WARNING: This project is a learning exercise. Do not assume your silo'ed processes are 'secure' in any way! There are both known security flaws and those due to my ignorance.

Colexd is a container host, which gets its instructions from the network (through use of colex-cli).

mkdir server
cd server
export GOPATH=`pwd`
go get github.com/twitchyliquid64/colex
go build github.com/twitchyliquid64/colex/colexd
cp github.com/twitchyliquid64/colex/busybox.tar ./ # Use busybox-arm.tar if running on ARM (eg: raspberry pi).
sudo ./colexd --addr localhost:8080 --ip-pool 10.21.0.1/24 # colexd must run as root for now, cuz containers + cbf

export GOPATH=`pwd`
go get github.com/twitchyliquid64/colex
go build github.com/twitchyliquid64/colex/colexd/colex-cli

Given a silo configuration like:

Silo "s" {
  name = "test-sh"

  binary {
    path = "/bin/sh"
    args = ["/script.sh"]
  }

  file "main script" {
    path = "test-script.sh"
    silo_path = "/script.sh"
  }

  network {
    internet_access = true
  }
}

Explanation:

• The silo will be called test-sh. Names are unique on each host, so if you colex-cli up again, the existing silo will be stopped and replaced.
• The binary section specifies what should be run in the container. path is the path to the binary to be invoked, args is it's list of arguments, and env allows you to specify environment variables.
• Each file section allows you to specify files on your current system which will be dropped in the silo. The main script block references a file on your system test-script.sh which will be dropped at /script.sh.
• The network section lets you configure networking within the silo. If internet_access is set to true, iptables MASQUERADE rules + routing table entries will exist to allow the silo to access the internet through the host.

To Start:

./colex-cli --serv <colexd address> <silo-file> up

|-----------|----------|-------|---------------------------|
|   SILO    |    ID    | STATE |        INTERFACES         |
|-----------|----------|-------|---------------------------|
| test-sh   | 3938a44e | UP    | v0-3938a44es (10.69.69.2) |
|-----------|----------|-------|---------------------------|

To stop:

./colex-cli --serv <colexd address> <silo-file> down

To see whats running:

./colex-cli --serv localhost:8080 list
|-----------|--------------|-------|------|------------|
|   NAME    |    ID ()     | CLASS | TAGS | ADDRESSES  |
|-----------|--------------|-------|------|------------|
| test-ping | 7d462601 (0) |       |      | 10.69.69.2 |
|-----------|--------------|-------|------|------------|

Now that you've had a taste, lets spend some time to write a proper config and enforce security properties.

Create two files in your server directory: config.hcl, and authorized-users.

Write the following into config.hcl:

name = "my host"
listener = ":8080"
address_pool = "10.21.0.1/24"

authentication {
  mode = "certs-file"
  certs_file = "authorized-users"
}

Start the server (using the command below) and stop it again once it has started listening. It would have spat out a block of text like this:

Add this section to your configuration file:
transport_security {
  key_source = "embedded"
  embedded_cert = "..."
  embedded_key = "..."
}

Do what it says, and copy paste the transport_security block into the bottom of config.hcl. From now on, the server will use the generated TLS certificate and key.

You can finally start the server for good, using sudo ./colexd config.hcl.

TL;DR: Run colex-cli --serv <addr> enroll and put in the enrollment key printed out in the servers log.

You may have noticed the authentication block in config.hcl. This tells colexd how to verify commands coming from colex-cli are legitimate. By default, mode is set to open, meaning anyone with network connectivity to colexd can run commands and manage silos.

By setting the mode to certs-file and pointing to another file with certs_file, colexd will check the file to verify the certificate presented by colex-cli, as well as add new entries to the file for successful enrollments.

To enroll a client:

1. Start the server if you havent already.
2. Copy paste the enrollment key (not including quotes) - this should have been printed soon after the server started.
3. Run colex-cli --serv <addr> enroll and paste the enrollment key when prompted. But be quick, enrollment is only available by default for 35 seconds! (configurable with blind_enrollment_seconds option).

Docker has a great selection of images, but they need to be exported to a flat tarball before colex can use them.

EG: busybox

export IMGNAME="busybox"

docker pull $IMGNAME
docker create $IMGNAME containerz
docker export $(docker ps --latest --quiet) > "${IMGNAME}.tar"
docker rm $(docker ps --latest --quiet)
docker image rm $IMGNAME

• Server configuration file
• Silo configuration file

This was exploratory / learning exercise, made with heavy reference to https://www.infoq.com/articles/build-a-container-golang & https://medium.com/@teddyking/linux-namespaces-850489d3ccf & https://blog.scottlowe.org/2013/09/04/introducing-linux-network-namespaces/ & https://google.com.