Security, scans and FIPS¶
For architecture across reliability, operations, performance, cost and sustainability as well as security, use the separate Well-Architected assessment: cs scan architecture. It evaluates saved configuration and local evidence, reports unknowns explicitly, and is excluded from the scan all security suite below.
Secure defaults are the point of cloudseed. Every environment follows the same rules, every scan saves a report with a verdict, and FIPS 140 mode is one switch for the whole environment.
The security model¶
%%{init: {"flowchart": {"nodeSpacing": 30, "rankSpacing": 38, "padding": 12}}}%%
flowchart TB
ip["Your IP only<br/>(--allow-ip, update-ip)"] -- "TCP/22" --> fw["Cloud firewall<br/>SG / VPC firewall / NSG"]
fw --> bastion["Bastion<br/>key-only SSH, nftables,<br/>fail2ban, auditd"]
bastion --> priv["Private subnets<br/>workloads, cluster nodes<br/>(egress through NAT)"]
priv --> data["Isolated data subnets (AWS)<br/>no internet route"]| Layer | What cloudseed does |
|---|---|
| Network | Only the bastion (and an optional VPN host) has a public address. The cloud firewall admits SSH from --allow-ip only: IPv4, never 0.0.0.0/0, nothing wider than a /8 and at most two /8s. Workloads live in private subnets behind NAT; AWS adds an isolated data tier with no internet route. Default security groups are stripped (AWS); GCP and Azure end in explicit deny-all rules. |
| Hosts | Key-only SSH, no root login or passwords, latest LTS images, encrypted disks. AWS: IMDSv2 and SSM Session Manager. GCP: Shielded VM with its own least-privilege service account, project SSH keys blocked. Azure: Trusted Launch (secure boot + vTPM) and a managed identity. |
| Provisioning | Ansible hardens sshd (strong ciphers, AllowUsers, MaxAuthTries), installs fail2ban, auditd rules, sudo logging, kernel sysctls, unattended security updates and a default-deny nftables firewall. |
| Logging | VPC or subnet flow logs, NAT and firewall logs (AWS, GCP), Activity Log to Log Analytics (Azure), auditd on the hosts, and cloudseed's own audit trail of every command. |
| Baseline | AWS: CloudTrail (multi-region, validated, KMS), S3 account public-access block, IAM password policy, GuardDuty, Access Analyzer, EBS encryption by default, optional Security Hub. GCP: log retention and optional Data Access audit logs. Azure: optional Defender for Servers / Storage. |
| State | Remote state is versioned, encrypted, private and TLS-only. No secrets in state or outputs. |
| Agents | Credentials are stripped from agents and redacted from their output (Agentic mode). |
cs help security prints the full model; cs explain bastion, cs explain network and cs explain security-baseline show exactly how each part is implemented.
Keep SSH access current¶
When your public IP changes, SSH to the bastion times out. One command re-detects it and changes only the SSH rule:
cs update-ip aws --env dev
cs update-ip azure --env dev --allow-ip 198.51.100.4,203.0.113.0/24 --auto-approve
Re-run the host hardening¶
Provisioning runs after every setup and is idempotent, so you can re-apply it any time, for example after changing something by hand on a host:
cs provision aws --env dev # the bastion (and the VPN host)
cs provision aws --env dev --host vpn # only the VPN host
cs provision vmware --env lab --host k8s # the Kubernetes nodes of a local cluster
cs provision aws --env dev --no-firewall # without the nftables host firewall (and remove an earlier one)
--no-harden, --no-firewall and --no-tools leave out a role; the first two also take back what an earlier run installed. Supported host images: Amazon Linux 2023, Debian 12, Ubuntu 22.04 and 24.04 (including Ubuntu Pro FIPS).
Baselines are singletons¶
Some settings belong to the whole account, project or subscription, so only one environment should manage them:
| Cloud | Setting | For the second environment |
|---|---|---|
| AWS | account-wide half: CloudTrail, S3 public-access block, password policy | --var enable_account_baseline=false |
| AWS | regional half: GuardDuty, Access Analyzer, EBS default encryption, Security Hub | one per account and region; an environment alone in another region adds --var enable_regional_baseline=true |
| GCP | _Default log retention, Data Access audit logs | --var enable_project_baseline=false |
| Azure | Defender plans | enable Defender in one environment only |
An existing GuardDuty detector or Security Hub is never adopted: pass --var enable_guardduty=false or --var enable_security_hub=false. The target references list every variable.
