Security Defaults
The VerteX Management Appliance ships with the security behavior on this page built in. Unless a setting is listed in Operator-Configurable Settings, no System Console page, Tenant Console page, Local UI page, API field, or install-time setting changes the behavior described here.
Most sections end with a command that you can run to confirm the behavior on your own installation.
Transport Encryption
The VerteX Management Appliance serves the following interfaces over Transport Layer Security (TLS) only.
| Interface | Port |
|---|---|
| Tenant Console and management API | 443 |
| System Console | 443 |
| Local UI | 5080 |
| Kubernetes API server | 6443 |
| etcd | 2379 |
| Kubelet | 10250 |
| Built-in pack registry | 30003 |
Port 80 answers every request with a redirect to the HTTPS address and returns no other content. A sign-in request sent over plain HTTP is redirected, not processed.
On every interface in the table:
- TLS 1.2 is the minimum protocol version, and TLS 1.3 is the maximum. TLS 1.1, TLS 1.0, SSL 3.0, and SSL 2.0 are refused.
- Only AES-GCM cipher suites are offered:
- TLS 1.3:
TLS_AES_128_GCM_SHA256andTLS_AES_256_GCM_SHA384. - TLS 1.2:
TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256,TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384,TLS_ECDHE_ECDSA_WITH_AES_128_GCM_SHA256, andTLS_ECDHE_ECDSA_WITH_AES_256_GCM_SHA384. Each interface negotiates the RSA or ECDSA pair that matches its certificate key.
- TLS 1.3:
- Key agreement uses ECDHE over NIST curves. Each interface refuses the X25519 curve and the ChaCha20, CBC, and 3DES suites.
- Handshakes are signed with SHA-256 or SHA-384.
When a client requires the Commercial National Security Algorithm Suite (CNSA) 1.0 parameters (AES-256-GCM, ECDH P-384, and SHA-384), the interfaces respond as follows:
-
The Tenant Console, System Console, management API, Kubernetes API server, etcd, and built-in pack registry negotiate all three parameters.
-
The Local UI negotiates AES-256-GCM and SHA-384 and offers only P-256 for key agreement.
-
The Kubelet negotiates AES-256-GCM and P-384 and signs the handshake with SHA-256 because its certificate carries a P-256 key. The cluster certificate authority signs that certificate with SHA-256 and RSA, like every other certificate in the cluster.
The platform database (MongoDB) accepts only TLS connections and runs TLS in FIPS mode.
Verify
From a workstation that reaches the appliance, confirm that TLS 1.1 is refused and that an AES-GCM suite is negotiated. Repeat with port 5080 for the Local UI.
echo | openssl s_client -connect <appliance-address>:443 -tls1_1 -cipher 'DEFAULT:@SECLEVEL=0' 2>&1 | grep --ignore-case 'alert protocol version'
echo | openssl s_client -connect <appliance-address>:443 2>/dev/null | grep --extended-regexp 'Protocol +:|Cipher +:'
Replace <appliance-address> with the IP address or domain name of your VerteX Management Appliance.
SSL routines:ssl3_read_bytes:tlsv1 alert protocol version
Protocol : TLSv1.3
Cipher : TLS_AES_128_GCM_SHA256
Cryptographic Modules
The VerteX Management Appliance performs its cryptography in FIPS 140-3 validated modules selected at build time:
- The management plane services, the Kubernetes control-plane components, etcd, and the Kubelet are built against the Spectro Cloud Cryptographic Library (Cryptographic Module Validation Program (CMVP) certificate #5061). Refer to FIPS-Compliant Components.
- The Traefik ingress proxy and the built-in pack registry are built against the Go Cryptographic Module v1.0.0 (CMVP certificate #5247), with FIPS 140-3 mode on by default. Traefik terminates TLS for the Tenant Console, the System Console, and the management API.
- Appliance nodes run the Ubuntu Pro FIPS kernel with FIPS mode enabled.
The tenant settings Allow non-FIPS packs, Allow non-FIPS features, and Allow non-FIPS cluster import govern the packs, features, and imported clusters a tenant uses. They do not change the cryptographic modules of the management plane. Refer to Enable non-FIPS Settings.
