Home › Topics › East-West Visibility & Lateral Movement › Responding to East-West

Responding to Suspicious East-West Activity


A signal has fired. A workstation is reaching servers it never touched, on an administrative port, at an hour when nobody should be working. The signals guide taught you to notice this. Now the harder question arrives, and it arrives with your pulse up: what do you actually do?


The instinct of most administrators is to pull the plug — unplug the host, kill the switch port, phone everyone. Sometimes that is exactly right. Often it destroys the evidence you would need to understand what happened. Often it tips off whatever is spreading before you have scoped it. And often it takes down a machine that turns out to belong to a colleague doing legitimate work at an odd hour. A good response is not the fastest one; it is the proportionate and reversible one, taken in an order that preserves your options.


This guide gives you that order: verify, scope, isolate, preserve, coordinate, restore. It is written for the generalist who is the first — and maybe only — responder. And it stays in the defender's lane throughout: what you observe, what you contain, what you preserve, and whom you call. By the end you will have a response checklist and a containment decision matrix you can keep beside the dashboard. You will also have a way to practice all of it before the night it counts.


In this guide

  • First, do no harm: the order matters
  • Step 1: Verify the signal
  • Step 2: Scope the blast radius with flows
  • Step 3: Isolate — proportionate and reversible
  • Step 4: Preserve evidence
  • Step 5: Coordinate
  • Step 6: Restore and watch for recurrence
  • The containment decision matrix
  • The response checklist
  • Drill it before you need it
  • Where to go next

First, do no harm: the order matters


The six steps are in their order for a reason, and the reason is that each early step protects the later ones. Verify before you isolate, so you do not quarantine a colleague. Scope before you isolate, so you isolate the right things and know what else to watch. Preserve before you clean, so you can still answer "how" and "how far" afterward. Skipping ahead feels faster and usually costs you later. The decision matrix makes one exception explicit: a spread moving so fast that containment cannot wait for perfect scoping. Even then you preserve what you can as you go.


One principle sits above all six: reversible beats drastic. Every containment action in this guide can be undone if the signal turns out to be innocent. That is deliberate. A response you can walk back is a response you can take quickly and confidently, because being wrong costs a few minutes rather than a day.


Step 1: Verify the signal


Before anything else, answer one question: is this an admin doing admin things? A startling share of east-west alerts resolve here, at the first check, and resolving one costs a phone call rather than a day of cleanup.


Verification is fast and it does not touch the suspect host. Identify the source in your inventory: whose machine is it, what is its role, is it in a zone that would make the activity normal? Check the calendar: is there a maintenance window, a deployment, a new application launch that explains it? And if you can, ask a human — the person whose machine it is, or the admin who might be working late. Many alerts end with "oh, that was me, I was patching from my laptop instead of the jump host." That is both a closed alert and a habit worth fixing so the alert does not fire the same way next week.


Verification has a time limit, though. If a few minutes of checking does not produce an innocent explanation, stop trying to explain it away and move to scoping. That is especially true if a second signal has fired on the same host or a neighbor. The failure mode here is not over-reacting; it is spending an hour convincing yourself it is probably fine while the thing spreads.


Tip: write down the moment you stopped believing the innocent explanation, and why. That single line — the time and the trigger — anchors your whole timeline later. It is the hardest thing to reconstruct from memory once adrenaline fades.


Step 2: Scope the blast radius with flows


Before you contain, understand what you are containing. The blast radius is the set of hosts potentially involved. And your flow store — built in the monitoring guide — is exactly the tool to measure it. It answers three questions without touching a single suspect machine.


What did the suspect host talk to, and when? Pull every conversation to and from the host over a generous window. Go back further than the first alert, because the alert is when you noticed, not when it started.


-- Everything the suspect host touched, last 3 days, newest first.
SELECT start_time, src_ip, dst_ip, dst_port, bytes
FROM   flows
WHERE  (src_ip = '10.10.1.57' OR dst_ip = '10.10.1.57')
  AND  start_time >= NOW() - INTERVAL '3 days'
ORDER  BY start_time DESC;

Did the hosts it reached start behaving the same way? This is the question that separates a contained problem from a spreading one. Take the destinations the suspect reached and ask whether they have since begun making the same kind of new connections. Look for the same admin port, the same fan-out shape. If host A reached B and C, and now B is reaching D and E the same way, you are watching a spread in progress. In that case, the blast radius is growing.


