Skip to content

Poseidon is a python-based application that leverages software defined networks (SDN) to acquire and then feed network traffic to a number of machine learning techniques. The machine learning algorithms classify and predict both the type of device and if the device is acting normally or abnormally.

License

Notifications You must be signed in to change notification settings

eclipsesrcool/poseidon

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Poseidon

License Build Status codecov Codacy Badge Docker Hub Downloads Latest Version @ Cloudsmith

Software Defined Network Situational Awareness

POSEIDON is now BlackDuck 2016 OpenSource Rookie of the year

Poseidon began as a joint effort between two of the IQT Labs: Cyber Reboot and Lab41. The project's goal is to explore approaches to better identify what nodes are on a given (computer) network and understand what they are doing. The project utilizes Software Defined Networking and machine learning to automatically capture network traffic, extract relevant features from that traffic, perform classifications through trained models, convey results, and provide mechanisms to take further action. While the project works best leveraging modern SDNs, parts of it can still be used with little more than packet capture (pcap) files.

Table of Contents

Background

The Poseidon project originally began as an experiment to test the merits of leveraging SDN and machine learning techniques to detect abnormal network behavior. (Please read our blogs posts linked below for several years of background) While that long-term goal remains, the unfortunate reality is that the state of rich, labelled, public, and MODERN network data sets for ML training is pretty poor. Our lab is working on improving the availability of network training sets, but in the near term the project remains focused on 1) improving the accuracy of identifying what a node IS (based on captured IP header data) and 2) developing Poseidon into a "harness" of sorts to house machine learning techniques for additional use cases. (Read: Not just ours!)

Prerequisites

  • A dedicated Linux System or Virtual Machine (A Debian-based distribution is preferred - Ubuntu 16.x is ideal)
    • Currently supported versions for the .DEB install are:
      • Ubuntu 14.04
      • Ubuntu 16.04
      • Ubuntu 18.04
      • Ubuntu 18.10
  • Docker - Poseidon and related components run on top of Docker, so understanding the fundamentals will be useful for troubleshooting as well. A Good Ubuntu Docker Quick-Start
  • ~10GB of free disk space
  • An SDN Controller - specifically BigSwitch Cloud Fabric or Faucet - if you want full functionality.

Note: Installation on OS X is possible but not supported, see the ./helpers/run file (above) as a starting point.

Installing

On Ubuntu, this will download and install our .deb package from Cloudsmith.

sudo usermod -aG docker $USER
sudo apt-get install -y apt-transport-https curl
curl -sLf "https://dl.cloudsmith.io/public/cyberreboot/poseidon/cfg/gpg/gpg.F9E23875C98A1F72.key" | sudo apt-key add -
sudo add-apt-repository "deb [arch=amd64] https://dl.cloudsmith.io/public/cyberreboot/poseidon/deb/ubuntu $(lsb_release -cs) main"
sudo apt-get update
sudo apt-get install poseidon

If you prefer a lightweight package that downloads the images from the latest build at install time, you can alternatively install: sudo apt-get install poseidon-net

Note: The installer has a Demo option in the installation wizard that will deploy and configure the full Poseidon package, the Faucet SDN contoller (and related components like Grafana and Prometheus), mininet, and openvswitch. We suggest the demo install as a starting point if much of this is new to you.

SDN Controller Configuration

If you opt to do a full install (NOT the demo mode), you need to first identify one of the two supported controllers (BigSwitch Cloud Fabric or Faucet). The controller needs to be running and accessible (via network API) by the Poseidon system. We recommend making sure the SDN portion is configured BEFORE the above Poseidon installation, but it's not a hard requirement.

BigSwitch Big Cloud Fabric Configuration

You will need to access the API of your Big Cloud Fabric (BCF) controller using authorized credentials (user/pass) and you will need to add support in your BCF controller for moving mirrored endpoint network data around your network. In BCF, allowing port mirroring for Poseidon is done using 1) the "span-fabric" feature and 2) identifying a switch interface to send the captured ("spanned") traffic out of. The BigSwitch config will need an admin to add:
  • span-fabric: you need to define a fabric-wide port mirroring mechanism and give it a name (e.g. 'poseidon')
  • interface-group: you need to identify which port the mirrored traffic is going to egress from, and name it (e.g. 'ig1')
Span Fabric

Replace <name> with the name of your span-fabric and <interface-group> with the name of your interface-group.