Verify
On a control-plane node, confirm kernel FIPS mode and the ingress proxy's module.
cat /proc/sys/crypto/fips_enabled
sudo grep --text --only-matching --extended-regexp 'build[[:space:]](GOFIPS140|DefaultGODEBUG)=[^[:space:]]*' /proc/$(pgrep --exact traefik)/exe | sort --unique
1
build DefaultGODEBUG=fips140=on
build GOFIPS140=v1.0.0-c2097c7c
Certificates and Signatures
- Every certificate in the appliance cluster's public key infrastructure is signed with SHA-256
(
sha256WithRSAEncryption). This includes the Kubernetes certificate authority, API server, etcd, front proxy, and Kubelet certificates, and the default certificates that the consoles present after installation. - Each TLS interface refuses a handshake when the client offers only SHA-1 signature algorithms.
- A certificate uploaded on the System Console Administration > System Address tab must verify against the certificate authority supplied with it and match the system address, or the upload is refused. A certificate signed with SHA-1 fails that verification. Selecting Insecure Skip Verify does not bypass this check.
Verify
From a workstation that reaches the appliance, confirm that a handshake limited to SHA-1 signatures is refused.
echo | openssl s_client -connect <appliance-address>:443 -tls1_2 -sigalgs 'RSA+SHA1:ECDSA+SHA1' -cipher 'DEFAULT:@SECLEVEL=0' 2>&1 | grep --ignore-case alert
Replace <appliance-address> with the IP address or domain name of your VerteX Management Appliance.
The output contains alert handshake failure and SSL alert number 40.
SSH on Appliance Nodes
The SSH daemon on each appliance node offers only the following ciphers, Message Authentication Code (MAC) algorithms,
and key exchange algorithms. The host image sets them in /etc/ssh/sshd_config.d/00-complianceascode-hardening.conf,
which /etc/ssh/sshd_config includes.
| Type | Algorithms |
|---|---|
| Ciphers | aes256-ctr, aes256-gcm@openssh.com, aes192-ctr, aes128-ctr, aes128-gcm@openssh.com |
| MAC algorithms | hmac-sha2-512, hmac-sha2-512-etm@openssh.com, hmac-sha2-256, hmac-sha2-256-etm@openssh.com |
| Key exchange | ecdh-sha2-nistp256, ecdh-sha2-nistp384, ecdh-sha2-nistp521, diffie-hellman-group-exchange-sha256 |
Compliance scanners that read only /etc/ssh/sshd_config can report these settings as missing. Use sshd -T to read
the effective values.
Verify
On an appliance node, print the effective SSH daemon configuration.
sudo sshd -T | grep --extended-regexp '^(ciphers|macs|kexalgorithms) '
ciphers aes256-ctr,aes256-gcm@openssh.com,aes192-ctr,aes128-ctr,aes128-gcm@openssh.com
macs hmac-sha2-512,hmac-sha2-512-etm@openssh.com,hmac-sha2-256,hmac-sha2-256-etm@openssh.com
kexalgorithms ecdh-sha2-nistp256,ecdh-sha2-nistp384,ecdh-sha2-nistp521,diffie-hellman-group-exchange-sha256
Authentication
Sign-in Feedback
The Tenant Console, System Console, and Local UI mask the password as it is typed. Every sign-in surface answers a failed sign-in with one message, whether the account does not exist or the password is wrong. The following table lists the message on each interface.
| Interface | Failure message |
|---|---|
| Tenant Console and System Console | Invalid credentials specified or too many failed attempts |
| Local UI | Login failed due to incorrect credentials |
| Palette Terminal User Interface (TUI) on the node console | invalid username or password |
API Keys
API keys are generated inside the management API from the random bit generator of the Spectro Cloud Cryptographic Library. Each key is a 128-bit random value. The key is shown once, when it is created. Palette VerteX stores only a salted one-way hash of the key, and no API response returns the key after creation. No field selects a generator, key length, or format. Refer to Create API Key.
Session Tokens
- Tenant Console, System Console, and password-authenticated API sessions use a JSON Web Token (JWT) protected with HS256 (HMAC-SHA-256). The signature key is generated at installation and stored as a secret on the management cluster.
- Local UI sessions use a JWT signed with RS256 under a key pair generated on each host.
- When Palette VerteX acts as an OpenID Connect identity provider for clusters, it signs the tokens it issues with RS256 under a per-tenant key pair.
- The only token setting an operator can change is the session lifetime, from 15 to 1,440 minutes. Values outside that range are refused. Refer to Session Timeout.
