Manual Enumeration and Attack-Path Mapping
An authorized assessment has produced an identity graph showing a route toward a privileged group. Several relationships in the graph came from automated collection. One edge says that a particular user has a permission that might allow control over another identity. What should you do next?
You could trust the graph and describe the route as an attack path. You could attempt to use the permission immediately. Or you could inspect the evidence, confirm what the permission actually allows, and decide whether any further validation is authorized. Only the last choice separates an interesting hypothesis from an unsupported finding.
PenTest+ objective 2.2 includes manual enumeration and attack-path mapping. Manual enumeration means human-directed discovery and validation rather than relying only on automated collection. It doesn’t mean rejecting automation.
Automated tools can gather and organize far more data than you could reasonably inspect one item at a time. The problem appears when convenient output is treated as complete, current, and correctly interpreted without further thought. Automated discovery can miss information, preserve stale information, or present a relationship whose practical meaning depends on conditions the output didn’t establish.
Manual enumeration fills that reasoning gap. You compare tool output with the environment you’re authorized to assess. You inspect the identities, privileges, hosts, services, and reachable resources that matter to the question. You also look for gaps.
Perhaps a crawler didn’t reach a resource that required a particular session. Perhaps network enumeration identified a service, but the service information needs confirmation. Perhaps an identity relationship was collected earlier and no longer reflects the effective permission. Manual work helps you validate what automation found and pursue relevant omissions without assuming that every missing result means the resource doesn’t exist.
Attack-path mapping uses those observations differently. An attack path is a chain of relationships or conditions that could allow an attacker to progress toward a target. The chain might connect a user to a permission, that permission to another identity, the second identity to a host, and the host to a privileged resource.
BloodHound describes attack paths as chains of abusable privileges and user behaviors connecting principals. A principal is an identity, such as a user or group, that can receive permissions. Mapping these chains can also reveal high-value remediation points where changing one important relationship may interrupt more than one possible route.
Manual enumeration and attack-path mapping therefore support each other, but they aren’t interchangeable. Manual enumeration asks whether an observation is accurate and whether relevant information is missing. Attack-path mapping asks how validated or suspected relationships might connect. Evidence that a user belongs to a group supports one relationship.
It doesn’t, by itself, prove that the user can reach a particular host, exercise a separate permission, or complete every later step in a chain. Likewise, a scanner’s failure to report a service is not sufficient evidence that no service exists. In both cases, you have an observation that must be interpreted within the engagement’s boundaries.
A useful decision process begins with the requirement. What question has the assessment been authorized to answer? If the requirement is to evaluate routes to a privileged group, identify the specific target and the starting identity you’re allowed to consider. Then state the expected condition.
If the graph’s disputed edge is accurate, what evidence should exist? You might expect the relevant permission to appear in an authoritative or corroborating source. You might also expect the permission to apply to the exact identity and object shown, rather than to a similarly named account or an outdated object.
Next, choose a bounded action that can produce that evidence. Begin with the least disruptive valid method. Reviewing collected permission data or obtaining another authorized observation may answer the question without exercising the permission.
If passive review can’t establish whether the relationship is effective, a more active validation might be considered, but only when the rules of engagement allow it. Authorization, target scope, timing, dependencies, and likely operational impact all influence that choice. A technically possible action is still the wrong action when it exceeds those constraints.
Suppose the graph shows that the account Riley can influence the account Morgan, and Morgan belongs to a privileged group. The proposed path is Riley, then Morgan, then the group. The uncertain part is the first edge.
Before doing anything stronger, predict what you need to confirm. You need evidence that the permission belongs to Riley, applies to Morgan, remains effective under current controls, and has the meaning assigned to it by the graph. Evidence of Morgan’s group membership confirms the second relationship, but it can’t repair uncertainty in the first.
Now check the engagement details. Riley and Morgan may both be in scope, yet active modification of either account may be prohibited. The testing window may exclude business hours. An account might support an important service, making even an authorized change unnecessarily disruptive.
If the permitted evidence can be gathered by reviewing current relationships, that lower-impact method comes first. If the permission can only be tested actively and the authorization is unclear, the correct next step is clarification or escalation through the engagement process. You don’t silently broaden the test because a graph made the path look promising.
After the bounded check, compare the result with your prediction. If current evidence confirms that Riley holds the relevant permission over Morgan, you can describe that edge as supported. If the relationship was removed, applied to a different object, or was interpreted incorrectly, the edge is not supported by current evidence.
If the available method can’t determine whether the permission is effective, the status remains unresolved. Those are three different outcomes. Confirmed, rejected, and unresolved should not collapse into the same statement merely because an automated graph drew the same line for all three.
This process produces more than a yes or no answer. It records the requirement, the predicted evidence, the method used, the observed result, and the remaining uncertainty. That record lets another person understand why you accepted or questioned the relationship. It also prevents a later analyst from interpreting a cautious hypothesis as a completed exploit chain. The value lies in defensible reasoning, not in making the graph contain as many edges as possible.
Conflicting results are especially important because they tell you where to investigate your assumptions. Imagine that the graph still shows Riley’s permission over Morgan, but a current manual review doesn’t. It’s tempting to declare either the graph wrong or the manual check conclusive. First ask whether the two methods examined the same target, identity, and relationship.
Confirm that the names refer to the same objects. Check whether the automated data and the manual observation were collected at different times. Consider whether the account performing the manual review had enough visibility to observe the permission.
You should also separate a collection problem from a security finding. A tool configured for the wrong target may return clean-looking output that has no bearing on the intended target. Insufficient privileges can hide information. Missing prerequisites can prevent a relationship from being evaluated.
Old data can preserve an edge that no longer exists, while incomplete current data can hide one that does. None of these conditions proves that the proposed path is exploitable or safe. They affect the quality of your observation, so troubleshoot them before reaching a stronger conclusion.
