Kubernetes Detection Pack
46 detection rules for Kubernetes API server audit logs — covering RBAC privilege escalation, container escape, cryptomining workloads, and admission control bypass. Supports EKS, AKS, GKE, and self-managed clusters.
Pack summary
- Detection rules
- 46
- Critical severity
- 12
- Cluster distributions
- EKS, AKS, GKE
- ATT&CK techniques
- 16
What's included
46 rules across 5 threat categories for Kubernetes security.
cluster-admin binding, wildcard roles, ServiceAccount token abuse, kube-system modification.
Privileged pods, hostPID, hostNetwork, host volume mounts, exec into sensitive containers.
Untrusted image registries, mutating webhook modification, Helm chart provenance.
CronJob creation, DaemonSet deployment, ConfigMap with credentials, API backdoor accounts.
API audit log disabling, admission controller removal, etcd direct access.
Detection rules (20 of 46 shown)
Showing 20 representative rules. All 46 rules activate with one click.
| Rule name | Severity |
|---|---|
| Kubernetes cluster-admin ClusterRoleBinding Created | Critical |
| Kubernetes Privileged Container Created | Critical |
| Kubernetes exec into Running Container | High |
| Kubernetes Anonymous Authentication Enabled | Critical |
| Kubernetes ServiceAccount Token Mounted from Sensitive Namespace | High |
| Kubernetes Namespace Created and Immediately Used for Sensitive Workload | High |
| Kubernetes RBAC Role with Wildcards — Broad Permission Grant | High |
| Kubernetes Pod with hostPID or hostNetwork | Critical |
| Kubernetes Secrets Enumerated by Unexpected Service Account | High |
| Kubernetes CronJob Created for Persistence | High |
| Kubernetes etcd Accessed Directly | Critical |
| Kubernetes ValidatingWebhookConfiguration Modified | High |
| Kubernetes Node Taint Removed — DaemonSet Bypass | Medium |
| Kubernetes Image Pulled from Untrusted Registry | High |
| Kubernetes API Server Authentication Failure — Brute Force | High |
| Kubernetes Namespace kube-system Modified by Non-Admin | Critical |
| Kubernetes Pod Termination Rate Spike — Ransomware Pattern | High |
| Kubernetes ConfigMap Containing Credentials Created | High |
| Kubernetes Admission Controller (PodSecurityPolicy) Deleted | Critical |
| Kubernetes API Server Audit Log Disabled | Critical |
Prerequisites
- Kubernetes API server audit logging enabled with an audit policy that captures RequestResponse-level events for sensitive resources
- For EKS: CloudWatch Logs enabled for API server and audit log types in cluster logging configuration
- For AKS: Diagnostic settings configured to send kube-audit logs to Log Analytics or Event Hub
- For GKE: Cloud Logging enabled for the cluster with API server audit logs forwarded
Kubernetes Detection Pack: frequently asked questions
What Kubernetes log sources does this detection pack use?
The Kubernetes detection pack uses the Kubernetes API Server audit log as its primary source. The audit log captures every request made to the API server — authentication, RBAC decisions, resource creation, modification, and deletion. Runtime security signals from Falco (if deployed) are also ingested for node-level detections. Cloud provider Kubernetes services (EKS, AKS, GKE) have specific audit log delivery mechanisms.
How does ManySignal ingest Kubernetes audit logs?
For cloud-managed Kubernetes (EKS, AKS, GKE), audit logs are streamed through the cloud provider's native log delivery service (CloudWatch, Azure Monitor, Cloud Logging). For self-managed clusters, ManySignal deploys a lightweight log forwarder as a DaemonSet that tails the API server audit log file and ships events. Both paths produce the same normalised event schema.
Can ManySignal detect container escape attempts?
Yes. Container escape detection covers privileged containers, hostPID/hostNetwork pod specs, volume mounts targeting host filesystems (/, /proc, /sys), and exec-based access to privileged containers. Runtime-level escape attempts (ptrace, kernel module loading) require Falco integration for syscall-level visibility.
Does this pack cover Kubernetes supply chain attacks?
Yes. Supply chain coverage includes: images pulled from untrusted registries, InitContainer scripts that download and execute payloads, mutating webhooks that modify pod specs at admission time, and Helm chart deployments from unverified repositories. Image signing verification via Cosign or Notary v2 integration is also supported.
How does ManySignal handle managed Kubernetes differences (EKS vs GKE vs AKS)?
ManySignal normalises audit events from all three managed Kubernetes services into a common schema. Cloud-specific variations (EKS uses CloudWatch, AKS uses Azure Monitor, GKE uses Cloud Logging) are handled by the cloud-specific connector. The detection pack rules run against the normalised schema and are identical across all three providers.
Detect Kubernetes privilege escalation and container escape
46 Kubernetes detection rules covering RBAC abuse, container escape, supply chain attacks, and cluster control plane tampering.