Verify
Sign in over the API and save the session token.
TOKEN=$(curl --silent --insecure --header 'Content-Type: application/json' --data '{"emailId":"<email-address>","password":"<password>"}' https://<appliance-address>/v1/auth/authenticate | jq --raw-output .Authorization)
Replace <email-address> and <password> with the credentials of a Palette VerteX user, and <appliance-address> with
the IP address or domain name of your VerteX Management Appliance.
Decode the token header.
echo "$TOKEN" | jq --raw-input --raw-output 'split(".")[0] | @base64d'
{"alg":"HS256","typ":"JWT"}
Audit Records
Audit Layers
The following audit layers ship enabled. None can be turned off from a console, the API, or an install option.
| Layer | Where to read it |
|---|---|
| Management API audit trail | Tenant Console Audit Logs page, or GET /v1/audits |
| Kubernetes API server audit log | /var/log/apiserver/audit.log on control-plane nodes |
Host audit daemon (auditd) | /var/log/audit/audit.log on each node |
| systemd journal | journalctl on each node |
| Local UI audit log | /var/log/stylus-audit.log on each node, or Local UI Diagnostics > Audit |
Timestamps
Every layer takes its timestamps from the system clock of the node that writes the record. When Network Time Protocol (NTP) servers are configured, they synchronize that clock. No layer uses another time source.
Every layer records UTC, and every node runs in UTC. The Tenant Console Audit Logs page displays times in the
viewer's local time zone, to the minute. By default, the Local UI displays times in the viewer's local time zone and
labels the offset, for example (Local: UTC-05:00). To display UTC or both, open the User Menu in the upper-right
corner of the Local UI, select Time/Date Settings, and then select Device (UTC) or Both (Local, Device). The
display setting does not change the stored records.
The following layers record fractions of a second.
| Layer | Format | Precision |
|---|---|---|
| Management API audit trail | RFC 3339 with Z | Milliseconds |
| Kubernetes API server audit log | RFC 3339 with Z | Microseconds |
| Host audit daemon | Seconds since the Unix epoch | Milliseconds |
| systemd journal | ISO 8601 with +0000 | Microseconds |
Verify
On a control-plane node, read the newest API server audit record and confirm the UTC, sub-second timestamp.
sudo tail --lines=1 /var/log/apiserver/audit.log | grep --only-matching --extended-regexp '"requestReceivedTimestamp":"[^"]+"'
"requestReceivedTimestamp":"2026-11-13T14:58:43.536483Z"
Maintenance and Diagnostic Sessions
- Tenant Console sign-ins and changes are recorded in the management API audit trail with the actor and the action. Failed sign-ins, refused requests, and reads are not recorded.
- System Console sign-ins and some System Console actions are not recorded, and System Console records do not identify the administrator who acted. Refer to Known Issues.
- Local UI sessions are recorded in the Local UI audit log with the actor, the action, and the outcome, including failed sign-ins.
- SSH sessions are recorded in the systemd journal with the user, the accepted key, and the source address. The host
audit daemon records when each session opens and closes. The SSH daemon also logs to
/var/log/auth.log, which only root and theadmgroup can read. - Commands run with
sudoare recorded by the host audit daemon and by the systemd journal, which logs the user, working directory, target user, and command. There is no separatesudolog file. - Kubernetes API requests are recorded in the API server audit log with the user, verb, resource, source address, and result code.
Verify
On an appliance node, list the audit rule for commands run with sudo.
sudo auditctl -l | grep /usr/bin/sudo
-a always,exit -S all -F path=/usr/bin/sudo -F perm=x -F auid>=1000 -F auid!=-1 -F key=privileged
Kubernetes API Audit Policy
The shipped Kubernetes API server audit policy records requests at the Metadata level or higher and ends in a
catch-all Metadata rule. A refused request is recorded with code 403. The policy does not record the following
routine requests:
-
Requests for
events. -
Requests to the
/healthz,/version, and/swaggerendpoints. -
Reads, watches, and updates that Kubernetes system components make:
-
kube-proxy watches of endpoints and services.
-
Node status reads by the Kubelet.
-
Endpoint reads and updates in
kube-systemby the controller manager, scheduler, and endpoint controller. -
ConfigMap reads in
kube-systembysystem:unsecured. -
ConfigMap and endpoint reads and updates in
kube-systemby the cluster autoscaler. -
Namespace reads by the API server.
