KCSA flashcards
KCSA · Build recall with a focused review session.
33 cards. Read the term and the definition here. The deck at the top of this page is optional.
- Four Cs model
- Cloud, Cluster, Container, Code. Each layer inherits trust from the layer below; a compromised cloud provider undermines all higher layers. Used to structure defense in depth for cloud native systems.
- Shared responsibility model
- Security duties split by service model. In IaaS the customer patches the OS; in managed Kubernetes the provider patches the control plane while the customer secures workloads and configuration.
- Defense in depth
- Layering independent controls so no single failure grants full compromise. In Kubernetes this means RBAC, network policy, admission control, and runtime detection all apply concurrently.
- Ephemeral workload assumption
- Cloud native workloads are short-lived and replaced, not patched in place. Controls must be automated, identity-based, and applied at admission rather than configured manually on hosts.
- Confidentiality, integrity, availability
- The CIA triad. In Kubernetes, confidentiality maps to RBAC and encryption, integrity to admission control and image signing, availability to quotas and disruption budgets.
- kube-apiserver role
- The only component that communicates with etcd and the single entry point for all cluster API requests. Authentication, authorization, and admission all occur here.
- etcd encryption at rest
- Configured through EncryptionConfiguration with aescbc, aesgcm, or a KMS provider. Without it, Secrets are stored base64-encoded and readable by anyone with etcd access.
- Kubelet port 10250
- The kubelet HTTPS API endpoint used for pod operations and exec. Anonymous authentication must be disabled and authorization delegated to the API server via webhook mode.
- Kubelet read-only port
- Port 10255 served unauthenticated metrics and pod data. It is deprecated and should be disabled with readOnlyPort set to zero.
- NodeRestriction admission
- Limits a kubelet to reading and modifying only its own Node object and pods scheduled to it. Reduces blast radius when a single node is compromised.
- Container runtime socket risk
- Mounting containerd or CRI-O sockets into a pod grants effective root on the node. Such mounts should be blocked by admission policy.
- API server secure flags
- Set authorization-mode to Node,RBAC, disable anonymous authentication, and enable the NodeRestriction admission plugin. These are baseline control plane hardening settings.
- RBAC additive model
- RBAC grants permissions only; there are no deny rules. Effective permissions are the union of all bound roles, so over-broad ClusterRoleBindings silently expand access.
- Pod Security Standards
- Three profiles: Privileged, Baseline, and Restricted. Baseline blocks known privilege escalations; Restricted adds non-root, seccomp, and dropped capabilities requirements.
- Pod Security Admission labels
- Namespace labels pod-security.kubernetes.io/enforce, audit, and warn select a profile per mode. Enforce rejects violating pods; audit and warn only record or report them.
- NetworkPolicy default behavior
- Pods are non-isolated until a policy selects them. Once selected, only explicitly allowed traffic passes; egress control requires listing Egress in policyTypes.
- Service account token automount
- Tokens are mounted into pods by default. Setting automountServiceAccountToken to false removes the credential where the workload does not call the API server.
- Secrets are not encrypted
- Kubernetes Secrets are base64-encoded by default. Confidentiality requires etcd encryption at rest plus RBAC restrictions on reading Secret objects.
- Admission controllers
- Intercept requests after authentication and authorization but before persistence to etcd. Gatekeeper, Kyverno, and ValidatingAdmissionPolicy enforce policy at this point.
- Hardened securityContext fields
- Key settings: runAsNonRoot, readOnlyRootFilesystem, allowPrivilegeEscalation false, capabilities drop ALL, and a seccompProfile of RuntimeDefault or Localhost.
- STRIDE threat categories
- Spoofing, Tampering, Repudiation, Information disclosure, Denial of service, Elevation of privilege. Applied per trust boundary when threat modeling Kubernetes components and data flows.
- Kubernetes threat actors
- External attacker, compromised container, compromised node, malicious insider with API access, and supply chain attacker. Each has different capabilities and detection signals.
- MITRE ATT&CK for Containers
- A technique catalog covering exposed dashboards, privileged containers, hostPath mounts, and cloud metadata access. Used to map detections and red-team scenarios.
- HostPath escape path
- A pod mounting hostPath or running privileged can reach the node filesystem. Combined with a mounted service account token it escalates to cluster-level access.
- Cloud metadata endpoint
- 169.254.169.254 issues instance credentials. Workloads reaching it can steal node IAM identity; block via egress NetworkPolicy or require IMDSv2.
- Lateral movement detection
- Stolen service account tokens and broad ClusterRoles enable movement between namespaces. API audit logs and admission logs are the primary detection sources.
- Supply chain threat
- Typosquatted, unsigned, or CI-compromised images. Mitigations include digest pinning, signature verification at admission, and provenance attestation.
- Image digest pinning
- Referencing an image by sha256 digest instead of a mutable tag prevents tag-rewrite attacks and guarantees the exact artifact that was scanned and signed is deployed.
- Image signing with cosign
- Sigstore cosign signs container images and stores signatures in a registry. Admission policies verify signatures before allowing a pod to be created.
- SBOM formats
- Software Bill of Materials in SPDX or CycloneDX lists components and versions. Used to trace vulnerable dependencies and satisfy supply chain audit requirements.
- Runtime threat detection
- Tools such as Falco and Tetragon observe syscalls and process behavior. They alert on signals like a shell spawned inside a container or unexpected binary execution.
- CIS Kubernetes Benchmark
- A configuration checklist for control plane, nodes, and policies. kube-bench automates the checks and produces pass/fail evidence for audits.
- Kubernetes conformance
- A program certifying API behavior and portability, not security posture. Passing conformance tests does not demonstrate compliance with any regulatory framework.