! span-fabric
span-fabric <name>
  active
  destination interface-group <interface-group>
  priority 1
Interface Group

Replace <interface-group> with the name of your interface-group. Additionally fill in the YOUR_LEAF_SWITCH and YOUR_INTERFACE_WHERE_VENT_WILL_RECORD_TRAFFIC_FROM.

! interface-group
interface-group <interface-group>
  description 'packets get mirrored here to be processed'
  mode span-fabric
  member switch YOUR_LEAF_SWITCH interface YOUR_INTERFACE_WHERE_VENT_WILL_RECORD_TRAFFIC_FROM

Poseidon will connect to the BCF controller using its REST API, so you will also need the BCF API hostname or IP address and credentials for the controller. If your controller is an HA pair and has a virtual IP address, we recommend using that virtual address. Also, because Poseidon will be making dynamic filter rule changes we will need an account that has administrative priveleges. (Poseidon only modifies the filter rules of the defined span-fabric, but until BigSwitch has more granular access control options this means admin privs!) Bringing the above configuration requirements together, below is an example of what the relevant parts of your BCF configuration could look like where the span-fabric is called 'poseidon', the user 'poseidon' is defined for API access, and the egress interface is interface '48' on switch 'leaf04' and labelled as interface group 'ig1':

! user
user poseidon
  hashed-password method=PBKDF2WithHmacSHA512,salt=M4534fcV1Ksg_fNm2pGQ,rounds=25000,ph=true,eWNHYUPVAUYosBVRguJnkmAzM

! group
group admin
  associate user poseidon

! interface-group
interface-group ig1
  description 'Mirroring for Poseidon'
  mode span-fabric
  member switch leaf04 interface ethernet48

! span-fabric
span-fabric poseidon
  active
  destination interface-group ig1
  priority 1

BCF is now configured and ready for use with Poseidon.

Faucet Configuration

Poseidon requires at least Faucet version 1.8.6 or higher.

Unless Poseidon and Faucet are running on the same host, Poseidon will connect to Faucet using SSH. So you'll need to create an account that can SSH to the machine running Faucet and that has rights to modify the configuration file faucet.yaml (currently Poseidon expects it to be in the default /etc/faucet/faucet.yaml location and dps must all be defined in faucet.yaml for Poseidon to update the network posture correctly).

Faucet needs to be started with the following environment variables set:

export FAUCET_EVENT_SOCK=1
export FAUCET_CONFIG_STAT_RELOAD=1

If using the RabbitMQ adapter for Faucet (recommended) make sure to also export FA_RABBIT_HOST to the IP address of the host where Poseidon is running.

Faucet is now configured and ready for use with Poseidon.

Usage

NEW: If you have used the .DEB installer previously, it is worth noting that Poseidon is now packaged as a standard Linux service, and ties in nicely to both systemctl and journalctl.

After installation you'll have a new command poseidon available for looking at the status, logs, changing the configuration, or stopping and starting the service.

$ poseidon help
Poseidon 0.3.6, an application that leverages software defined networks (SDN) to acquire and then feed network traffic to a number of machine learning techniques. For more info visit: https://github.com/CyberReboot/poseidon

Usage: poseidon [option]
Options:
    -c,  config        display current configuration info
    -h,  help          print this help
    -i,  info/status   display current status of the Poseidon service
    -l,  logs          display the information logs about what Poseidon is doing
    -L,  system-logs   display the system logs related to Poseidon
    -R,  reconfig      reconfigures all settings (uses sudo, will restart the Poseidon service)
    -r,  restart       restart the Poseidon service (uses sudo)
    -s,  start         start the Poseidon service (uses sudo)
    -S,  stop          stop the Poseidon service (uses sudo)
    -v,  viz/visualize get url to visualize Poseidon with CRviz
    -V,  version       display the version of Poseidon and exit
    -Z,  reset         reset the configuration (uses sudo)

Poseidon makes heavy use of a sister project, vent. With a successful installation you should minimally see a combination of Poseidon and Vent components, to include:

