Choosing Lateral-Movement and Tunneling Tools
An authorized tester controls one workstation inside a lab network. The engagement permits access to one internal application, but the tester’s current system can’t connect to that application directly. Several tools could help. One could route traffic through the workstation.
Another could establish remote execution on a second Windows host. A larger framework could manage sessions and support several kinds of activity. Which one is technically appropriate?
The answer doesn’t begin with the longest feature list. It begins with the authorized outcome. The tester needs a network path to one in-scope service. If that path can be validated through a bounded tunnel or proxy, taking control of another host would add execution, evidence, operational impact, and cleanup without answering a necessary question. A capable tool can still be the wrong choice when its extra capabilities exceed the requirement.
PenTest+ objective 5.2 addresses lateral movement in a scenario and names tools that cover command and control, Windows and SMB operations, protocol libraries, tunneling, proxy routing, remote execution, exploitation, and credential access. The complete product inventory is preserved in the companion episode reference. The names are not interchangeable, and recognizing one does not justify using it. Choose according to the authorized outcome, the mechanism required, the evidence produced, and the cleanup burden.
A useful first step is to classify tools by their broad operational roles. Covenant fits the command-and-control role. Command and control, often shortened to C2, concerns maintaining and directing communications with systems under control during an authorized test.
CrackMapExec, often abbreviated CME, fits broad Windows and Server Message Block operations. Server Message Block, or SMB, is commonly associated with Windows network resources and remote services. Impacket fits the protocol-library role, giving a tester components associated with network protocols rather than forcing every task into one monolithic application.
Netcat belongs in the general network-connection category. sshuttle is associated with tunneling, which means carrying traffic through another connection to reach a network path that isn’t directly available. Proxychains belongs to proxy-based routing.
A proxy chain sends traffic through one or more intermediary proxy hops. Metasploit fits exploitation and session management. PsExec fits remote execution.
Mimikatz fits credential access. These are broad classifications, not exclusive boxes. Tools can have overlapping capabilities, and the behavior selected in a particular engagement matters more than the category label alone.
Those distinctions help separate activities that may appear together but answer different questions. A tunnel can show that traffic from the tester’s system reaches an internal service through a pivot. Remote execution can show that an authorized identity or condition permits actions on another host.
Credential access concerns obtaining or interacting with authentication material. Command and control concerns managing communications and activity. All of these behaviors can contribute to an attack path, but evidence for one doesn’t automatically prove the others.
Suppose a proxy connection reaches an internal web service. That supports a conclusion about network reachability through the pivot. It doesn’t prove that the tester can execute code on the application server, access credentials there, or administer every system on the same subnet. Likewise, successful remote execution on one Windows host doesn’t prove that an unrestricted proxy path exists through it. The result has to be interpreted according to the behavior that actually occurred.
This is why tool selection should follow a decision process rather than brand recognition. First, identify the decision the evidence must support. State what you expect to become true if your understanding is correct.
Choose a bounded action that can test that expectation. Then verify the result using evidence that’s appropriate to the behavior. Finally, document both what the evidence supports and what remains uncertain.
Return to the tester who needs one internal service. The requirement is reachability from the authorized test position through an existing pivot. The predicted state is narrow: traffic routed through the pivot should reach that one service. The tester doesn’t yet need a command shell on another computer, a collection of credentials, or a general-purpose session across the environment. Those may be technically related to lateral movement, but they aren’t part of this immediate requirement.
A SOCKS proxy or another approved tunneling method can be the narrower choice. It provides an intermediate route for supported traffic without treating remote execution as the default solution. Before using it, the tester still has to confirm that the pivot, destination, service, testing time, and traffic type are within scope. The tester should also consider operational constraints. A permitted target may still have timing restrictions or dependencies that make even ordinary-looking traffic inappropriate during a sensitive period.
Next comes prediction. If the tunnel is functioning and the intended route is correct, the connection attempt should travel through the approved pivot and reach the named internal service. If it fails, the failure doesn’t immediately prove the service is secure or unreachable. The problem could be the proxy configuration, the route, the pivot’s own access, the destination selection, or an unmet prerequisite. A failed tool operation and a validated security control are not the same conclusion.
