Documenting Attack Paths, Dependencies, and Scope
An authorized penetration test has identified an internet-facing web application with a possible route toward a valuable domain resource. Between those two points sits a production database marked as sensitive in the rules of engagement. What should happen next?
It might be tempting to draw an arrow from the web application to the database, another arrow to the domain resource, and treat the route as established. But those arrows could represent very different things. One might record a connection you observed.
Another might express a hypothesis that still needs testing. A third might cross a boundary you aren't permitted to test at all. Before choosing an attack, you need documentation that makes those differences unmistakable.
This is part of analyzing output to prioritize and prepare attacks in PenTest+ PT0-003. The relevant objective includes attack-path documentation, low-level diagrams, storyboards, dependencies, scope limitations, and labels for sensitive systems. Operationally, that means turning scattered technical information into a defensible picture of what could happen, what has happened, and what you're allowed to do next.
It doesn't mean that every possible route should be exploited. It also doesn't mean that a polished diagram proves its own accuracy. The artifact helps you make decisions, but its value depends on the evidence and constraints attached to it.
An attack path connects conditions and actions to an outcome. A useful path might begin with exposed web access, depend on a particular weakness or privilege, require communication with another system, and end with access to a protected resource. Each connection is an edge in the path.
Each system, account, service, or security state can be treated as a point the path passes through. The important part isn't the drawing convention. It's whether another reviewer can understand why each connection exists and how certain you are about it.
A low-level diagram captures technical relationships with enough detail to support a decision. For this engagement, it could identify the exposed application, the production database, and the domain resource, along with the relevant connections, required access, and boundaries. If the rules of engagement identify the database as sensitive, that label belongs directly on the artifact.
It shouldn't be left in a separate document and trusted to the tester's memory. The path should also show whether the database is in scope, excluded from active testing, or subject to special handling. A route can be technically plausible while still being unusable under the engagement terms.
A storyboard expresses the path as an ordered progression. It records the precondition for a step, the proposed or observed action, the evidence associated with it, and the resulting state. For example, step one might state that the web application is reachable and identify the observation supporting that statement. Step two might propose using an authorized, bounded method to determine whether the application can reach the database.
Step three would depend on the outcome of step two. That ordering exposes dependencies. If the second step isn't validated, later steps that rely on it remain hypothetical.
The diagram and storyboard serve related but different needs. The diagram emphasizes relationships and boundaries. The storyboard emphasizes sequence, preconditions, decisions, and outcomes. They can describe the same proposed route without becoming interchangeable.
A diagram may make it easy to notice that every route to the domain resource passes through the sensitive database. A storyboard may make it easier to see that a planned action assumes credentials or connectivity that haven't been demonstrated. Used together, they help prevent an attractive end state from hiding the unresolved conditions underneath it.
You can also use a common vocabulary for describing adversary behavior. MITRE ATT&CK organizes behavior into tactics and techniques, which can help different reviewers describe actions consistently. That language can clarify what kind of behavior a proposed step represents. It doesn't replace engagement-specific evidence, though.
Assigning a technique name to an arrow doesn't demonstrate that the action succeeded against this target. The label describes behavior. Your notes must still identify the target, preconditions, observations, authorization, and result.
Start the decision process with a requirement or an observation. Suppose your current observation is limited to this: the approved test has confirmed that the web application is externally reachable. The engagement question is whether that exposure could contribute to access to the domain resource. You haven't yet proved application compromise, database access, or domain access.
Your artifact should say exactly that. The confirmed reachability is an observed fact. The route through the application and database is a proposed path. Keeping those categories separate prevents possibility from quietly becoming reported fact.
Next, predict what evidence should exist if the proposed connection is real. If one step depends on access from the application environment to the database, evidence should address that specific dependency. An unrelated vulnerability identifier or a high severity score doesn't prove the connection. A public vulnerability record can help identify and define a disclosed vulnerability, but it doesn't establish that the target has the affected condition or that exploitation would produce the required position in this path. The documentation should connect evidence to the particular claim it supports rather than collecting impressive-looking facts around the diagram.
Then examine authorization, target scope, timing, dependencies, and operational impact. The production database has been marked sensitive because interaction with it may carry business, privacy, regulatory, safety, or operational consequences. That label changes the decision even if the database is technically reachable. The rules might allow passive review of existing evidence while prohibiting active exploitation.
They might require coordination during a maintenance window. They might exclude the system completely. If the applicable boundary isn't clear, the next step is clarification or escalation through the engagement process, not an assumption that the broader route is permitted.