Compliance and vulnerability scans¶
One command per scanner, each with a saved report and a verdict:
| Command | Scanner | What it checks |
|---|---|---|
cs scan cis | kube-bench | CIS Kubernetes Benchmark with the right profile per distro (EKS, GKE, AKS, RKE2, kubeadm) |
cs scan kube | kubescape | NSA and MITRE ATT&CK frameworks by default (add --framework cis-v1.10.0, soc2, ...) with a compliance score |
cs scan images | trivy | vulnerabilities in running workloads: trivy-operator reports when installed, else a one-off scan |
cs scan host | OpenSCAP + SCAP Security Guide | CIS level 1/2 server profile (or --profile stig) on the bastion, the VPN host and local Kubernetes nodes |
cs scan stig | OpenSCAP, kube-bench | DISA STIG on Ubuntu 24.04, Ubuntu 22.04 with Ubuntu Pro and RHEL 8/9 hosts; EKS Kubernetes STIG |
cs scan cloud | prowler | the newest CIS benchmark for your account, project or subscription |
cs scan fips | cloudseed | end-to-end FIPS 140 verification (N/A on a non-FIPS environment) |
cs scan all | all of the above | everything applicable, then a summary |
cs scan all
cs scan kube --framework nsa,mitre,cis-v1.10.0
cs scan host vmware --env lab --host bastion,k8s
cs scan stig aws --env prod --host vpn # the Ubuntu VPN host (the AL2023 bastion has no STIG content)
cs scan cloud gcp --env prod
cs scan reports --last 5
- Reports go to
<workdir>/scans/<kind>-<run>.jsonand.md, raw tool output to<workdir>/scans/raw/, and they show up in the web console's Reports view. - The exit code is 1 when a verdict is FAIL or
scan allcould not run one of its scans, 3 for incomplete evidence without failures, and 0 for PASS/N/A. CI must handle incomplete scans explicitly. - Saved JSON and Markdown preserve findings and remediation. The console reports when its preview is limited and links to the full report. Check the scanner's failure policy and coverage limits: a threshold pass does not mean there are no lower-severity findings or that every requirement was inspected.
- Hosts without content for a profile report n/a rather than failing: the Amazon Linux 2023 bastion and the Debian 12 GCP bastion have no STIG content, and managed GKE / AKS / EKS nodes are not SSH-reachable (use
cisthere). - kube-bench's policy checks run with a temporary read-only ClusterRole that cannot read Secrets, removed when the scan ends.
- Scanners (kubescape, trivy) are installed on first use.
- For continuous scanning, install the operators:
cs platform install trivy-operator kubescape-operator.
FIPS 140 mode¶
FIPS mode is chosen when the environment is created (the SSH key type depends on it) and cannot be toggled later. Everything cloudseed puts into the environment must then be FIPS-capable, or setup and install refuse it.
| Target | What changes |
|---|---|
| AWS | the provider and the S3 state backend use FIPS endpoints (so only us-east-1/2, us-west-1/2 and GovCloud); EKS nodes run Bottlerocket FIPS AMIs; the Amazon Linux 2023 bastion runs fips-mode-setup --enable and reboots; the Ubuntu VPN host attaches Ubuntu Pro and enables fips-updates |
| GCP | the bastion and VPN host use the Ubuntu Pro FIPS image (metered, no token needed); GKE nodes stay on Container-Optimized OS |
| Azure | the bastion and VPN host use the Ubuntu Pro FIPS marketplace image; the AKS node pool is created with fips_enabled |
| VMware | every VM attaches Ubuntu Pro and enables fips-updates, then reboots; the guest OS must be Ubuntu and Kubernetes must be RKE2 |
| All hosts | sshd and OpenVPN offer only FIPS-approved algorithms; environments get an RSA-4096 SSH key instead of ed25519 |
Refused in FIPS mode
kubeadm, Tailscale (WireGuard / ChaCha20) and ed25519 keys are refused. VMware, AWS with a VPN host and a GCP bastion on a plain Ubuntu image need UBUNTU_PRO_TOKEN (free for personal use): export it or store it with cs creds set UBUNTU_PRO_TOKEN before setup applies. A setup that would apply stops without it; --dry-run and --plan-only only warn.
Platform items in a FIPS environment¶
| Tier | Items | Behaviour |
|---|---|---|
| compatible | controllers and services that terminate no user TLS with their own crypto | installed |
| tls-restricted | Envoy Gateway, ingress-nginx | installed with TLS 1.2+ and FIPS cipher suites; flagged by cs scan fips because the proxy crypto is not a validated module |
| crypto-restricted | cert-manager, sealed-secrets, velero, cloudnative-pg | installed (the platform needs them) and flagged |
| the rest | application stacks that ship their own crypto | refused unless --force, and listed by cs scan fips |
The Platform catalog shows each item's tier. cs scan fips verifies the whole chain: configuration, SSH key type, kernel fips_enabled and sshd / OpenSSL algorithms on every host, FIPS endpoints and node images, the RKE2 build, the TLS policy on the shared Gateway and the FIPS capability of every installed item.
Related¶
- Scenarios: compliance scans, FIPS 140 mode
- Credentials: how cloudseed keeps secrets local
cs help security,cs help fips,cs help scan,cs explain fips,cs explain scan