Verification should match the prediction. The tester can compare the expected connection with the observed result and collect suitable evidence of the network behavior. Where packet capture is authorized and appropriate, network traffic can help corroborate which systems communicated.
Service responses or controlled logs may offer additional confirmation. One output message from the chosen tool is evidence, but it isn’t automatically the complete truth. Corroboration matters, especially when a result will support a reportable finding.
Now compare that approach with full remote execution. Remote execution could produce strong evidence about the ability to act on another system, but it changes the question. It may create host activity, use privileges, expose authentication material to a different process or system, and leave artifacts requiring cleanup. If network reachability is all that needs validation, those consequences aren’t justified merely because PsExec, Metasploit, or another capable option is available.
If the tunnel result is inconclusive, the next action isn’t automatically to escalate. First check the assumptions. Was the correct target selected? Was the service expected to be available at that time?
Did the pivot have the route being assumed? Was the proxy-aware application actually using the proxy path? Is the evidence current?
If a boundary or dependency is unclear, clarification is a valid next step. Quietly expanding the activity to another host isn’t troubleshooting. It’s a scope change.
The same discipline applies when choosing among the other tool roles. If the authorized objective is to assess SMB-related access across named Windows systems, a tool oriented toward Windows and SMB operations may fit more closely than a general exploitation framework. If the task requires controlled interaction with particular network protocols, a protocol library may provide a focused approach. If an already-approved session must be managed, a session framework may be relevant. The label guides selection, but the required behavior and evidence make the final decision.
This also changes how you interpret detection. Different choices create different kinds of observable behavior. Proxying changes the apparent network path. Remote execution creates host and remote-service activity.
Command-and-control activity introduces managed communications. Credential-access activity concerns sensitive authentication material. You shouldn’t assume that all lateral movement looks alike to defenders. The selected mechanism influences where evidence may appear, what monitoring might observe, and which artifacts need to be considered during cleanup.
Credential exposure deserves particular care. A technique that involves authentication may place credentials or reusable authentication material into a path that a simple connectivity test wouldn’t require. Mimikatz is broadly associated with credential access, but that association isn’t permission to collect credentials whenever lateral movement is being assessed. The tester needs explicit authorization, a defined purpose, suitable handling requirements, and a reason that lower-impact evidence won’t establish the condition. Sensitive evidence also changes storage, access, and reporting concerns.
Imagine that CME reports successful authentication against an authorized Windows target. It would be tempting to write that the organization has a CrackMapExec vulnerability. That statement misidentifies the problem.
A tool name isn’t a vulnerability. The relevant weakness could involve a credential, an authorization decision, a trust relationship, a configuration, or a software condition that allowed the observed behavior. The tester must determine which explanation the evidence actually supports.
The initial result needs context. Was the credential expected to have access to that target? What privilege was validated? Did the action succeed against the intended system, or was the target information stale?
Was the result limited to authentication, or did it also establish authorization for a particular operation? Authentication evidence alone doesn’t justify claiming unrestricted control. A supported report distinguishes the successful behavior from stronger possibilities that weren’t tested or weren’t observed.
Conflicting evidence makes that distinction even more important. Suppose the tool indicates success, but the expected service behavior isn’t present and the target’s approved records don’t corroborate the event. The right response is to examine configuration, privileges, prerequisites, target identity, and data quality. Perhaps the output was misunderstood.
Perhaps a cached or old result was used. Perhaps access existed but the expected follow-on action required a separate permission. The contradiction calls for bounded validation, not an inflated finding.
You can apply the same reasoning to a failed PsExec attempt. Failure doesn’t prove that remote execution is impossible throughout the environment. It proves only that this attempt, with these conditions, didn’t establish the predicted result. The cause might involve the identity, authorization, target, remote service, configuration, or another prerequisite. Until the cause is validated, the professional conclusion is unresolved rather than secure, vulnerable, or exploitable.
Metasploit presents a related decision challenge because a broad framework may support more than the engagement needs at that moment. Its breadth doesn’t make it the automatic best option. A focused protocol interaction, proxy route, or approved remote-execution check may produce cleaner evidence with fewer unrelated effects.
