Skip to content

KCSA · Domain 1: Overview of Cloud Native Security · 14% of the exam

Task 1.4: Isolation Techniques

Namespaces, node pools, sandboxed runtimes, Linux kernel security features and multi-tenancy models, and how strong each boundary is.

Study it

  • Isolation: containers versus virtual machines, sandboxes and multi-tenancy

    Partly covered by: Containers vs VMs

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 SaaS provider gives each tenant its own set of nodes, using taints and node selectors so that no two tenants' Pods share a node, while all tenants still use one cluster control plane. According to the Kubernetes multi-tenancy guide, which risk does node isolation alone leave in place?

  1. ANoisy-neighbor contention for node resources between tenants becomes worse than before
  2. BA container escape still gives direct access to the containers of other tenants running on the same node
  3. CThe kubelet and API server are still shared, so an attacker could use their permissions to move laterally
  4. DPer-tenant billing becomes harder to work out than with sandboxed containers
Show the answer and why
  • ANoisy-neighbor contention for node resources between tenants becomes worse than before

    Incorrect

    The guide says node isolation reduces the noisy tenant issue, because all Pods on a node belong to one tenant.

  • BA container escape still gives direct access to the containers of other tenants running on the same node

    Incorrect

    With node isolation every Pod on a node belongs to one tenant, so an escape reaches only the containers and volumes on that tenant's node.

  • CThe kubelet and API server are still shared, so an attacker could use their permissions to move laterally

    Correct

    The guide warns that with node isolation the kubelet and the API server are still shared services, which a skilled attacker could use to move laterally to other tenants' workloads.

  • DPer-tenant billing becomes harder to work out than with sandboxed containers

    Incorrect

    The guide says node isolation is easier to reason about from a billing standpoint than sandboxing.

Node isolation dedicates nodes to one tenant, which limits what an escape exposes and reduces noisy neighbors. The multi-tenancy guide still flags shared services: the kubelet and the API server can be used to move laterally, and suggests compensating controls if that matters.

Question 2 · choose 1

A team loads a custom AppArmor profile on its nodes and applies it to a container. Tests show that file writes the profile should forbid still succeed, while the system log records each violation. What explains this?

  1. AThe container must set allowPrivilegeEscalation to false before AppArmor rules take effect
  2. BAppArmor filters only system calls, so blocking file writes needs a seccomp profile instead
  3. CThe container also needs an SELinux label before an AppArmor profile can block anything
  4. DThe profile is loaded in complain mode, which reports violations without blocking them
Show the answer and why
  • AThe container must set allowPrivilegeEscalation to false before AppArmor rules take effect

    Incorrect

    allowPrivilegeEscalation controls the no_new_privs flag. AppArmor profiles are applied through the container's AppArmor profile setting, independently of that field.

  • BAppArmor filters only system calls, so blocking file writes needs a seccomp profile instead

    Incorrect

    Seccomp is the syscall filter. AppArmor is a mandatory access control module that restricts access to resources such as files through per-program profiles.

  • CThe container also needs an SELinux label before an AppArmor profile can block anything

    Incorrect

    SELinux is a separate Linux security module. An AppArmor profile does not depend on SELinux labels to enforce its rules.

  • DThe profile is loaded in complain mode, which reports violations without blocking them

    Correct

    The security checklist says each AppArmor profile runs either in enforcing mode, which blocks access to disallowed resources, or in complain mode, which only reports violations.

AppArmor profiles have two modes. In enforce mode the kernel blocks what the profile disallows; in complain mode it only logs. A profile that logs every violation yet lets the writes through is running in complain mode.

Question 3 · choose 1

Which Linux kernel security feature works by assigning security labels to objects for more manageable policy enforcement?

  1. Aseccomp
  2. BAppArmor
  3. CSELinux
  4. DLinux capabilities
Show the answer and why
  • Aseccomp

    Incorrect

    seccomp filters which system calls a process can make.

  • BAppArmor

    Incorrect

    AppArmor restricts the access privileges of individual programs through profiles.

  • CSELinux

    Correct

    The Kubernetes page on kernel security constraints describes SELinux as assigning security labels to objects for more manageable security policy enforcement.

  • DLinux capabilities

    Incorrect

    Capabilities divide root privileges into separate units that can be granted or dropped.

Kubernetes can use three kernel security modules and features: seccomp for syscall filtering, AppArmor for per-program access restrictions and SELinux for label-based policy, each configured through the securityContext.

Question 4 · choose 1

In the Kubernetes multi-tenancy guide, what does the term "hard" multi-tenancy usually describe?

  1. AStrong isolation, often for tenants that do not trust each other
  2. BTenants that share one namespace with separate RBAC roles
  3. CA cluster that has one dedicated tenant per node pool and no other controls
  4. DTenants that are all teams of the same company
Show the answer and why
  • AStrong isolation, often for tenants that do not trust each other

    Correct

    The guide says hard multi-tenancy implies strong isolation and is often used where tenants do not trust each other.

  • BTenants that share one namespace with separate RBAC roles

    Incorrect

    Sharing a namespace gives weak boundaries; the guide calls weaker isolation soft multi-tenancy.

  • CA cluster that has one dedicated tenant per node pool and no other controls

    Incorrect

    Node pools are one technique; the term describes the strength of isolation, not a specific setup.

  • DTenants that are all teams of the same company

    Incorrect

    Multi-team tenancy describes who the tenants are, not how strong the isolation is.

The guide contrasts "hard" multi-tenancy, strong isolation typical when tenants do not trust each other, with "soft" multi-tenancy, weaker isolation, while noting there is no single definition of either term.

Practice domain 1 →Practice all domains →