-- Did anything the suspect contacted then reach NEW hosts on the same admin port?
SELECT src_ip AS now_spreading, dst_ip AS new_target, start_time, bytes
FROM   flows
WHERE  src_ip IN ('10.20.0.11','10.20.0.12','10.20.0.14')  -- what the suspect reached
  AND  dst_port = 5985                                     -- the admin port in question
  AND  start_time >= NOW() - INTERVAL '1 day'
ORDER  BY start_time;

From these you can draw the blast radius on paper. Put the suspect at the center, with a ring of hosts it contacted. Then show whether any of that ring has started its own ring. That is the part that sets your severity. A single host with a quiet ring is one problem. An expanding chain across zones is another, and the decision matrix later treats them differently.


Step 3: Isolate — proportionate and reversible


Now you contain, matched to what you scoped. Containment comes in two grains, and you choose the smallest one that fits the blast radius.


Host-level containment cuts off a single machine. The cleanest network method is to move its switch port into a quarantine VLAN, or to shut the switch port entirely. A quarantine VLAN is an isolated segment with no path to anything but, ideally, the tools you need to investigate. Both options are reversible with a single command. Both leave the machine powered and intact for evidence. Neither requires you to touch the possibly-compromised host itself. Prefer the quarantine VLAN when you want to keep observing the machine's attempts; prefer the port shutdown when you want it silent immediately.


Zone-level containment tightens the inter-zone rules for a whole boundary. If the spread is crossing from users into servers on an admin port, add a rule that blocks that port on that boundary for everyone, not just the suspect. This is the right move when the blast radius is several hosts, or when you cannot yet enumerate every involved machine. It is also right when the spread is faster than you can chase host by host. It is broader, so it carries more disruption. But a temporary block on an admin port between two zones rarely bothers ordinary users. It buys you time to work host by host behind the line.


Whichever grain you pick, write down what you changed and how to undo it, before you make more changes. Reversibility is only real if you can find your way back. And resist the urge to escalate straight to the biggest hammer. Pulling a core switch to stop one laptop takes out the whole building and destroys the flow visibility you are relying on to scope the problem.


Step 4: Preserve evidence


Containment and investigation pull in opposite directions. The fastest way to stop a host is to wipe it. The fastest way to lose the answer is the same act. So preserve before you clean, and preserve the things that survive containment cheaply.


  • Export the flows for every involved host, over a wide window, to a file that lives outside the affected zone. Flow records are your most durable evidence. They are already central, they cannot be altered from a compromised host, and they hold the who-talked-to-whom timeline that memory cannot.
  • Save the logs — firewall session logs, authentication logs, the alerts that fired — with their timestamps, to the same safe place. Note the time zone; mismatched clocks are the classic reason a timeline falls apart later.
  • Leave the host as it is where you can. A quarantined machine that is still running holds volatile state that a power-off erases. Unless something actively harmful is in progress that only a shutdown stops, isolate it on the network. In that case, leave it powered rather than turning it off or starting to "clean" it. The decision to capture more from the host itself is one to make with the security help you are about to call, not ahead of them.

You are not performing forensics — that may be someone else's job — but you are making sure forensics is still possible. A response that stops the spread and leaves no way to learn how it started fixes today and invites the same tomorrow. The rule of thumb is simple: capture what is cheap and durable now. Decide about the expensive, host-level capture together with the help you are about to call.


Step 5: Coordinate


You are not the whole response, and pretending otherwise is its own risk. Two audiences need to hear from you, in different registers.


Security or incident-response help may mean an internal security team or a managed provider. It may mean an external responder on a retainer, or the person who holds that hat part-time for your organization. Reach them early, while the situation is still small, with the facts you have. Tell them what fired, what you scoped, what you contained, and what you preserved. Suspected lateral movement is precisely the kind of event where a second set of eyes with more depth changes the outcome. The earlier they are looped in, the more options remain open.


Management needs a short, honest status: what is affected, what you have done, what you need, and when the next update comes. Resist both the urge to downplay ("probably nothing") and to catastrophize; state what you know and what you do not. The broader shape of running those first minutes is the subject of the first fifteen minutes guide. It covers who to tell, how to size severity, and how to keep a clean timeline. An east-west incident slots straight into that frame.


Step 6: Restore and watch for recurrence


Restoration is not the same as "the alert stopped." Bring things back deliberately, and only once you can answer why the activity happened and are confident the way in is closed. Return quarantined hosts to their normal segment one at a time, not all at once. That way, if something resumes you know exactly which machine it was. Lift a zone-level block only after the individual hosts behind it are cleared.