  1. The following 14 containers with a "(healthy)" STATUS listed (NOTE: this is truncated output):
# docker ps
CONTAINER ID        IMAGE                                  COMMAND                  STATUS
8c07adf421fb        cyberreboot/poseidon:master            "/bin/sh -c '(flask …"   Up 2 hours (healthy)
0a4f947f299b        cyberreboot/vent-file-drop:master      "/bin/sh -c '(flask …"   Up 2 hours (healthy)
511f90c6ddd3        cyberreboot/crviz:master               "serve -s build -l 5…"   Up 2 hours (healthy)
fb250044ff17        cyberreboot/poseidon-api:master        "/bin/sh -c '(flask …"   Up 2 hours (healthy)
8e898fd68c08        cyberreboot/vent-network-tap:master    "/bin/sh -c '(flask …"   Up 2 hours (healthy)
552f65d7a982        cyberreboot/vent-rq-worker:master      "/bin/sh -c '(flask …"   Up 2 hours (healthy)
8dbabe78d1b9        cyberreboot/vent-rq-worker:master      "/bin/sh -c '(flask …"   Up 2 hours (healthy)
e076452c1515        cyberreboot/vent-rq-worker:master      "/bin/sh -c '(flask …"   Up 2 hours (healthy)
d0f406f240b1        cyberreboot/vent-rq-worker:master      "/bin/sh -c '(flask …"   Up 2 hours (healthy)
6229e46723a9        cyberreboot/vent-rq-dashboard:master   "/bin/sh -c '(flask …"   Up 2 hours (healthy)
5c695040603b        cyberreboot/vent-redis:master          "docker-entrypoint.s…"   Up 2 hours (healthy)
004e5fdde96e        cyberreboot/vent-syslog:master         "/usr/sbin/syslog-ng…"   Up 2 hours (healthy)
c0cd7c1f881c        cyberreboot/vent-rabbitmq:master       "docker-entrypoint.s…"   Up 2 hours (healthy)
d81f5509628c        cyberreboot/vent                       "/bin/sh -c '(flask …"   Up 2 hours (healthy)

(NOTE: On our test systems it usually takes 1-2 minutes for all containers to spin up)

If you performed the demo installation, you should also see the following Faucet-related containers running (NOTE: Faucet has not yet implemented docker-friendly health checks, so the "(healthy)" reference will not be shown):

196f53632485        grafana/grafana:5.2.1                  "/run.sh"                Up 2 hours
cfb7d68b66e0        prom/prometheus:v2.3.1                 "/bin/prometheus --c…"   Up 2 hours
86f4188d67f6        faucet/gauge:latest                    "/usr/local/bin/entr…"   Up 2 hours
2e186573532e        faucet/event-adapter-rabbitmq          "/usr/local/bin/entr…"   Up 2 hours
c0652a6ccd44        influxdb:1.6-alpine                    "/entrypoint.sh infl…"   Up 2 hours
2a949b5b1687        faucet/faucet:latest                   "/usr/local/bin/entr…"   Up 2 hours
  1. You should see "Poseidon successfully started, capturing logs..." in your syslog output:
# journalctl -u poseidon | grep capturing
Jul 25 15:42:20 PoseidonHost poseidon[4273]: Poseidon successfully started, capturing logs...

To continue to test (assuming demo installation), please see /opt/poseidon/docs/demo.txt, also referenced in the repo above.

Developing

Modifying Code that Runs in a Docker Container

If installed as described above, poseidon's codebase will be at /opt/poseidon. At this location, a .vent_startup.yml file can be edited to point to a fork of the original repository. Develop and commit changes on the poseidon fork and use poseidon -r to reload and see your changes. You can verify that it's building against your fork by doing a docker ps and the poseidon container will be named off of your fork.

Modifying Code that Runs on the Host Machine

To make changes to anything outside of the poseidon subdirectory you will need to build a new .deb and reinstall.

git clone <YOUR-POSEIDON-FORK>
cd poseidon
make build_debian
sudo dpkg -i dist/poseidon*.deb

Related Components

Additional Info

About

Poseidon is a python-based application that leverages software defined networks (SDN) to acquire and then feed network traffic to a number of machine learning techniques. The machine learning algorithms classify and predict both the type of device and if the device is acting normally or abnormally.

Resources

License

Code of conduct

Stars

Watchers

Forks

Packages

No packages published

Languages

  • Python 85.4%
  • Shell 11.7%
  • Makefile 2.3%
  • Dockerfile 0.6%