Skip to content

Latest commit

 

History

History
64 lines (42 loc) · 2.55 KB

README.md

File metadata and controls

64 lines (42 loc) · 2.55 KB

net-forward

Simple plugin that lest you create a local port listener on your personal machine that redirects to an arbitrary TCP service that the Cluster can see. This is similar to kubectl port-forward without the restriction that forwarding can only use Pods or Services as destinations.

This is intended to be flexible for people debugging network issues, testing whether objects are able to communicate with eachother, and from a security perspective, making it easier for pentesters to target other services on adjacent subnets from the comfort of their own machine.

Examples

Show help:

kubectl net-forward -h

Let the plugin prompt you for information:

kubectl net-forward

Make a local forwarder on port 9999 that redirects to https://10.24.0.1 in the cluster:

kubectl net-forward -i 10.24.0.1 -p 443

Make a local forwarder on port 8888 that redirects to http://169.254.169.254 from within the cluster:

kubectl net-forward -i 169.254.169.254 -p 80 -l 8888

Make a local forwarder inside the "testing" namespace and have it connect to an IP in the "default" namespace (e.g. 10.23.10.3):

kubectl net-forward -n testing -i 10.23.10.3 -p 80

Details

This works the same way as port-forward, when you run this, it will create a Pod running the alpine:socat image and then configure it to redirect to the service of your choosing.

Future plans

  • Rewrite into Go to support common cli stuff
  • Suggest some useful endpoints to users (API endpoint, discovered services, metadata service)
  • Add support for full proxy mode (like ssh -D)
  • Support a service discovery mode (ie nmap)

FAQ

How is this different than port-forward kubectl port-forward requires that you give it a Pod or service and a port. Like kubectl port-forward Pod/my-app 9999 This restriction is useful when you're a cluster-admin but for security testing it's useful to make arbitrary connection to other services that you've discovered.

Why would you want to use this? One example is the case where you wanted to verify whether the network controls you've setup properly prevent a Pod in a namespace from accessing the Pod of another namespace. Maybe you've configured a Network Policy that prevents Default/Pod/appA from accessing Secure/Pod/appB.

For security folks, this lets you proxy communications to any host in the cluster for a variety of things. I personnly like doing it because I have all of my favorite testing tools on my laptop and now can just point it another service in the cluster without having to tool up a Pod or deploy a custom image.