Conversely, if the authorized requirement genuinely involves managing an established testing session, choosing a framework for that role may be reasonable. Narrowest doesn’t always mean smallest. It means no broader than the objective requires.
Covenant also shouldn’t be reduced to a synonym for lateral movement. Its broad command-and-control role concerns managed communications and activity. That behavior can support lateral movement in an engagement, but the finding still needs to identify the condition that permitted movement or control. Saying that Covenant worked identifies a tool result. Explaining that an authorized test identity reached a specified system through an allowed remote path, under a validated configuration or trust condition, describes the security-relevant behavior.
Netcat, sshuttle, and Proxychains illustrate why a simple-looking network utility can still affect evidence and cleanup. Routing through an intermediary changes how a connection traverses the environment. If the tester forgets that path, later observations can be attributed to the wrong source.
Temporary listeners, forwarding arrangements, proxy settings, and associated test artifacts must be tracked according to the approved plan. Cleanup isn’t merely closing a window. It means accounting for the changes and temporary conditions created by the test.
Impacket illustrates another important distinction. A collection of protocol-oriented components may support several operations, but a report shouldn’t substitute the library’s name for the behavior performed. The evidence might concern authentication, a remote service, execution, or another protocol interaction. Precise reporting says which approved action occurred, which system and identity were involved, what evidence confirmed it, and which underlying condition enabled it. That description remains meaningful even if a defender reproduces the issue with a different tool.
Tool independence improves remediation. If a report says only that Mimikatz, PsExec, or Metasploit worked, a reader may focus on blocking that executable or product name. That response may miss the credential exposure, excessive authorization, reachable remote service, weak trust condition, or vulnerable software that made the behavior possible. Product-specific controls can still have defensive value, but removing one tester tool doesn’t necessarily remove the attack path. Remediation should address the validated weakness.
Consider one final choice. A tester is authorized to determine whether an approved support account can reach and administer a particular Windows server during a maintenance window. One option is to establish a broad command-and-control deployment across several systems.
Another is to perform a bounded validation against the named server using a mechanism suited to the permitted remote behavior. Both choices may sound technically capable. The second fits the stated target and purpose more closely because it doesn’t silently enlarge the systems, activity, or persistence involved.
That answer could change if the engagement requirement changed. If the organization explicitly authorized evaluation of command-and-control detection across a defined group of test systems, the broader mechanism might fit. If the account’s credential use were prohibited, even a focused remote test could be wrong. If the server supported an operationally sensitive process, the maintenance window and safety conditions could outweigh technical convenience. The correct choice depends on scope, evidence, mechanism, and impact together.
The work product should preserve that reasoning. Record the authorized objective, the selected target and identity, the expected result, the mechanism used, and the evidence observed. Separate validated facts from hypotheses.
Note limitations and unresolved questions. Describe any temporary route, session, or host-side condition that needs cleanup. Then frame the finding around the weakness and behavior, not around the fact that a famous tool produced output.
This approach also gives defenders something actionable. They can connect the validated path to privileges, configurations, remote-service exposure, credential handling, or trust relationships. They can decide where remediation interrupts the path most effectively. A clear record of network path and host behavior also helps them understand what monitoring detected and what it missed. The penetration test becomes more useful when the evidence explains both the tester’s decision and the environment’s response.
The names in this objective become useful when you connect each one to a role and then choose according to the authorized outcome. A tunnel or proxy can answer a reachability question without automatically justifying remote execution. A credential-access or command-and-control capability brings different evidence, exposure, detection, and cleanup concerns.
Whatever tool you select, its output is only part of the proof. Validate the behavior, preserve the limits of what you observed, and report the configuration, credential, trust, authorization, or software condition that enabled it. The defensible decision is the narrowest action that answers the approved question while keeping scope, operational impact, verification, and cleanup visible from beginning to end.
One final question should stay with you: Which one is technically appropriate? The companion episode reference provides exact terminology, named tools or standards, source-derived evidence checkpoints, and review material for Selecting Lateral-Movement Tools and Frameworks. This lesson was produced by Bare Metal Cyber dot com.