What Shadow IT Looks Like on the Wire
There is something on your network right now that you do not know about. I can say that without having seen your network, because I have never audited one where it was not true. It might be a file-sharing account the sales team set up because the official file server was painful to reach from home. It might be a small storage box under a desk in engineering. Someone bought it on a company card while the request for more shared space was "still in review." It might be a project-tracking service one department has paid for by expense report since before you were hired. Nobody involved thinks they did anything wrong. Most of them think they were being helpful, and in a narrow sense they were.
That is shadow IT: technology adopted and used inside an organization without the IT department's knowledge or approval. The name makes it sound sinister. It is usually ordinary people solving ordinary problems with whatever tool was closest to hand. It is not harmless. It puts company data outside your backups, credentials outside your control, and surprises into your future support queue. But the right first reaction is curiosity, not a crackdown. A crackdown only teaches people to hide the next one better.
By the end of this guide you will be able to recognize the common forms shadow IT takes. You will be able to explain in plain language why each one is a risk. You will understand why network data is the most honest census you have. That means DNS queries, traffic flows, and the list of devices your switches can see. The guides that follow show you how to run that census and what to do with the results.
In this guide
- A definition without judgment
- The common forms shadow IT takes
- Why it happens
- The risks, in plain language
- Why the network is the honest census
- The form, signal, and risk table
- A few boundaries before you start
- Where to go next
A definition without judgment
Start with the word "shadow." It does not mean hidden on purpose. It means outside the light — specifically, outside the light of the inventory, the support process, and the backup schedule. Think of a house where the family keeps buying furniture and nobody tells the person who handles the insurance. Nothing is being stolen and nobody is lying. The insurance policy is simply wrong, and everyone will find that out on the worst possible day. Shadow IT is the same shape of problem: the organization's picture of its own technology is incomplete. The gap only becomes visible when something breaks, leaks, or gets audited.
A simple three-part test tells you whether something counts. First, is it used for work? Second, does it touch company data or the company network? Third, is it missing from IT's list? If all three are true, it is shadow IT. A personal phone on the guest Wi-Fi checking the weather fails the first test and is nobody's business. The same phone syncing the department's shared folder to a personal cloud account passes all three. The distinction matters because it keeps the topic about services and devices, not about what people do on their lunch break.
Notice that the third test assumes a list exists. "Sanctioned" means more than "someone in IT has heard of it" (someone in IT has heard of everything). A sanctioned service has an owner, a support path, and a known way of handling data. It also has a plan for what happens when an employee leaves. Suppose you do not yet have a written list of the cloud services your organization depends on. The guide on building a SaaS dependency inventory is the place to start. Everything in this topic is measured against that list.
The common forms shadow IT takes
Shadow IT is not one thing. It is a family of habits, and each member leaves a different footprint on the network. Here are the forms you will meet most often.
Unsanctioned SaaS
SaaS (software as a service) means an application that runs in someone else's data center and is reached through a browser or a small client. The unsanctioned kind might be a team's own project-tracking tool or a survey service used for customer feedback. It might be a design tool with a free tier, or a note-taking app where meeting minutes quietly accumulate. Or it might be a file-transfer site used to get a large deliverable to a client. On the wire it looks like several workstations looking up the service's name in DNS. That is the Domain Name System, which turns a name into an address and logs the asking. Encrypted web connections to the service's addresses follow, clustered in business hours. One person trying a tool is noise. A whole department using it every day is a pattern.
Personal cloud storage
This deserves its own entry because it is the most common form and carries the most data. A file-sync client installed on a work laptop pushes a copy of everything in a chosen folder to a personal account. It is convenient for working from home, and it is also a complete, unmanaged copy of whatever was in that folder. On the wire a sync client is steady, chatty, and upload-heavy: many small connections through the day and a large outbound volume. That is the traffic shape of a machine pushing data out rather than pulling it in. A complication: the consumer and business versions of the same storage family often use the same names. The network therefore cannot always tell a personal account from the sanctioned company one. That question gets answered in a conversation, not a log. The account, when you ask, is usually named after a pet.
Unknown hardware
The physical branch of the family includes a NAS under a desk. That is network-attached storage — a small box of disks that serves files over the network. There is the five-port mini-switch bought because one wall jack was not enough. There is the personal router providing "better Wi-Fi" in a corner office. Add a wireless printer from home, a smart speaker in the break room, and a streaming stick on the conference-room television. On the wire each shows up as a hardware address whose vendor prefix names a consumer electronics maker rather than your standard laptop vendor. The device often has a DHCP hostname that is the product's default. That is the name a device announces when it asks the network for an address. Sometimes the device answers file-sharing or media ports that no laptop should be answering.
