Most cyberattacks on small businesses aren’t over in a minute. Someone’s password is stolen on a Tuesday. The attacker signs in quietly, reads email for a week to learn who pays whom, sets up a rule that hides replies from the bank, and only then sends the fake invoice with new wiring instructions. Or they get onto one computer, spend days finding the server and the backups, and launch ransomware on a Friday night when nobody’s around.
In both cases there was a window, often days long, when the attacker was inside and nothing had been lost yet. Threat detection is about using that window.
How long attackers go unnoticed
Mandiant, Google’s incident response firm, publishes figures every year from the breaches it investigates. In M-Trends 2026, covering 2025 investigations, the median time an attacker spent inside a network before being discovered was 14 days, up from 11 the year before. Organizations spotted the attack themselves 52% of the time; the rest of the time they heard about it from someone outside, such as a customer, a bank or law enforcement (Google Cloud, M-Trends 2026).
Those figures come from breaches serious enough to bring in a specialist firm, and a small business with nobody watching has no reason to expect to do better. And when nobody inside is watching, the person who tells you is usually a customer or the bank, after the damage is done. Finding out from your bank is the slow way.
Why prevention isn’t enough
Firewalls, email filtering, updates and multi-factor sign-in stop the great majority of attacks, and they’re where most of the effort should go. But each has gaps. A phishing page can be brand new. A patch can come out days after attackers start using the hole. A password can be stolen from another website where the same person used it. The same M-Trends report found that exploited software weaknesses were the most common way in for the sixth year running (32% of intrusions), with voice phishing, where a caller talks someone into handing over access, now second (11%).
Detection assumes something will get through eventually, and makes sure that when it does, somebody knows.
How detection works
Detection tools use a few methods, usually together.
Known patterns
The tool compares activity against signatures of known attacks: a file that matches known malware, network traffic that matches a known exploit. Fast and reliable for anything already catalogued, and blind to anything new.
Unusual behavior
The tool learns what normal looks like and flags departures from it. A user who always signs in from Daytona Beach suddenly signs in from overseas. A computer that normally talks to the file server starts connecting to every machine on the network. A program starts renaming hundreds of files a minute. This catches new attacks that have no signature yet, at the cost of some false alarms.
Threat intelligence
Lists, shared across the security industry and by agencies such as CISA, of addresses, domains and files known to be used by attackers. If something in your network connects to one, that’s worth a look.
Correlation
One odd event might be nothing. A failed sign-in, then a successful one from a new country, then a new mail forwarding rule, all within ten minutes, is a pattern. Better tools join those events up into one alert.
Finding out from your bank is the slow way.
Signs of an attack in progress
You don’t need special software to notice some of these. Mention any of them to whoever looks after your IT:
- A colleague says they got an email from you that you didn’t send.
- Someone’s mailbox has rules they didn’t create, especially rules that forward or delete messages.
- Multi-factor sign-in prompts arrive when nobody is trying to sign in.
- Antivirus or security software has been turned off on a computer.
- A new administrator account appears that nobody set up.
- A computer is suddenly slow, with the fan running hard and nobody using it.
- Files have odd new extensions, or won’t open.
- A vendor or customer asks about a payment change you didn’t request.
The first and the last are often the earliest signs of email compromise, and they come from outside the network entirely.
Where to watch in a small office
The phrase “network threat detection” suggests watching the network cables, and in large organizations that’s part of it. For most small businesses the attacks show up first in three places:
- The accounts. Sign-ins to Microsoft 365, mailbox rules, new admin rights. This is where email compromise starts.
- The computers. Endpoint detection and response software sees what programs actually do and can stop ransomware mid-run.
- The edge of the network. Firewall and VPN logs show attacks against the doors into the office.
Traffic analysis inside the network adds value as the network grows, but if you have to choose, start with accounts and devices. Our guide to network security monitoring tools explains each kind of tool and which ones a small office needs.
Detection needs a response
An alert that nobody reads detects nothing. Before you rely on any detection tool, know who receives its alerts, how quickly they look, and what they’re allowed to do on their own: isolate a computer, disable an account, reset a password. Managed detection and response (MDR), where a provider’s team watches the alerts and acts on them, exists because most small businesses can’t staff that themselves. What happens in the minutes after an alert is the subject of our post on real-time threat detection.
What we put in place
For our Managed IT customers, detection is one of the layers in our cybersecurity service. Endpoint detection and response runs on every computer, with alerts reviewed by our team and threats isolated. Identity threat detection and response watches Microsoft 365 for unusual sign-ins and locks the account down. And if something does get through, we help you contain it, work out what happened and restore from clean backups. If you don’t know whether anyone is watching your accounts and devices today, a free technology review will tell you.