Now choose a bounded action that can answer the immediate question with the least necessary impact. You don't need to prove the entire route in one attempt. Existing logs, approved configuration information, previously collected output, or a narrowly authorized validation may be enough to establish whether one dependency is present.
The exact method depends on the engagement and the evidence available. The governing principle is stable: when a lower-impact method can establish the needed evidence, use it before escalating. A technically capable action isn't the right action if it exceeds the approved target, timing, or operational constraints.
After the action, compare the result with the prediction. Imagine that approved evidence confirms communication from the application environment to the database but says nothing about access from the database environment to the domain resource. The diagram can now mark the first relationship as observed or validated to the level supported by that evidence. The second relationship must remain proposed or unresolved.
The storyboard records the result of the bounded check, the state it established, and the remaining question. You haven't failed to complete the drawing. You've improved its accuracy by refusing to imply more than the evidence demonstrates.
This distinction also changes prioritization. If the only plausible route crosses a sensitive system that can't be actively tested, the best next action might be to seek an approved alternative validation, request clarification, or investigate another in-scope route. It wouldn't be defensible to select a more intrusive exploit merely because it could move the storyboard forward. Attack preparation is not just choosing a capability. It includes deciding whether the preconditions are present, whether the action can answer the engagement question, and whether the likely consequences are acceptable under the authorization.
Documentation created before exploitation is especially useful because it exposes hidden assumptions while there is still time to correct them. Suppose the path assumes that one set of credentials will work on a later resource. Writing that precondition down reveals the dependency. Suppose it assumes that two systems can communicate.
The diagram makes that connection available for validation. Suppose an action would cross from a test environment into production. A clearly marked boundary can stop that action before it occurs. Reconstructing the route afterward may explain what happened, but it can't provide the same preventive decision support.
Now consider conflicting evidence. One data source suggests that the application can connect to the sensitive database. Another indicates that the relevant connection is blocked. That conflict doesn't justify selecting the interpretation that makes the attack path more interesting.
First, identify exactly what each item shows. One result may be older than the other. One may concern a different target, network position, configuration, or privilege level.
A tool may have failed to collect the expected information. The security condition itself may also have changed. Until those possibilities are examined, the edge should be marked unresolved rather than validated or rejected.
Troubleshooting begins with the assumptions behind the observation. Was the correct target selected? Was the relevant configuration examined? Did the data come from the system and time period represented in the path?
Did the collection method have enough privilege to observe the condition? Were the required prerequisites present? These questions separate an actual security finding from incomplete data, a collection failure, or a mistaken assumption. They also help you avoid responding to uncertainty with a stronger technique that may add risk without improving the quality of the answer.
Choose the next validation according to the disputed claim. If the question is whether a dependency exists, gather evidence that addresses that dependency and no more. Keep the sensitive label, authorization limits, timing restrictions, and operational concerns visible while deciding.
If existing approved evidence can resolve the discrepancy, there's no reason to interact more aggressively with the production database. If no permitted method can answer the question, record the limitation. An unresolved condition honestly documented is more useful than an unsupported edge presented with false confidence.
The central misconception is simple but dangerous: a plausible path diagram is not evidence that every edge has been successfully validated. Plausibility means the route appears possible given current assumptions and information. Validation means appropriate evidence supports the specific relationship or state. A path can contain a mixture of confirmed observations, supported inferences, proposed actions, rejected branches, and unresolved questions. If the artifact uses the same unqualified arrow for all five, a reviewer can't tell what the test actually demonstrated.
Return to the route through the production database. Perhaps the application is reachable, a known weakness appears relevant, and architecture information suggests a connection to the database. Those facts may support further analysis, but they don't prove successful application compromise, usable database access, or access to the domain resource. The public identifier describes a disclosed vulnerability.
The architecture information suggests a dependency. Neither item alone validates the complete route. The next authorized check must be chosen according to the missing evidence, not according to the visual momentum of the diagram.
This is also why observations and proposed next steps need explicit language. A statement such as “the application reached the database” should only appear when the collected evidence supports that outcome. If the step hasn't occurred, describe it as proposed and include its preconditions.
If evidence only partially supports it, state the limitation. If active validation is prohibited, identify the scope restriction. A reviewer should be able to look at the artifact and distinguish what was observed, what was inferred, what is planned, and what cannot be tested under the current authorization.
