KODE CLOUD CKS Challenge 2

Take me to the lab!

Please note that the competition status for CKS Challenges is ended. Please do not submit a solution. It will not be scored.

A number of applications have been deployed in the dev, staging and prod namespaces. There are a few security issues with these applications.

Inspect the issues in detail by clicking on the icons of the interactive architecture diagram in the lab and complete the tasks to secure the applications. Once done click on the Check button to validate your work.

Do the tasks in this order:

dockerfile

• Run as non root(instead, use correct application user)
• Avoid exposing unnecessary ports
• Avoid copying the Dockerfile and other unnecessary files and directories in to the image. Move the required files and directories (app.py, requirements.txt and the templates directory) to a subdirectory called app under webapp and update the COPY instruction in the Dockerfile accordingly.
• Once the security issues are fixed, rebuild this image locally with the tag kodekloud/webapp-color:stable

The first two subtasks involve cleaning the Dockerfile...

cd /root/webapp
vi Dockerfile

Change the Dockerfile this:

We are asked to move the application to subdirectory app, so change to COPY instruction in anticipation of this

COPY ./app /opt

Run as non root

A user has been created with RUN adduser -D worker, but we are not switching to it, but are instead switching to root.

Change the line USER root to USER worker

Avoid exposing unnecessary ports. We don't need port 22 (SSH) for the app, so delete the following lines

## Expose port 22
EXPOSE 22

Now save and exit vi

Move the app and related files to a new subdirectory app

mkdir app
mv app.py app/
mv requirements.txt app/
mv templates app/

Rebuild image

docker build -t kodekloud/webapp-color:stable .

Return to home directory

cd ~

kubesec

• Fix issues with the /root/dev-webapp.yaml file which was used to deploy the dev-webapp pod in the dev namespace.
• Redeploy the dev-webapp pod once issues are fixed with the image kodekloud/webapp-color:stable
• Fix issues with the /root/staging-webapp.yaml file which was used to deploy the staging-webapp pod in the staging namespace.
• Redeploy the staging-webapp pod once issues are fixed with the image kodekloud/webapp-color:stable

When running kubesec we can use jq to extract the part of the JSON output that's relevant to identifying critical issues with the scanned manifest. Run without | jq and everything after to see the whole report.

dev-webapp.yaml

kubesec scan /root/dev-webapp.yaml | jq '.[] | .scoring.critical'

Note that CapSysAdmin and AllowPrivilegeEscalation are called out.

Edit the manifest: Remove the SYS_ADMIN capability Set allowPrivilegeEscalation to false Set the container's image to kodekloud/webapp-color:stable (which we built earlier)

Don't recreate the pod yet. There's more to do in the next stage.

staging-webapp.yaml

kubesec scan /root/dev-webapp.yaml | jq '.[] | .scoring.critical'

Note that this has exactly the same issues as dev-webapp.yaml. Perform exactly the same steps as for staging-webapp.yaml.

dev-webapp

Ensure that the pod dev-webapp is immutable:

• This pod can be accessed using the kubectl exec command. We want to make sure that this does not happen. Use a startupProbe to remove all shells before the container startup. Use initialDelaySeconds and periodSeconds of 5. Hint: For this to work you would have to run the container as root!
• Image used: kodekloud/webapp-color:stable (We have already done this above)
• Redeploy the pod as per the above recommendations and make sure that the application is up.

Check what shells are present in the container - shell commands are found in /bin directory and usually end with sh, e.g. sh itself, bash etc.

kubectl exec -n dev dev-webapp -- ls /bin | grep sh

Output:

fdflush isn't a shell, but the other two are. ash is a shell normally packaged with Alpine Linux.

Create a startup probe according to the specification, and ensure the startup probe can run as root. Note that the probes aren't affected by the USER command in the Dockerfile.

Edit dev-webapp.yaml Add the following under securityContext, if it is not already there

runAsUser: 0

Insert the probe

startupProbe:
  exec:
    command:
    - rm
    - /bin/sh
    - /bin/ash
  initialDelaySeconds: 5
  periodSeconds: 5

Now recreate the running pod with everything we changed in step 2 and this step

kubectl replace -f dev-webapp.yaml --force

staging-webapp

Ensure that the pod dev-webapp is immutable:

• This pod can be accessed using the kubectl exec command. We want to make sure that this does not happen. Use a startupProbe to remove all shells before the container startup. Use initialDelaySeconds and periodSeconds of 5. Hint: For this to work you would have to run the container as root!
• Image used: kodekloud/webapp-color:stable (We have already done this above)
• Redeploy the pod as per the above recommendations and make sure that the application is up.

Follow the same steps as for dev-webapp above, adjust staging-webapp.yaml and recreate the pod.

prod-web

• The deployment has a secret hardcoded. Instead, create a secret called prod-db for all the hardcoded values and consume the secret values as environment variables within the deployment.

Examine the deployment manifest to see what this secret is

kubectl get deployment -n prod prod-web -o yaml

We can see there are 3 environment variables with values.

Create a secret for these vars

kubectl create secret generic prod-db -n prod \
    --from-literal DB_Host=prod-db \
    --from-literal DB_User=root \
    --from-literal DB_Password=paswrd

Edit the deployment and change the env section to get the values from the secret

kubectl edit deployment -n prod prod-web

Replace the variables under the env block with

    - name: DB_User
      valueFrom:
        secretKeyRef:
          key: DB_User
          name: prod-db
    - name: DB_Host
      valueFrom:
        secretKeyRef:
          key: DB_Host
          name: prod-db
    - name: DB_Password
      valueFrom:
        secretKeyRef:
          key: DB_Password
          name: prod-db

Test this by pressing the prod-web button above the terminal. After you apply the network policy next, this will no longer work

prod-netpol

• Use a network policy called prod-netpol that will only allow traffic only within the prod namespace. All the traffic from other namespaces should be denied.

Note that all namespaces have a predefined label kubernetes.io/metadata.name which is very useful when creating namespace-restricted network policies.

apiVersion: networking.k8s.io/v1
kind: NetworkPolicy
metadata:
  name: prod-netpol
  namespace: prod
spec:
  podSelector: {}       # apply to all pods in prod namespace
  policyTypes:
    - Ingress
  ingress:
    - from:
        - podSelector: {}                        # any pod...
          namespaceSelector:
            matchLabels:
              kubernetes.io/metadata.name: prod  # ...that is only in prod namespace

Once all the above tasks are completed, click the Check button.