Home › Topics › Network Monitoring as a Security Layer › Who Owns What

Who Owns What: Security Monitoring in a Team of One (or Three)


Every article about security monitoring seems to assume a room. There is a wall of screens and a rotation of analysts. There is also a person whose entire job is to stare at alerts and decide which ones are real. I have read a great many of those articles, and I have never once worked in that room.


Your reality is different. You run the switches and reset the passwords. You answer the phone when the printer jams. And somewhere in the gaps you are also expected to notice if someone hostile is on the network. There is no room. There is you, maybe two colleagues, and a monitoring system that was bought to keep the file server up.


This guide is about making that work honestly. It does not pretend you can run a security operations center in your spare time. Instead it shows you how to define the smallest set of security signals you will actually look at. It shows how to assign the jobs of watching, triaging, and escalating when the same person does all three. It covers who to call when something is beyond you and how to write the one page that makes an incident survivable. It also shows how to tell management what you can and cannot see without overselling or underselling. It ends with a responsibility table sized for a team you can count on one hand.


In this guide

  • The reality, stated without shame
  • Minimum viable security monitoring: the five signals
  • The four hats, and who wears them
  • Escalation paths: who to call, and when
  • The one-page incident sheet
  • Setting expectations with management
  • A RACI table for small teams
  • Where to go next

The reality, stated without shame


A SOC (security operations center) is a team whose job is continuous watching. People work on shift, around the clock, looking at security alerts as they arrive. You do not have one. That is not a failing; it is the normal state of most organizations, and the design consequence is simple. You cannot watch. You can check. A SOC is built around a person being present when the alert fires. Your security monitoring has to be built around a person who shows up later, on a schedule. That person needs the system to have kept good notes in the meantime.


That single reframing drives everything else in this guide. Alerts must be few enough to read in a sitting. Suspicious alerts must wait for you without silently clearing. The first check for each signal must be written down, because you will do it tired. And the decision about who to call must be made now, in a calm moment. Do not wait until 4:45 PM on a Friday when something has started encrypting the file server.


Minimum viable security monitoring: the five signals


The temptation, once you understand that ops data doubles as security data, is to build rules for everything. Resist it. The constraint in a small shop is not data; it is attention. A view with thirty security signals gets skimmed and then ignored. A view with five gets read. So the question is not "what could I detect?" but "what will I honestly look at every week, and what will I get out of bed for?" Here is a defensible starting five.


#SignalComes fromHow often you lookWhat "look" means
1New devices, especially in server or management networksDiscovery and inventoryWeekly review; alert on server/management VLANsClaim or remove every unknown device
2Authentication failure bursts, and any success after a burstDomain, VPN, and Wi-Fi auth logsAlert on bursts; success-after-burst gets a call the same dayGroup by source and account; confirm with the user
3New external destinations and outbound volume per hostFlow data or firewall logsDaily report, weekly reviewExplain every new destination for a workstation
4Monitoring agents, log sources, or backups going quietHeartbeats, log collector, backup reportsAlert immediately when several go quiet togetherCheck reachability and collector health; if hosts are fine, escalate
5DNS failure storms or query volume from one hostResolver logsAlert on bursts; weekly top-clients reviewReadable names or noise? One host or everyone?

Why these five? Between them they cover the stages that almost every intrusion passes through. Something arrives (new device, new destination), and something tries to log in (auth bursts). Something looks around and calls home (DNS, new destinations). And something prepares to do damage (silencing agents and backups). Each one has a cheap first check, and each one is produced by data you already collect. If you are only going to do five things, these five earn their place. The longer catalog in Security Signals Hiding in Your Ops Monitoring is there for when the first five are boring and trusted.


Notice that the table also sorts the five into two speeds. Three of them — new devices, new destinations, DNS oddities — are mostly weekly-review material. That is because the boring explanation is common and a few days' delay rarely changes the outcome. Two of them are get-out-of-bed material: a successful logon right after a burst of failures, and agents or backups going quiet on several hosts at once. Those two are the closest thing ops data has to a fire alarm. They are the only security alerts that should be allowed to interrupt an evening. Deciding that split in advance is what keeps the phone quiet enough that you still answer it.


The four hats, and who wears them


Even in a team of one, security monitoring involves four distinct jobs. Naming them matters, because when one person does all four the jobs blur together, and the blur is where things get dropped. Think of them as hats you put on deliberately, one at a time.


The Watcher

Looks at the security view on a schedule and reads the suspicious alerts that arrived. The Watcher's job is to make sure nothing sits unread. It is a calendar appointment, not a mood.


The Triager

Does the first check on each suspicious alert and decides: boring, or not boring. The Triager works from a written first-check list, within a bounded time, and produces a one-line verdict for each alert. Most alerts die here, which is the point.


The Decider

Takes the not-boring verdicts and chooses what to do. The options are to contain now (disable a port, disable an account), escalate to outside help, or gather more before acting. The Decider needs authority that has been granted in advance. That is because asking permission to unplug a machine mid-incident is how a small incident becomes a large one.


