Quick answer
A botnet is a group of internet-connected devices that malware has taken over and that one operator controls remotely. To a website it looks like many unrelated visitors from many countries, so it shows up as a cluster of requests with a shared signature, not as one bad address.
Formula: Lift = signature share in the spike ÷ signature share at baseline
What is a botnet?
A botnet is a group of internet-connected devices that malware has taken over and that one operator controls remotely. The name joins "robot" and "network"; each infected device is a "bot", and the person running the network is the bot herder. NIST's glossary describes it as a set of infected computers that a criminal controls remotely and uses together. The devices are usually ordinary ones, such as home computers, routers and cameras, and their owners often do not know.
For a website owner, the point is what a botnet looks like from the outside. Its requests come from thousands of unrelated addresses, in many countries, on normal-looking consumer connections. In your reports it can pass for a crowd of strangers.
How a botnet is built and run
- Infection. Malware spreads through phishing emails or by scanning for devices with unpatched weaknesses or default passwords. It targets whatever is vulnerable, not specific victims.
- Control. Infected devices take orders from a command-and-control server run by the herder. Some botnets avoid a central server: the bots find each other and pass commands along, which is harder to shut down.
- Use. The herder sends the same instruction to many bots at once, or rents the botnet out. CISA notes that compromised devices may be rented to other attackers, sometimes through "attack-for-hire" services that let people with little skill run attacks.
Scale is the reason botnets matter. The 2016 Mirai botnet, mostly embedded and Internet-of-Things devices, reached about 600,000 infected devices at its peak and was used for large distributed denial-of-service (DDoS) attacks in late 2016. The researchers who studied it concluded that simple infection methods against weak, low-end devices were enough to threaten well-defended targets.
What botnets do to a website
DDoS is the best-known use, but a site can meet a botnet in other ways, and each leaves different marks in analytics:
| What the botnet is doing | What you may see in your data | Where to look |
|---|---|---|
| Flooding the site (DDoS) | A sharp, short spike in requests; many may never run your script, so analytics shows little or none of it | Server logs and your host's monitoring first; analytics only for bots that execute JavaScript |
| Testing stolen logins (credential stuffing), fake sign-ups | A burst of sessions on the login or sign-up page with no later pages | Entry page and session depth |
| Click and ad fraud | Paid visits that arrive and leave without doing anything | Channel and campaign sources, bounce and duration |
| Scraping | Many pageviews across your catalogue pages, from many sources | Page report, session depth |
Scraping is one use among several; for how to deal with it, see how to prevent web scraping.
Why botnet traffic looks like many unrelated people
A single bot is easy to spot. It sends thousands of requests from one address, with one user agent, at a speed no person could keep up. A botnet defeats those tests by spreading the work.
Sample data: a botnet of 600,000 devices sends 5 requests each. That is 600,000 × 5 = 3,000,000 requests, yet no address sends more than 5. Per-address counting sees a normal visitor. Each device has its own country, internet provider and sometimes even its own browser version, so the traffic does not group by place either.
What the nodes do share is their behaviour, because one program drives them. They often hit the same page, send the same headers, run the same browser version, stay for the same fraction of a second and touch nothing. A botnet shows up as a cluster of requests with a common signature, not as one bad visitor. You find it by grouping, not by hunting for a single culprit.
A worked example: finding the cluster
Sample data: a site normally gets about 400 sessions in the 14:00 hour. Last Tuesday at 14:00 it got 3,000. You group the spike hour's sessions by browser and version, entry page, number of events and duration.
| Group | Sessions | Share of the hour |
|---|---|---|
Chrome 131 on Windows, entry /login, 0 events, duration 0 s | 1,860 | 1,860 ÷ 3,000 = 62.0% |
| Everything else | 1,140 | 38.0% |
For comparison, the same signature in the same hour a week earlier (400 sessions) was 12 sessions, or 3.0%. The signature is now 62.0 ÷ 3.0 = 20.7 times as common as normal.
The 1,860 sessions come from 38 countries. The biggest country has 148 sessions, 148 ÷ 1,860 = 8.0% of them, and the average is 1,860 ÷ 38 = 48.9. No country stands out, which is exactly what a botnet does and what a marketing win does not.
The three signals together are the evidence: one dominant signature, spread across many places, and no interaction. Any one alone is weak. A viral post also spikes and spreads across countries, but its visitors do not share a browser version and an entry page, and some of them scroll and click. A reasonable rule, to adjust to your own traffic: if more than half of a spike shares one signature whose normal share is in single digits, treat it as automated until shown otherwise.
Bot, crawler and botnet
| Term | What it is | Identifies itself? | How to recognise it |
|---|---|---|---|
| Bot | Any automated visitor | Sometimes | Behaviour, user agent, network, automation signals |
| AI crawler | A declared bot from an AI company that fetches pages | Yes, with a named user agent | Published user-agent name and, for some, a verified network |
| Botnet | Many compromised devices controlled together | No; looks like ordinary visitors | A cluster with a shared signature, not one visitor |
Common mistakes
- Reading a multi-country spike as a campaign win. Check what the visitors do before you celebrate.
- Looking for one bad address or one bad user agent. A botnet has none, so any single-feature filter lets most of it through or catches innocent people.
- Assuming a botnet means DDoS. The same botnets send spam, commit click fraud and steal credentials, so the damage may be quiet.
- Expecting analytics to see every request. Flood traffic that never runs your script is invisible to a page script. Check server logs.
- Counting bot sessions in your conversion rate. Bots sit in the denominator and pull the rate down. Keep recognised bots out of human reports.
What TapCub can and cannot show about botnets
Suppose the example above happened on your site and you want to know whether to trust last Tuesday's numbers.
- Open Web analytics › Bots. New projects default to detecting bots and reporting them separately, so recognised bots go only to this report and do not inflate the human overview. The All bots tab lists each bot with its operator, purpose, requests, pages visited, most common status code, the detection layers it matched and a trust level.
- Remember what the page script can see. It sees only bots that execute JavaScript. A browser with automation signals, such as
navigator.webdriveror a headless user agent, is classified as a bot and counted separately. A botnet node that behaves like an ordinary browser may not be marked that way, and traffic that never runs the script is not seen at all. - Look for the cluster in the human data. In Web analytics › Realtime, the feed splits people from bots. In Web analytics › Audience, click a browser version or a country to filter, then open Web analytics › Sources to see where the filtered traffic came from. In the Sessions list, compare entry page, event count and duration, as in the table above.
- Cover what the script cannot see. Requests that do not run JavaScript need server collection or a log import to show up. Log import keeps only lines classified as bots.
- TapCub does not block anything. It identifies, separates and counts traffic. Any blocking happens elsewhere, in your host, CDN or firewall. If a row on the bots page is wrong, report a false positive for review.
See also: Smurf Attack: How It Works and Why It Is Rare Now.
See also: What Is a Bot Farm? Definition and the Traffic Signs It Leaves.
Sources
- NIST, Computer Security Resource Center Glossary: botnet.
- CISA, Understanding Denial-of-Service Attacks.
- Antonakakis et al., Understanding the Mirai Botnet, USENIX Security 2017.
- TechTarget, What is a botnet?
Was this explanation clear?
We use your answers to fix and expand this page.
Couldn’t send. Please try again in a minute.
Thanks for the feedback. We’ll use it to improve this page.You’ve already sent feedback on this page. Thanks.

Data Analyst
Explores traffic sources, conversion funnels, and retention, using clear metric definitions and concrete examples to explain what the numbers mean.