The most dangerous misconception here is that a graph edge proves an exploitable step. An edge indicates a relationship or a potential abuse condition. Its practical use may depend on prerequisites, effective permissions, reachability, current controls, and user behavior. Even if each edge describes a real relationship, the entire chain isn’t automatically proven.
One link may require a condition that isn’t present. Another may be blocked by a current control. Two individually valid observations may also come from different moments and fail to describe one usable path at the same time.
Return to the Riley and Morgan case. Assume Morgan’s privileged group membership is current, but the permission attributed to Riley came from older collection. The graph contains a complete-looking route, yet its appearance doesn’t answer the important question. Does Riley currently have the specific effective permission needed for the first transition?
If a current, authorized review shows that the permission was removed, the path fails at that edge. The group membership remains true, but it’s no longer reachable through the proposed chain. If the review can’t see the relevant permission, you still can’t say the edge is absent. You report the limitation and identify what authorized evidence would distinguish absence from lack of visibility.
Now consider the opposite conflict. Manual enumeration finds a relationship that the automated collection missed. Perhaps the automated data was incomplete or the relevant object wasn’t included. Before adding the edge, confirm that you’re examining the intended identities and that the relationship has the expected meaning.
Then update the hypothesis. Manual enumeration has filled a gap, but the newly observed relationship is still one part of a larger path. Every later transition needs its own support, and the path as a whole needs to be described according to the evidence actually established.
When stronger validation is permitted, it should answer a defined question rather than become open-ended exploitation. You predict the evidence, identify the minimum action needed to observe it, consider operational impact, and remain within the approved target and time. If a lower-impact check already establishes the relationship, escalating adds risk without adding necessary confidence.
If an active check is required but prohibited, document the unverified condition. Professional restraint doesn’t weaken the assessment. It preserves the distinction between what the evidence supports and what you merely suspect.
That distinction carries directly into reporting. Your working material might label a relationship as observed, corroborated, stale, contradicted, or unresolved. The exact vocabulary can follow the organization’s reporting process, but the meaning must stay clear. A report shouldn’t say that an attacker can reach a privileged group when you only identified a potential route with an unverified edge. A more defensible description explains the confirmed relationships, identifies the uncertain condition, states how the data was collected, and records why further testing was or wasn’t performed.
Attack-path mapping also helps prioritize remediation without pretending that every route has been exploited. If several potential paths depend on the same privilege relationship, that relationship may deserve focused review. Removing or constraining it could interrupt multiple routes.
If a privileged group is reachable through several independent relationships, fixing only one edge may leave alternatives. The graph helps analysts see those connections, while manual validation helps determine which connections are current and meaningful. Remediation decisions become stronger when they’re based on validated relationships and clearly marked uncertainty.
The artifact passed to the next phase should therefore be more useful than raw scanner output. It can identify the relevant identities, hosts, services, privileges, and resources. It should show which relationships were observed, which were manually corroborated, and which remain hypotheses. It should also preserve engagement constraints that affected validation.
If active testing wasn’t allowed, that limitation matters. If evidence was collected at different times, that matters too. The next analyst needs enough context to avoid treating a conditional relationship as an established route.
Consider another authorized assessment with two plausible next actions. An automated graph suggests that a standard user may have a route to a sensitive administrative resource. One choice is to exercise the whole route immediately because every edge appears in the graph. The other is to identify the least certain edge, predict the evidence that should support it, and validate it through the lowest-impact authorized method.
The second choice fits the available evidence. The first assumes that graph construction has already performed validation and that authorization for enumeration automatically permits every stronger action. Neither assumption is justified.
What information could change that decision? Clear authorization for active validation could permit a stronger check. Fresh, corroborated evidence might remove the need for one. Operational constraints could prohibit action even when it’s technically feasible.
Evidence that the target identity or resource is outside scope would end that branch of testing. The correct choice isn’t determined only by which technique can produce an answer. It depends on whether that technique addresses the uncertainty, remains inside the engagement, and has an acceptable impact.
This same reasoning applies when automated output seems empty rather than alarming. If a scanner or crawler reports nothing relevant, first decide what it was expected and authorized to observe. Then examine coverage, target selection, configuration, data freshness, and visibility. Manual enumeration may reveal a missed host, service, resource, or relationship.
It may instead confirm that the automated result is consistent with the available evidence. Either outcome is useful when the method and limitations are documented. An empty result without those checks is just an observation, not proof that the attack surface is empty.
By the end of this work, you should be able to trace why each important relationship is present in your analysis. Some edges will rest on direct, current observations. Others will remain conditional because a prerequisite couldn’t be verified or because the permitted testing stopped short of active use. Marking that difference is part of technical accuracy. It keeps attack-path mapping focused on evidence, helps remediation owners understand where action may have the greatest effect, and prevents an attractive diagram from claiming more certainty than the engagement established.
Manual enumeration gives you a way to question, corroborate, and complete automated discovery. Attack-path mapping then connects those observations into possible routes through identities, privileges, systems, services, and resources. The connection between them is disciplined validation.
A line in a graph can support a hypothesis, but it doesn’t prove that every condition is current or that the complete chain can be used. When you define the question, predict the required evidence, choose the least disruptive authorized check, and document both results and uncertainty, your map becomes a defensible assessment artifact rather than a collection of assumptions. That’s what lets later decisions about testing and remediation rest on evidence without crossing the engagement’s boundaries.
One final question should stay with you: What should you do next? The companion episode reference provides exact terminology, named tools or standards, source-derived evidence checkpoints, and review material for Manual Enumeration and Attack-Path Mapping. This lesson was produced by Bare Metal Cyber dot com.