conceptual tour of abuse paths for defenders


0 user ratings
2026-11-18 17:29:23
milo
Red Team (CNA)
"Mapping

A conceptual essay on how Active Directory privilege relationships form abuse paths, written as educational analysis for defenders and authorized security professionals.








TopicActive Directory attack surface — the graph of identity and permission relationships that attackers traverse
CategoryIdentity infrastructure analysis
Primary UseUnderstanding how misconfigurations in AD objects, ACLs, and trust relationships chain into escalation paths during authorized assessments
Safe UseConceptual analysis only, for defenders and authorized security teams hardening their own directories in labs and owned environments
Telemetry NoteAbuse paths in AD are observable through authentication logs, Kerberos ticket anomalies, directory object change auditing, and anomalous LDAP query patterns

Active Directory is best understood not as a product but as a graph. Every user, computer, group, organizational unit, and Group Policy object is a node, and every membership, permission, delegation, and trust relationship is an edge. Attackers who target AD are rarely breaking cryptography; they are walking this graph, finding edges that the defenders did not know existed, and chaining them into a path from a low-privilege foothold to domain dominance. This essay is a conceptual map of those abuse paths, written for defenders and authorized security professionals who need to reason about their own directories as an adversary would, without reproducing operational exploitation detail.


The foundational insight is that in AD, compromise is transitive in ways that are hard to see from any single console. A service account with a spn attribute, a group with an overly broad WriteDacl grant, a GPO linked at the wrong scope, or a one-way trust to a less-trusted forest each look unremarkable in isolation. The danger emerges only when they compose. This is why the modern tooling ecosystem — enumeration frameworks and graph analyzers that model ACLs, GPOs, kerberoastable principals, constrained and unconstrained delegation, gMSA password reads, and ADCS certificate templates — is fundamentally graph tooling rather than scanner tooling. The tool answers the question: given what you already control, what can you reach?


Start at the outermost ring: reconnaissance of the directory itself. AD is structurally verbose; an authenticated user with no special privileges can enumerate users, groups, computers, and service principal names through standard LDAP queries. From a defensive standpoint, this means you cannot rely on the directory being secret — its shape is effectively public to anyone inside the perimeter. The observability angle matters here: heavy, patterned LDAP enumeration, especially queries sweeping entire subtrees for spn attributes or sddl information, is one of the more reliable signals that someone is building a map of your environment. Enabling query logging on domain controllers and baselining normal directory access patterns turns this reconnaissance phase into a detection opportunity rather than a blind spot.


The first major family of abuse paths concerns credentials stored or exposed within the directory. Service accounts with spn attributes present a well-documented structural problem: any authenticated principal can request a service ticket for such accounts, and the ticket's offline crackability depends entirely on the entropy of the underlying password. Defensively, the fixes are architectural rather than reactive — long machine-generated passwords, gMSA accounts whose credentials rotate and are readable only by authorized principals, or the removal of spn attributes from human-service accounts entirely. The same reasoning extends to password storage in attributes like description fields or GPP cpassword blobs in SYSVOL, historical weaknesses that persist in uncleaned environments because directory hygiene is cumulative and rarely gets a dedicated budget.


The second family is delegation, and it deserves its own conceptual treatment because it is where intuition fails most often. Unconstrained delegation allows a server to hold a user's full ticket-granting-ticket, effectively making that server a proxy for the user's authority everywhere. Constrained delegation and resource-based constrained delegation narrow the blast radius but introduce their own configuration edges — who can delegate to whom, under which protocols — that become graph edges an adversary can traverse if the configuration was made permissively during a troubleshooting session and never reverted. ADCS certificate templates extend this family dramatically: a template that allows low-privileged enrollment with an ESC-permissive configuration can mint a credential that authenticates as anyone, converting a certificate server into an implicit path to domain escalation. The common thread is that each of these features was designed to make administration easier, and every convenience in identity infrastructure is a potential edge in the abuse graph.