Then watch, specifically and for a while. The same signals that first fired are now your recurrence detectors. Keep a tighter threshold on the involved hosts and their neighbors for days, not hours. Something that was not fully cleaned tends to announce itself by trying the same move again. Catching that second attempt is far easier than catching the first, because now you know exactly where to look. Only when the watch stays quiet do you loosen back to normal — and even then, update your baselines and allow-lists to reflect anything you learned.


The containment decision matrix


The single hardest judgment in the moment is how much to contain. Scope it to the blast radius you measured in step 2, using this matrix as your guide rather than your instinct.


What you scopedContainment grainTypical actionDisruption
One host, quiet ring (contacted a few hosts, none spreading)Host-levelMove the port to a quarantine VLAN; keep observingOne machine, reversible in seconds
One host, active behavior (still reaching new targets now)Host-level, immediateShut the switch port; preserve flows firstOne machine, offline
Several hosts, same behavior, one zoneZone-level, one boundaryBlock the admin port on that boundary; quarantine the known hostsOne port between two zones; users rarely notice
Spread crossing zones, chain expandingZone-level, multiple boundariesTighten inter-zone rules to the affected zones; call for help in parallelBroader; coordinate before, not after
Spread faster than you can enumerateZone-level, contain firstBlock the offending protocol between zones now; scope behind the lineHighest; the one time to act before perfect scoping

Read the matrix top to bottom as severity rising, and note that the disruption rises with it. You are always trying to sit as high as the evidence demands and no higher — the smallest containment that actually holds the line.


The response checklist


When the moment comes, working memory is unreliable. Keep this beside the dashboard and follow it in order.


StepActionDone when
1. VerifyIdentify the source; check role, zone, calendar; ask a human if you canYou have either an innocent explanation or a decision to proceed — and the time you decided is written down
2. ScopeFlow-query what the host touched and whether those hosts now behave the sameYou can draw the blast radius and say whether it is growing
3. IsolateApply the smallest containment that fits the matrix; record how to undo itThe spread cannot continue and the action is reversible
4. PreserveExport flows and logs to a safe place; leave hosts powered and untouchedEvidence is safe outside the affected zone; clocks and time zones noted
5. CoordinateBrief security/IR help and management with facts and a next-update timeThe right people know, early, and know what you need
6. RestoreReturn hosts one at a time; keep tight thresholds on them; update baselinesThe watch stays quiet and you understand how it started

Drill it before you need it


The first time you run this procedure should not be during a real incident at 3 a.m. The cheapest way to find the gaps is a tabletop. Gather the two or three people who would respond. Describe a scenario aloud: "the reception desktop is reaching three servers on a remote-management port at 2 a.m." Talk through each step without touching anything. Who verifies? Do we know how to move a port to the quarantine VLAN, and does that VLAN even exist yet? Where do flows get exported to, and who has access? Who is our security help, and do we have their number after hours?


Every tabletop turns up something. It might be a quarantine VLAN that was never built, or a flow retention window too short to scope a slow spread. It might be a contact list a year out of date, or a switch command nobody present actually knows. Finding those on a calm afternoon costs an hour. Finding them mid-incident costs the incident. Run one when you finish reading this. Run it again when the responders change. Keep the checklist and matrix where the whole team can reach them. A procedure only one person knows is a procedure that fails the night that person is on vacation.


Where to go next


A good east-west response is calm and ordered. Verify the signal before you disrupt anyone. Scope the blast radius with the flows you already collect. Isolate with the smallest reversible action that holds the line. Preserve the evidence that lets you learn. Coordinate early with security help and management. Restore carefully while watching for the second attempt. None of it is heroic, and that is the point — heroics are what you resort to when you have no procedure.


This completes the arc of the topic. You defined trust zones. You learned to monitor between them and to verify the boundaries hold. You learned to recognize the signals of a spread, and now to respond to one. It all lives together on the East-West Visibility & Lateral Movement pillar page.


Put this into practice with SysPeep Monitor

Response depends on the flows and the baselines being there before the incident, not scrambled together during it. SysPeep Monitor keeps a live map of your network and watches the boundary devices you would isolate at. It alerts on the health and reachability changes that accompany a spread. So when you reach for the scope and containment steps, the ground is already prepared. The free trial installs on any Windows machine.

Download Free TrialLearn About SysPeep