The Communicator

Tells management what is happening in plain language, tells affected users what to do, and keeps a timeline. During a real incident this hat alone can consume a person. That is exactly why it should be a named job and not something that happens between other things. The timeline is the Communicator's most valuable product: a running list of what was seen, what was done, and when. Every outside party you later involve will ask for it first. A timeline written from memory two days later is worth a fraction of one written as it happened.


In a team of one, you wear all four, and the trick is to switch consciously. "I am now triaging" and "I am now deciding" are different mental modes. Mixing them produces both hasty decisions and endless triage. In a team of three, split them: one person watches and triages, a second decides and contains, and the third communicates. Whatever the split, every hat needs a named backup for vacations and sick days. An escalation path that routes through one phone that is switched off on a beach is not a path.


At Tollgate Foods the IT administrator wore all four hats and had never needed to say so. One Monday the odd-hours logon rule fired on an administrator account used at twenty past eleven on Sunday night. The first check said to confirm with the account's owner, who was him. He had been patching, and he knew it. He still had to write a verdict. The verdict he wrote was "confirmed with user", which was true and was also the least useful sentence in the ticket. The manager, reading it, asked who else could have confirmed it, and there was a silence. A colleague in finance now holds the second card, and the odd-hours rule for administrator accounts goes to both of them.


Escalation paths: who to call, and when


The Decider's most important decision is usually "this is beyond us." Small teams get this wrong in both directions: calling for help too late because they wanted to be sure, or never having anyone to call. The fix is to define the tiers now, generically, and write down the trigger for each.


Management

Call when a suspicious alert survives triage and you cannot explain it. Call when any personal or financial data may be involved. Or call when you are about to disable something that people depend on. Management is not there to diagnose. They are there to make business decisions: whether to take a service down, whether to tell customers, whether to spend money. They can only do that if they hear early.


A managed detection provider

Some organizations contract an outside company (often called an MDR provider, for managed detection and response). It does the watching and triaging around the clock. Typically it places its own sensors or agents and has its analysts call you when something is real. If you have one, they become your Watcher and Triager for the signals they see. In that case, your job shifts to being the Decider they call. What they need from you is context: an accurate inventory, a contact list, and honest notes about what is normal on your network. What they cannot do is see data you never send them, so the ops signals in this pillar still matter.


An incident response retainer

A retainer (the "IR retainer" on the sheet below, for incident response) is a pre-arranged agreement with a specialist firm. The firm will investigate and help recover from a confirmed incident. The paperwork is already signed so the call takes minutes instead of days. The trigger is a confirmed intrusion, active encryption, or evidence that data has left. If you have one, the number goes at the top of the incident sheet. If you do not, that is a conversation to have with management now, in calm weather. The general question of when a one-person IT department should bring in outside help is covered in When to Get Help.


Insurers and law enforcement

Two parties you may be obliged to involve. If the organization carries cyber insurance, the policy usually specifies who must be notified, how quickly, and which responders you are allowed to use. The policy may also forbid certain actions, such as wiping affected systems before they are examined. Read the policy before an incident and put its requirements on the sheet. Law enforcement is appropriate when a crime has clearly occurred, and management should make that call with the insurer's guidance. Neither party will help you triage. Both will care very much about what you did in the first hours, which is why the next section exists.


Tip: the most common small-shop mistake is not "called too early." It is wiping and rebuilding a compromised machine before anyone looked at it, because getting the user working again felt urgent. Contain by disconnecting; preserve by leaving it powered on; then call. Rebuilding can wait a day. Evidence cannot come back.


The one-page incident sheet


Everything above collapses into a single page that lives somewhere you can reach when the network is down. It is printed, in a drawer, and in a personal note on a phone. That is because the wiki is on the network you are trying to fix. The page is the difference between an incident that is stressful and one that is chaotic. Here is what goes on it, in order.


  • Who wears each hat, with a named backup for each and a phone number that is not an office extension.
  • The escalation tiers with their triggers and contacts: manager, provider, retainer, insurer, and the person authorized to call law enforcement.
  • The first five containment moves are ones you are pre-authorized to make without asking. They are to disable a switch port, disable an account, block an address at the firewall, isolate a VLAN, or pull a server off the network.
  • What never to do before the responder says so: wipe, reboot, reimage, or pay anyone anything.
  • Where the evidence lives: flow history, DNS logs, auth logs, monitoring history, and how long each is kept.
  • The timeline template: a blank table with time, what was seen, what was done, and by whom. Start filling it in at the first suspicious alert, not after.

A skeleton you can copy and fill in looks like this. Every blank is a decision you can make in a quiet hour; none of them can be made well during the incident.