Unofficial Wi-Fi
A personal router plugged into a wall jack creates a wireless network you did not design. So does a phone acting as a hotspot with a laptop bridged into the wired network. That unofficial wireless network may bypass your access controls, extend your network to the parking lot, or hand out addresses that collide with your own. The mechanics of detecting these are a subject of their own, covered by the sibling topic on rogue device detection. Here it is enough to know the form exists and that it usually starts as someone's convenience.
Browser extensions and consumer messaging
Extensions that help with grammar, screenshots, PDFs, or coupons run inside the browser and often send page content to a service to do their work. Consumer messaging apps get used to move files between colleagues because they are faster than email. The distinctive signal here is population, not volume: an extension spreads by word of mouth. So you see many browsers on the floor making small, regular connections to the same unfamiliar name. One device doing that is a curiosity. Forty devices doing it is a deployment nobody approved.
Why it happens
Groundskeepers have a name for the dirt trails that appear across lawns where people actually walk: desire paths. You can fence them off, or you can pave them. Shadow IT is a desire path. It appears where the official route was too slow, too far, or did not exist, and it tells you something true about what people need. The reasons are almost always mundane:
- Slow procurement. The request went in months ago. The project is due Friday. A credit card and a free trial take four minutes.
- Missing capability. There is no sanctioned way to share a large file with an outside client, sign a document, or run a survey, so people find one.
- Convenience. The sanctioned tool exists but is clumsy, requires a VPN, or was never explained; the alternative just works.
- Habit. New hires bring the tools from their last job, and a whole team can inherit a workflow from one person.
- Nobody knew it counted. Many people genuinely do not think of a free web tool or a small storage box as "IT." To them it is stationery.
Keep this list in mind when you find something, because it changes what a finding means. A finding is not evidence of a rule-breaker. It is evidence of a gap, and the gap is often yours to close.
The risks, in plain language
If shadow IT is so ordinary, why care? Because "outside the light" has consequences that only show up later, and by then they are expensive. Here are the ones that matter, stated the way you would explain them to a colleague rather than to an auditor. The auditor's version is the same list with the verbs taken out.
Data outside backups and retention. Your backup jobs copy the file server, the mail system, and the sanctioned cloud services. They do not copy the personal storage account, the desk NAS, or the notes app. When that account is closed, that box's disk dies, or that employee leaves, the data is simply gone. The first anyone hears of it is a request for a document that no longer exists anywhere you can reach.
Credentials in unmanaged places. A team tool signed up for with one person's email and a shared password has no offboarding. When that person leaves, they keep the password and the company loses the admin account. Multiply by every service nobody wrote down.
Support surprises. "The thing stopped working" is a hard ticket when the thing is a service you have never heard of. It holds data you did not know existed, with a login you do not have. You inherit the problem at its worst moment, with none of its history.
Cost sprawl. Three departments pay separately for the same category of tool through expense reports. That is money spent with no negotiation, no volume pricing, and no way to cancel what nobody remembers subscribing to. (The renewal notice goes to someone who left.)
Compliance. Regulated data has rules about where it may live and who may see it. Contracts with customers often promise specific handling. Records requests and legal holds require you to produce documents from every place they exist. You cannot honor any of those obligations for places you do not know about.
Security exposure. Consider a consumer NAS that has never been patched. Also consider a personal router with a default administrator password, or a file-sharing link that was meant to be temporary. Each is a door in your building that is not on the floor plan. You cannot lock a door you do not know is there.
At Northgate Retail the marketing team kept its customer list in a personal file-sharing account. The shared drive could not be reached from the trade shows where the list was actually used. The file-sharing arrangement worked for three years. Then the account's owner left, the account left with her, and the list left with the account. IT heard about it when the new marketing manager raised a ticket asking where the customer list was kept. The honest answer was "nowhere we can reach." The incident review recorded it as "an unplanned change of data custodian."
Tip: when you explain these risks to someone outside IT, pick the one that applies to their situation and stop there. "If that box dies, your team's designs are gone" lands. A six-item list sounds like a lecture, and people stop listening to lectures.
Why the network is the honest census
You could survey people. Surveys are incomplete, because people forget, do not know what counts, or worry about the answer. You could rely on endpoint management software. It knows everything about the machines it manages and nothing about the ones it does not. The ones it does not are the interesting ones. You could read expense reports, which catch paid subscriptions and miss everything free.
Or you could look at the network, which has a property none of those share: everything talks eventually. A cloud service is a DNS name before it is a connection. A storage box has to ask for an address before it can serve a single file. An extension has to reach its service to do its job. None of this requires anyone's cooperation, memory, or honesty. The network does not know or care whether a device is "official"; it just records who asked for what and who talked to whom. That makes it the closest thing you have to a complete census.
Four sources do most of the work, and each answers a different question:
- DNS query logs answer "which names did our machines ask for?" Every service appears here first. Treating your resolver's logs as a sensor is its own topic, covered under DNS as a threat sensor.
