Skip to content

KCSA · Domain 4: Kubernetes Threat Model · 16% of the exam

Task 4.1: Kubernetes Trust Boundaries and Data Flow

Where trust changes in a cluster: users to the API server, the control plane to nodes, Pod to Pod and workloads to external services.

Study it

  • Trust boundaries and data flow

    Lesson coming

Sample questions

Try each one before opening the answer. Every option is explained, with the CNCF documentation page that proves it.

Question 1 · choose 1

A cluster's worker nodes sit in a remote site, and traffic between the control plane and the nodes crosses a network the company does not trust. Traffic from the nodes to the control plane already uses TLS with client credentials. What is true of the API server's own connections to the kubelets, and what should be configured?

  1. AThey are already protected by the API server's kubelet client certificate, so nothing more is needed
  2. BThey run through SSH tunnels by default, which already protect them over untrusted networks
  3. CBy default they do not verify the kubelet's serving certificate; set --kubelet-certificate-authority
  4. DThey use plain HTTP unless kube-proxy runs in IPVS mode on the nodes
Show the answer and why
  • AThey are already protected by the API server's kubelet client certificate, so nothing more is needed

    Incorrect

    A client certificate authenticates the API server to the kubelet; it does not make the API server check who answers, so the connection can still be intercepted.

  • BThey run through SSH tunnels by default, which already protect them over untrusted networks

    Incorrect

    SSH tunnels are not the default and are deprecated; the Konnectivity service is their replacement.

  • CBy default they do not verify the kubelet's serving certificate; set --kubelet-certificate-authority

    Correct

    The documentation says the API server does not verify the kubelet's serving certificate by default, which makes the connection open to man-in-the-middle attacks; --kubelet-certificate-authority provides the root bundle to verify it.

  • DThey use plain HTTP unless kube-proxy runs in IPVS mode on the nodes

    Incorrect

    The API server reaches the kubelet's HTTPS endpoint directly; kube-proxy is not involved in this path.

Node-to-control-plane traffic is secured by default, but the API server-to-kubelet connection checks no serving certificate unless --kubelet-certificate-authority is set. Over untrusted or public networks the documentation asks for that flag or, failing that, tunneling.

Question 2 · choose 1

An engineer reaches a billing Service from a laptop through the API server's proxy, using a URL such as /api/v1/namespaces/shop/services/https:billing:443/proxy/. The control plane runs in a cloud network and the nodes in a branch office over the internet. What does the Kubernetes documentation say about the leg from the API server to the Service?

  1. AThe API server authenticates to the Service with its kubelet client certificate
  2. BWith the https: prefix too, the API server neither validates certificates nor sends credentials
  3. CThe leg inherits the TLS session and identity of the engineer's request to the API server
  4. DThe API server refuses to proxy to Services whose Pods run outside the control plane's own network
Show the answer and why
  • AThe API server authenticates to the Service with its kubelet client certificate

    Incorrect

    Connections from the API server to nodes, Pods and Services through its proxy provide no client credentials.

  • BWith the https: prefix too, the API server neither validates certificates nor sends credentials

    Correct

    The documentation says these connections default to plain HTTP; with the https: prefix they are encrypted but the certificate is not validated and no client credentials are provided.

  • CThe leg inherits the TLS session and identity of the engineer's request to the API server

    Incorrect

    The engineer's TLS session ends at the API server; the onward connection is a separate one with the properties the documentation describes.

  • DThe API server refuses to proxy to Services whose Pods run outside the control plane's own network

    Incorrect

    The proxy reaches Services wherever their Pods run; the documentation warns only that the path is not safe over untrusted networks.

The API server's proxy connections to nodes, Pods and Services are plain HTTP by default, so neither authenticated nor encrypted. An https: prefix adds encryption but no certificate validation or client credentials, so the documentation calls the path unsafe over untrusted or public networks.

Question 3 · choose 1

For the Develop lifecycle phase, which architecture does Kubernetes' cloud native security overview suggest adopting to minimize attack surfaces, even for internal threats?

  1. AA zero trust architecture
  2. BA flat network in which internal traffic is trusted
  3. CA single shared administrator account for all services
  4. DA perimeter firewall as the only boundary
Show the answer and why
  • AA zero trust architecture

    Correct

    The overview suggests adopting an architecture, such as zero trust, that minimizes attack surfaces even for internal threats.

  • BA flat network in which internal traffic is trusted

    Incorrect

    Trusting all internal traffic is the opposite of minimizing attack surface against internal threats.

  • CA single shared administrator account for all services

    Incorrect

    Shared powerful credentials widen the attack surface.

  • DA perimeter firewall as the only boundary

    Incorrect

    A perimeter alone does nothing against internal threats.

Zero trust removes implicit trust inside the system: each interaction across a trust boundary is authenticated and authorized. The overview recommends it, with threat models that identify those boundaries, from the Develop phase onward.

Question 4 · choose 1

A cluster's API server presents a certificate signed by the organization's private CA. What does a kubectl client need so that it can verify the server?

  1. AThe API server's private key
  2. BA copy of that CA certificate in its kubeconfig
  3. CAn anonymous request to fetch the CA at each connection
  4. DNothing, because TLS verification is off by default for private CAs
Show the answer and why
  • AThe API server's private key

    Incorrect

    Clients never need the server's private key; they verify its certificate with the CA.

  • BA copy of that CA certificate in its kubeconfig

    Correct

    The documentation says that with a private CA you need a copy of the CA certificate in ~/.kube/config on the client so that the connection can be trusted.

  • CAn anonymous request to fetch the CA at each connection

    Incorrect

    The trust anchor must already be on the client; fetching it over an unverified connection would defeat verification.

  • DNothing, because TLS verification is off by default for private CAs

    Incorrect

    Clients verify the server certificate; a private CA must therefore be trusted explicitly.

The client side of the user-to-API-server trust boundary verifies the server's TLS certificate. With a private CA, that CA's certificate is placed in the kubeconfig so that kubectl can confirm it is talking to the real API server.

Practice domain 4 →Practice all domains →