INCIDENT SHEET  (review every quarter; last reviewed: ______)

HATS            primary            backup             mobile
  Watcher       ____________       ____________       ____________
  Triager       ____________       ____________       ____________
  Decider       ____________       ____________       ____________
  Communicator  ____________       ____________       ____________

ESCALATION      contact            trigger
  Manager       ____________       survives triage / data involved / taking
                                   a service down
  MDR provider  ____________       (if contracted) any confirmed-hostile verdict
  IR retainer   ____________       confirmed intrusion / encryption / data left
  Insurer       ____________       policy says: notify within ____ hours
  Law enforcement  via manager     with insurer guidance

PRE-AUTHORIZED  disable port | disable account | block address at firewall |
                isolate VLAN | disconnect server   (approved by: ____________)

NEVER WITHOUT RESPONDER OK   wipe | reimage | reboot suspect host | pay anyone

EVIDENCE        flows: ____ days   DNS logs: ____ days   auth logs: ____ days
                monitoring history: ____ days   location: ______________

TIMELINE        time | observed | action taken | by whom
                ---- | -------- | ------------ | -------

Setting expectations with management


The conversation with management is the part most admins skip, and it is the one that protects you. Management, left to assume, will assume the monitoring system is a security system and that "we would know." When an incident arrives, that assumption turns into "why didn't the monitoring catch it?" The honest conversation, held in advance, replaces the assumption with a shared understanding of what the layer is and is not.


Keep it to three plain statements and one request. What we can notice: the five signals, described in business terms — unknown devices, password guessing, computers sending data somewhere new, our safety systems being switched off, and computers behaving strangely on the network. What we cannot detect: known malware by name, anything hidden inside encrypted traffic, a stolen password used correctly, or anything on systems we do not monitor. How fast: alerts are read within one business day, urgent ones sooner, and there is no coverage at 3:00 AM unless someone pays for it. The request: the pre-authorizations on the incident sheet, a decision about outside help, and enough storage to keep evidence for a useful period.


This gives management a real choice: accept the coverage as described, or fund more. Either answer is fine. What is not fine is the third option, where nobody decides and everyone assumes. In my experience the third option is the most popular. Put the three statements in writing, get a reply, and keep both. When the incident comes, the conversation starts from "this was inside or outside the scope we agreed," not from blame.


A RACI table for small teams


A RACI table is a simple grid with answers for each task. It shows who is Responsible (does the work). It shows who is Accountable (owns the outcome and makes the call — exactly one person per task). It shows who is Consulted (asked before it happens) and who is Informed (told after). Large organizations use them to untangle committees. Small teams use them for the opposite reason: to make sure a task with only two possible owners actually has one. Here is a starting grid for a team of one to three plus a manager and an outside responder. Adjust the letters to match your reality; the value is in having had the argument.


TaskAdminSecond admin (if any)Manager / ownerOutside responder
Read the security view weeklyR, AR (backup)I—
Triage a suspicious alertR, AR (backup)I (if not boring)—
Contain: disable port, account, or addressRRA (pre-authorized)C
Declare an incidentR, CCAC
Engage the retainer or providerRIAI
Notify the insurerC—R, AC
Contact law enforcementC—R, AC
Communicate to staff and usersCRAC
Preserve and hand over evidenceR, ARIC
Recover and rebuildR, ARIC
Post-incident reviewRCAC
Tune rules and update the sheetR, ACI—

Three of the letters are deliberate choices. Containment is responsible to the admin but accountable to the manager, because the manager pre-authorized it. That is what lets you pull a port without a phone call. That authority also makes the manager own the consequences of having given that authority. Declaring an incident is accountable to the manager on purpose. It is a business decision with insurance and legal consequences, and it should not sit on an admin's shoulders alone. And "second admin" appears as a backup on almost every row. That is the quiet reminder that a team of one needs to borrow a second person from somewhere. That could be a colleague in another department who can follow a written list, if nothing else.


If there is no second admin, write "—" in the column and look hard at the rows where that leaves a single point of failure. The weekly review can survive a missed week. Containment cannot survive nobody being reachable. That is the row that justifies giving a trusted colleague a laminated card and ten minutes of training.


Where to go next


Security monitoring without a SOC is not a lesser version of the real thing; it is a different design, built around checking instead of watching. Keep the signal set small enough to read, and wear the four hats deliberately. Decide the escalation tiers before you need them. Write the one page, and have the honest conversation with management. There is still no room. There is a page in a drawer and a colleague with a card, and that turns out to be enough. The guides that surround this one fill in the rest. Where Monitoring Ends and Detection Begins explains why the Triager's cheap look is what makes the whole thing affordable. And Building a Security View in Your Monitoring System turns the five signals into the dashboard the Watcher reads. Both are part of Network Monitoring as a Security Layer.


Put this into practice with SysPeep Monitor

A team of one needs a monitoring system that keeps good notes while nobody is looking. SysPeep Monitor auto-discovers devices and interfaces. It raises new-device and threshold alerts with dependencies and rules designed to keep noise down. It routes email and other notifications where you choose. It keeps the history on dashboards and reports you can read in a weekly sitting. The hats and the phone calls are yours; the free trial installs on any Windows machine.

Download Free TrialLearn About SysPeep