- Flow records answer "who talked to whom, how much, and when?" They are the traffic shape without the content, which is exactly the level of detail you want. If flow reports are new to you, start with reading flow reports.
- Switch MAC tables and DHCP leases answer "what is physically here?" They are how you find hardware that never talks to the internet at all.
- TLS server-name hints, where a firewall or proxy records them, answer "which service was this encrypted connection for?" without opening the connection.
Here is the kind of thing a resolver log shows when a service has quietly become part of daily life. The names are invented; the shape is real.
time client query type 09:02:11 10.20.4.31 files.quickshare-example.net A 09:02:14 10.20.4.31 api.quickshare-example.net A 09:04:50 10.20.4.58 files.quickshare-example.net A 09:05:03 10.20.4.17 files.quickshare-example.net A 09:05:03 10.20.4.17 cdn.quickshare-example.net A ... (daily summary) quickshare-example.net distinct clients: 23 queries: 4,180
Twenty-three workstations asking for the same unfamiliar name, every working day, is not one person's experiment. It is a tool a team has adopted, and by now it has a champion, a shared login, and a nickname. You did not need to read a single file or open a single connection to learn that.
The form, signal, and risk table
This table is the summary you will come back to: each form, the signal that gives it away, and the risk that matters most. The later guides explain how to find each signal in practice.
| Form | Typical example | Signal on the wire | Main risk |
|---|---|---|---|
| Unsanctioned SaaS | A team's own project tool | Many hosts resolving and connecting to one unfamiliar name in business hours | Data and credentials outside IT's control |
| Personal cloud storage | Consumer sync client on a work laptop | Steady, upload-heavy traffic from one host to a storage service | Unmanaged copy of company data; no backup or retention |
| Unknown NAS | Storage box under a desk | Consumer vendor prefix; default hostname; serves file ports to many peers | Unpatched device; data that vanishes when the disk does |
| Mini-switch | Desk switch on one wall jack | Several hardware addresses on a single access port | Loops, bypassed port security, unknown devices behind it |
| Personal router / hotspot | "Better Wi-Fi" in a corner office | Device NATing others; second DHCP server; new wireless network | Bypassed access control; network extended beyond your walls |
| Browser extensions | Grammar or screenshot helper | Many browsers making small regular connections to one service | Page content leaving the organization unnoticed |
| Consumer messaging | Chat app used to pass files | Client connections plus periodic file uploads from many hosts | Records outside retention; files outside control |
A few boundaries before you start
Everything in this topic is a census of services and devices. It is not, and must not become, a record of what individual people read, wrote, or watched. The data sources above are chosen precisely because they show the shape of traffic without its content. And the way you use them should keep that boundary intact: aggregate by service and count hosts. Do not build browsing histories of named individuals. This is partly about respect, partly about law in many places, and partly about practicality. A program that feels like surveillance loses the cooperation it depends on, and the desire paths just move somewhere you cannot see.
A short sequence sets you up properly:
- Get the sanctioned list. You cannot find "unknown" without a definition of "known." Even a rough list beats none.
- Write down the privacy boundary. One paragraph, agreed with your manager: what data you will look at, at what level of detail, and what you will never do with it.
- Pick your sources. Start with whichever of DNS logs, flow records, or switch tables you already have. You do not need all four on day one.
- Decide who reviews findings. A finding needs a person and a regular time, or it becomes another report nobody opens.
- Decide what happens next. Before the first finding lands, agree that the default response is a conversation, not a block. The guide on the diplomatic conversation shows what that looks like.
One more piece of honesty: you will find things that are fine. A department's project tool may be genuinely better than anything IT offers. And the right outcome may be to sanction it, not to remove it. Discovery tells you what exists. It does not tell you what to do about it, and pretending otherwise is how a visibility program turns into an unpopular one.
Where to go next
You now have the vocabulary: what shadow IT is and the forms it takes. You know why people create it, why it matters, and why the network is where you look. The rest of this topic is the practice. Finding unsanctioned SaaS turns DNS and flow data into a ranked list of unknown services. Finding unofficial hardware does the same for the boxes under desks. The diplomatic conversation covers what to say when you find something. To turn one-off hunts into a routine, see the program guide. All of them live under Discovering Shadow IT Through Network Data. The thing on your network that you do not know about is still there. It was never hiding; nobody had looked.
Put this into practice with SysPeep Monitor
The hardware half of shadow IT shows up first as a device your monitoring has never seen. SysPeep Monitor auto-discovers the devices and interfaces on your network and polls your switches over SNMP. A new box under a desk appears in the console instead of in a ticket months later. The free trial installs on any Windows machine in minutes.
Download Free TrialLearn About SysPeep