Attacks & threats

5 min readUpdated

What is a botnet?

Quick answer

A botnet is a collection of internet-connected devices that have been compromised and can be commanded remotely by whoever controls them. The owners of those devices are usually unaware: the machine keeps working normally while quietly taking orders. Botnets are rented out and used for distributed denial-of-service attacks, credential stuffing, scraping, click fraud and spam. Their defining property for a defender is distribution: instructions come from one place, but traffic arrives from thousands of unrelated addresses at once.

01How a botnet is assembled

Most botnets are not built by breaking into anything sophisticated. They are assembled by scanning the internet for devices with known vulnerabilities or default credentials and installing a small program, the bot, that phones home for instructions.

Consumer routers, IP cameras, DVRs and other embedded devices are the classic recruits: permanently online, rarely updated, often shipped with a default password, and almost never monitored by their owners. A compromised camera can spend years in a botnet without anyone noticing.

Mirai, the botnet that defined the modern pattern, was built from exactly such embedded and IoT devices and grew to a peak of about 600,000 infections, according to the 2017 USENIX Security study of it. The study's conclusion was that unsophisticated techniques can compromise enough low-end devices to threaten even well-defended targets.

  • Scanning for default or reused credentials on exposed management ports
  • Exploiting unpatched vulnerabilities in embedded firmware
  • Malicious downloads and email attachments on desktop machines
  • Free apps that quietly bundle a proxy component
  • Compromised or abandoned cloud instances, which have far better bandwidth

02How large botnets get

Size is easiest to grasp through the cases that were measured or dismantled. The figures come from the researchers, providers and authorities listed in the sources.

Attack sizes keep climbing. Cloudflare's report for the fourth quarter of 2025 records a 31.4 Tbps attack that lasted just 35 seconds, and attributes that quarter's hyper-volumetric floods to the Aisuru-Kimwolf botnet, an estimated 1 to 4 million infected hosts, mainly Android TVs.

BotnetWhenReported sizeBuilt fromUsed for
Mirai2016–2017Peak of about 600,000 infectionsEmbedded and IoT devicesDDoS attacks
911 S5Dismantled May 2024More than 19 million IP addresses in about 200 countriesComputers infected through free VPN programsA paid residential proxy service
Aisuru2025An estimated 1 to 4 million hostsCompromised devices worldwideHyper-volumetric DDoS, including a 29.7 Tbps attack

03How they are controlled

Every bot needs to receive instructions, and the mechanism it uses is the botnet's biggest weakness. Older designs had every bot connect to a single command-and-control server, which meant seizing that one server ended the whole network.

Modern botnets are more resilient. They use algorithmically generated domain names so there is no fixed address to seize, peer-to-peer designs with no central server at all, or ordinary services such as chat platforms, code hosting and social media as a hiding place for commands.

DesignHow it worksHow it is taken down
Central C2All bots connect to one serverSeize the server; the whole network dies
Domain generationBots compute new domains dailyRegister the domains before the operator does
Peer-to-peerBots relay commands to each otherVery hard: no single point to remove
Third-party servicesCommands hidden in normal platformsRequires the platform's cooperation

04What they are used for

DDoS is the most visible use but far from the only one. A botnet is really just a large pool of internet connections that someone else is paying for, which makes it useful for anything that benefits from looking like many different people.

  • DDoS attacks, including HTTP floods that look like ordinary page views, sold openly as "stresser" or "booter" services
  • Credential stuffing, where each attempt appears to come from a different address
  • Scraping at scale, spread thin enough to defeat per-address rate limits
  • Click and impression fraud against advertising networks
  • Spam relaying and phishing distribution
  • Residential proxy networks, sometimes sold as a legitimate product

05Why they are hard to defend against

A botnet's advantage is that every individual request looks ordinary. One connection from a home broadband address requesting your homepage is exactly what a customer looks like. Only the aggregate reveals the pattern.

This is why rate limiting per address does so little: each bot only needs to send a handful of requests. Ten thousand bots making two requests a second each will pass any sane per-client limit while delivering twenty thousand requests a second to your origin.

A quick look at distribution helps. Grouping requests by /24 network shows whether traffic is concentrated, as a single abusive host would be, or spread thinly across many networks, which is the botnet shape. Treat it as a first look rather than a verdict, because a popular site has a long tail too.

  • Look for one signature repeated across many unrelated addresses
  • Compare against your own baseline, not a fixed threshold
  • Use challenges such as a proof-of-work challenge: a bot on an embedded device usually cannot run JavaScript
  • Corroborate before acting, so one unusual visitor is never treated as an attack
Requests per IPv4 /24 from an access log (tested on a sample log)
awk '{split($1, o, "."); print o[1] "." o[2] "." o[3] ".0/24"}' access.log \
  | sort | uniq -c | sort -rn | head -20

06Botnets for hire, and the law

Most people who launch a botnet attack never built a botnet. They rent one through a booter or stresser site, often marketed as a load-testing service. Law enforcement treats that as what it is: in December 2024 Europol coordinated Operation PowerOFF across 15 countries, taking 27 such platforms offline, arresting administrators and identifying around 300 users, some of whom face prosecution.

Using these services against infrastructure you do not own is a criminal offence in most jurisdictions, whatever the marketing says; is DDoS illegal covers the legal position in more detail.

07Keeping your own devices out of one

  • Change default credentials on anything with a management interface
  • Do not expose router, camera or NAS admin ports to the internet
  • Keep firmware updated, and retire devices the vendor no longer patches
  • Be wary of free VPN and proxy apps that ask for broad network permissions
  • Watch for outbound connections from devices that should have no reason to make them

Questions

Frequently asked questions

It varies enormously, from a few thousand devices to millions. Size matters less than bandwidth: a few thousand compromised cloud servers can produce far more traffic than a hundred thousand home cameras.

Usually not. Well-written bot software keeps the device working normally, because a device that stops working gets replaced or rebooted. Unexplained outbound traffic is the most reliable sign.

Not on its own. Each bot stays under any reasonable per-address limit; the damage is in the aggregate. Detection has to look at patterns across sources rather than at any single one.

Yes. Stresser and booter services are marketed as testing tools, but using one against infrastructure you do not own is a criminal offence in most jurisdictions, and operators and customers alike have been prosecuted.

Sources

Sources and verification

awk query run with GNU awk 5.2 against a sample nginx access log on Debian 12, October 2026.

  1. 01Understanding the Mirai Botnet (USENIX Security 2017) · USENIX
  2. 02Botnet down and administrator arrested in 911 S5 case, FBI says · The Record
  3. 03DDoS threat report for 2025 Q3 · Cloudflare
  4. 042025 Q4 DDoS threat report: A record-setting 31.4 Tbps attack · Cloudflare
  5. 05Operation PowerOFF shuts down 27 DDoS-for-hire platforms · BleepingComputer

Last updated