bot-cooker
A defensive HTTP tarpit that makes automated scanning expensive.
bot-cooker runs on a hostname of your own and answers the scripts that sweep
the internet for leaked .env files, open .git directories, database dumps
and admin panels. It answers them slowly, with decoy content convincing enough
to keep them reading, so each attempt costs minutes of their time and returns
nothing of value.
Documentation ·
Quick start ·
How it works ·
[Contributing](CONTRIBUTING.md)
Philosophy
Automated attacks are cheap. A scanner tries the same few hundred paths on
every address it finds, and a blocked request costs it nothing: it simply moves
on to the next target. Defence that only says "no" leaves that economy intact.
bot-cooker takes the opposite approach. It protects infrastructure by raising
the cost of attacking it: every request a scanner sends to a trap holds one
of its connections for minutes and yields only decoys. Spread across many
traps, that is less time and money for indiscriminate attacks against everyone.
It does this within strict limits, the same ones defensive security work has
always kept:
- We defend, we do not attack. The trap never scans, never connects out,
never "hacks back". It only answers connections that clients chose to open.
- Time is the only cost. Responses waste a scanner's time and nothing
else: no exploits, no malware, nothing aimed at anyone's machine, no attempt
to track the people behind a scanner.
- Decoys are decoys. Every credential, record and name is generated; none
is anyone's real data or leads to a real system.
- Your own infrastructure only. Traps belong on servers and hostnames you
control and that nobody depends on.
Read more in Ethics and responsible use.
What a scanner meets
Each instance derives a whole imaginary organisation from a secret key: a
company, its staff, an application stack (Laravel, Symfony, Django, Rails,
Express, Spring Boot, ASP.NET Core, WordPress, Joomla or Drupal), a web server
(nginx, Apache, OpenResty, Caddy, IIS or LiteSpeed), real released versions,
habits and timing. Everything it serves is rendered from that persona, so the
pieces agree with each other, and two instances look like two unrelated
systems:
[... continued on GitHub]