The third family is ACL abuse, and it is the most generative because discretionary access control is arbitrarily expressive. Generic rights such as GenericAll, GenericWrite, WriteDacl, WriteOwner, and ForceChangePassword on sensitive objects — the domain root, Domain Admins, a GMSA principal, a GPO — allow a holder to rewrite the rules that govern the target. The subtle and dangerous cases are indirect: writing a spn onto an account you have GenericWrite over converts it into a ticket-harvesting target; adding a member to a group that itself holds a sensitive right chains two individually-explainable grants into one escalation. Because ACL assignments accumulate over years of tickets and projects, virtually every mature directory contains edges nobody intended. Defensive review means periodically exporting the security descriptors on tier-zero objects and asking, for every grant, who benefits and whether the grantee is itself trustworthy transitively.


The fourth family is tiering failure — the human architecture layer. AD abuse paths are only exploitable if an attacker holds a credential somewhere in the graph, and the cheapest credentials come from endpoints and habits. Administrators who log into workstations with privileged accounts place those credentials within reach of any compromise on that machine; cached credentials, LSA secrets, and automation passwords on jump boxes turn each endpoint into a fragment of the identity graph. Tiered administration models exist precisely to cut these edges: privileged identities exist only on dedicated hosts, never browse, never read email, never hold spns. Conceptually, tiering is the practice of disconnecting subgraphs so that a foothold in one cannot reach a credential in another.


Trust relationships form the fifth family and the one most often ignored in scoping. Forest and domain trusts are edges between entire graphs, and the direction, transitivity, and selective-authentication settings of each trust determine whether abuse paths cross organizational boundaries. Mergers and acquisitions typically produce hastily-created wide trusts that are never revisited. A defender's map of attack surface must therefore extend beyond the forest boundary: enumerate trusts, understand which side authenticates where, and treat any bidirectional transitive trust to a forest with weaker controls as a first-class escalation edge — even though no single object in your directory appears misconfigured.


How should defenders act on this conceptual map? The discipline that has matured across the industry is to model the directory as data: export objects, memberships, ACLs, delegation settings, and certificate templates into a graph, compute shortest paths from plausible footholds to tier-zero assets, and remediate the shortest paths first. The value is prioritization — most directories contain too many theoretical weaknesses to fix everything at once, and path analysis tells you which few edges actually connect an attacker's likely starting position to the crown jewels. Periodic re-runs matter because directories are living systems; every joiner-mover-leaver cycle and every delegated admin task can add edges back.


Detection deserves equal conceptual weight to prevention. The abuse paths described here all manifest as directory activity that can be logged: modifications to ACLs and group memberships on sensitive objects, Kerberos ticket requests for unusual service principal names, unusual ticket encryption types, certificate template enrollment outside baseline, and anomalous LDAP patterns. The classic mistake is logging domain controllers generically and never writing detections for the specific object classes and operations that matter. Tier-zero change alerting, in particular, converts the highest-value abuse edges into high-signal alarms, because legitimate changes to those objects are rare and reviewable.


There is also a hardening dimension that closes whole families at once rather than edge-by-edge. Removing legacy authentication paths (NTLM where feasible), requiring strong ticket encryption types, eliminating reversible-encryption stores, auditing ADCS templates against published hardening baselines, enabling LDAP channel binding and signing, and adopting least-privilege administrative models each reduce the number of usable edges in the graph. None of these are novel; the recurring finding in authorized assessments of real enterprises is not exotic technique but the persistence of long-documented weaknesses in environments where identity hygiene was never assigned an owner.


The takeaway for defenders and authorized security teams is to abandon the fortress mental model. Active Directory security is not a wall but a weighted graph, and your risk is the shortest path from any plausible compromise to domain control, weighted by how detectable each hop is. Map that graph in your own lab and production directories with your favorite enumeration and path-analysis tooling, cut the shortest edges first, instrument the tier-zero nodes, and re-map continuously. The adversary's advantage has always been that they see the graph while defenders see consoles; the entire modern discipline of AD attack surface management is about closing that perceptual gap — and it is a gap that closes one honest look at your own directory at a time.



Educational analysis for authorized security professionals. Use only in controlled, authorized environments.






Source: OffensiveSec
Source Link: https://www.offsecblog.com/2026/11/mapping-active-directory-attack-surface.html


Comments
new comment
Nobody has commented yet. Will you be the first?
 
Forum
Red Team (CNA)



Copyright 2012 through 2026 - National Cyber Warfare Foundation - All rights reserved worldwide.