Good attack-path documentation becomes part of the wider penetration-testing workflow. During planning, it records boundaries, sensitive targets, and approval conditions. During evidence collection, it gives observations a place in the developing route.
During analysis, it exposes missing prerequisites and competing explanations. During attack preparation, it helps select a bounded action. During reporting, it supports a clear account of demonstrated exposure, unresolved risk, testing limitations, and the relationships that matter to remediation.
That reporting connection deserves attention. A finding shouldn't overstate impact merely because a severe outcome appears at the end of a proposed path. If only the early steps were validated, the report needs to preserve that qualification.
At the same time, an incomplete path may still reveal an important dependency or control weakness. The artifact lets the reviewer see where evidence is strong, where the path remains conditional, and where scope prevented further validation. This makes the result useful without pretending that an unperformed action succeeded.
The same artifact can improve remediation discussions. If a route depends on several conditions, defenders can examine where the path could be interrupted. The penetration tester's job in this episode isn't to promise that one change eliminates every risk. It's to document the relationships and evidence accurately enough for the organization to understand the demonstrated and conditional parts of the route. When a sensitive database sits at a critical junction, the label helps ensure that its operational importance remains visible during both technical review and reporting.
Consider another authorized situation. Analysis identifies two possible next actions. The first is an intrusive attempt against a sensitive production system during normal operations. The second is a lower-impact review of already approved evidence that could confirm whether the required dependency exists.
Both might sound technically relevant. Under the stated conditions, the second choice is stronger because it addresses the immediate evidence gap while respecting impact and scope. The first choice would require additional authorization or different operating conditions before it could become defensible.
What information could change that answer? Explicit approval for the intrusive action could matter. So could an authorized testing window, a verified recovery plan, or evidence that the lower-impact method can't answer the engagement question.
Those details wouldn't make intrusive testing automatically necessary. They would change the constraints used to evaluate it. Attack-path documentation should capture those decision points so the choice doesn't appear arbitrary when another tester, engagement lead, system owner, or report reviewer examines it later.
Now reverse the situation. Suppose existing evidence is stale and refers to a retired configuration, while a narrowly authorized check against the current environment has low expected impact. Relying on the old evidence may be the weaker choice because it could support the wrong conclusion.
The bounded current check may be justified if it remains within authorization and operational limits. Least impactful doesn't mean never interact with a target. It means use the least disruptive valid action that can establish the evidence you actually need.
A mature artifact therefore contains both technical and engagement logic. Technically, it links systems, access states, preconditions, actions, and outcomes. Professionally, it identifies authorization, scope boundaries, sensitive systems, timing restrictions, evidence quality, and unresolved uncertainty.
Removing either side makes the artifact less useful. A technically detailed path without constraints can encourage unauthorized action. A scope-aware document without technical dependencies can't explain why one action follows another.
Before any proposed step becomes an executed step, you should be able to answer a short set of connected questions in ordinary language. What condition has actually been observed? What outcome are you trying to establish? Which preconditions remain uncertain?
What evidence would confirm or weaken the proposed connection? Is the target authorized at this time and with this level of impact? How will you record the outcome without overstating it? Those answers turn attack-path documentation from a picture made after the fact into a control for what happens next.
For the web application, sensitive database, and domain resource, the responsible decision isn't to follow every arrow. It is to classify each arrow. The confirmed observations remain distinct from the hypotheses. The database carries a visible sensitive-system label and the applicable scope restriction.
Each proposed action names its prerequisites and expected evidence. When evidence conflicts, the edge remains unresolved until a permitted validation addresses the conflict. That artifact can guide attack preparation, support later reporting, and show exactly why the team acted or declined to act.
Attack-path documentation is valuable because it makes both technical dependencies and engagement boundaries visible before an action is taken. A diagram can show systems, relationships, sensitive targets, and scope limits. A storyboard can show the order of preconditions, actions, evidence, and outcomes. Neither one proves a route merely by making it look plausible.
You still need to distinguish observed facts from proposed steps and validate each important connection with evidence appropriate to the claim. When a path crosses a sensitive system, authorization, timing, and operational impact remain part of the technical decision. The result is more than a cleaner picture. It is a defensible basis for choosing the next bounded action and reporting exactly what the engagement did and did not demonstrate.
Before moving on, ask yourself: What should happen next? Use the companion episode reference for exact terminology, named tools or standards, source-derived evidence checkpoints, and review material for Documenting Attack Paths, Dependencies, and Scope Limits. This lesson was produced by Bare Metal Cyber dot com.