Endlessh: an SSH tarpit

Endlessh is an SSH tarpit that very slowly sends an endless, random SSH banner. It keeps SSH clients locked up for hours or even days at a time. The purpose is to put your real SSH server on another port and then let the script kiddies get stuck in this tarpit instead of bothering a real server.

Since the tarpit is in the banner before any cryptographic exchange occurs, this program doesn't depend on any cryptographic libraries. It's a simple, single-threaded, standalone C program. It uses poll() to trap multiple clients at a time.

Usage

Usage information is printed with -h.

Usage: endlessh [-vh] [-d MS] [-f CONFIG] [-l LEN] [-m LIMIT] [-p PORT]
  -4        Bind to IPv4 only
  -6        Bind to IPv6 only
  -d INT    Message millisecond delay [10000]
  -f        Set and load config file [/etc/endlessh/config]
  -h        Print this help message and exit
  -l INT    Maximum banner line length (3-255) [32]
  -m INT    Maximum number of clients [4096]
  -p INT    Listening port [2222]
  -v        Print diagnostics to standard output (repeatable)

Argument order matters. The configuration file is loaded when the -f argument is processed, so only the options that follow will override the configuration file.

By default no log messages are produced. The first -v enables basic logging and a second -v enables debugging logging (noisy). All log messages are sent to standard output.

endlessh -v >endlessh.log 2>endlessh.err

A SIGTERM signal will gracefully shut down the daemon, allowing it to write a complete, consistent log.

A SIGHUP signal requests a reload of the configuration file (-f).

A SIGUSR1 signal will print connections stats to the log.

Sample Configuration File

The configuration file has similar syntax to OpenSSH.

# The port on which to listen for new SSH connections.
Port 2222

# The endless banner is sent one line at a time. This is the delay
# in milliseconds between individual lines.
Delay 10000

# The length of each line is randomized. This controls the maximum
# length of each line. Shorter lines may keep clients on for longer if
# they give up after a certain number of bytes.
MaxLineLength 32

# Maximum number of connections to accept at a time. Connections beyond
# this are not immediately rejected, but will wait in the queue.
MaxClients 4096

# Set the detail level for the log.
#   0 = Quiet
#   1 = Standard, useful log messages
#   2 = Very noisy debugging information
LogLevel 0

# Set the family of the listening socket
#   0 = Use IPv4 Mapped IPv6 (Both v4 and v6, default)
#   4 = Use IPv4 only
#   6 = Use IPv6 only
BindFamily 0

Build issues

Some more esoteric systems require extra configuration when building.

RHEL 6 / CentOS 6

This system uses a version of glibc older than 2.17 (December 2012), and clock_gettime(2) is still in librt. For these systems you will need to link against librt:

make LDLIBS=-lrt

Solaris / illumos

These systems don't include all the necessary functionality in libc and the linker requires some extra libraries:

make CC=gcc LDLIBS='-lnsl -lrt -lsocket'

If you're not using GCC or Clang, also override CFLAGS and LDFLAGS to remove GCC-specific options. For example, on Solaris:

make CFLAGS=-fast LDFLAGS= LDLIBS='-lnsl -lrt -lsocket'

The feature test macros on these systems isn't reliable, so you may also need to use -D__EXTENSIONS__ in CFLAGS.