Spamdetector

A demo of how Wallaroo can be used alongside a chat installation to detect, in near-real-time, the presence of malicious clients.

Howto (Docker-compose)

$ make deps
$ docker-compose up
    (.. wait for docker to download the world ..)
$ tail -f sink_output/sink.log

Open http://localhost:4000 to see the runtime statistics for your pipeline.

Development & running w/out Docker

• Install Erlang 18.x .. 20.x and the development libraries for: ssl, expat, zlib, and unixodbc. MongooseIM will not currently run with Erlang 21.

• Run make deps build-chat

• You now have two directories: mongooseim and amoc.

• The chat server lives in mongooseim and is preconfigured for:
  a) Allowing logins from any <USER>@localhost, where password=<USER>
  b) Sending a copy of every XMPP stanza to a TCP socket defined by the environment variable WALLAROO_TCP_HOSTPORT

• The spam simluation script lives amoc/spambots.erl and can be launched with the run.sh script found in amoc.
  Use CHAT_SERVER_HOSTNAME=127.0.0.1 ./run.sh spambots 1 13 to launch 12 regular users and 1 spammer, as every 13th user will be a spammer.

• Run make setup to set up a Virtualenv in .env

• As instructed, copy the machida executable and wallaroo.py into the virtualenv. If you don't have these, follow the Wallaroo Tutorial to get them.

• Run make run-locally to launch all the required components locally. If you modify anything under mongooseim, you will have to make build-chat to recompile the chat server.

• Take a look at log/local_sink.log to see the spam detector publishing its results.