-
Metrics reads by the controller manager.
-
Access to Runtime Objects
The container runtime socket, /run/containerd/containerd.sock, is accessible to root only. The host audit daemon
records every attempt to create, open, or truncate a file that fails with EACCES or EPERM, including an attempt to
open this socket. The rule applies to every account that logs in with a user ID of 1000 or higher, and it still applies
after the account runs su or sudo. System and service accounts that never log in are not covered.
Verify
On an appliance node, list the audit rules for refused file access. The four rules cover EACCES and EPERM for 32-bit
and 64-bit system calls.
sudo auditctl -l | grep key=access
-a always,exit -F arch=b32 -S open,creat,truncate,ftruncate,openat,open_by_handle_at -F exit=-EACCES -F auid>=1000 -F auid!=-1 -F key=access
-a always,exit -F arch=b64 -S open,truncate,ftruncate,creat,openat,open_by_handle_at -F exit=-EACCES -F auid>=1000 -F auid!=-1 -F key=access
-a always,exit -F arch=b32 -S open,creat,truncate,ftruncate,openat,open_by_handle_at -F exit=-EPERM -F auid>=1000 -F auid!=-1 -F key=access
-a always,exit -F arch=b64 -S open,truncate,ftruncate,creat,openat,open_by_handle_at -F exit=-EPERM -F auid>=1000 -F auid!=-1 -F key=access
Workload Isolation
-
The management plane's security services are authentication, user and role management, the management API, and system administration. Each runs as a separate Kubernetes Deployment under its own service account and as a non-root user. Each uses the
RuntimeDefaultsecure computing mode (seccomp) profile and the containerd AppArmor profile. Their namespace enforces thebaselinePod Security Standard. The console front end, ingress, image registry, storage, and networking run in their own namespaces. -
No container shares the host PID or mount namespace.
-
Only the following infrastructure components, shipped in the appliance image, use the host network, UTS, IPC, or cgroup namespaces, or host paths:
-
The Kubernetes control-plane components.
-
kube-vip.
-
kube-proxy.
-
The Calico network agent.
-
The LINSTOR and DRBD storage components.
-
The cluster agent, for its pack cache at
/usr/local/spectrocloud/packs.
-
-
Every unprivileged container runs under the containerd AppArmor profile. The only privileged containers are the Calico network agent, kube-proxy, and the LINSTOR storage components.
-
The Kubelet runs with
protectKernelDefaults: true. -
Workloads are grouped by function in dedicated Kubernetes namespaces. Nothing runs in the
defaultnamespace.
Verify
On a control-plane node, count the pods that share the host PID namespace. Kubernetes has no pod setting that shares the host mount namespace.
sudo kubectl --kubeconfig /etc/kubernetes/admin.conf get pods --all-namespaces --output jsonpath='{range .items[?(@.spec.hostPID==true)]}{.metadata.name}{"\n"}{end}' | wc --lines
0
API Input Validation
The management API declares a type for every request parameter and request body in its published
API specification. It binds each request body to its declared type and refuses a body of the wrong
type with a 400 response that names the field. The management API also checks typed query parameters on resource
endpoints and refuses a value of the wrong type with a 422 response that names the parameter. The Local UI API refuses
request bodies of the wrong type the same way.
Verify
Send a query parameter of the wrong type.
curl --silent --insecure --write-out ' %{http_code}\n' --header "ApiKey: <api-key>" "https://<appliance-address>/v1/audits?limit=abc"
Replace <api-key> with a Palette VerteX API key, and <appliance-address> with the IP address or domain name of your
VerteX Management Appliance.
{"code":601,"message":"limit in query must be of type int64: \"abc\""} 422
Operator-Configurable Settings
The following related settings can be changed. Nothing else described on this page is configurable.
| Setting | Where |
|---|---|
| Console certificate, key, and certificate authority | System Console Administration > System Address |
| Session lifetime, 15 to 1,440 minutes (default 240) | Tenant Console Tenant Settings > Platform > Platform Settings > Session Timeout |
| Local UI time display (UTC, local time, or both) | Local UI User Menu > Time/Date Settings |
| NTP servers | Local UI Cluster > Actions > Cluster Settings > Network Time Protocol (NTP) |
| Pack registry certificates, or an external registry in place of the built-in one | Local UI Cluster > Configuration > Container Registry |
| Audit log forwarding | Tenant Console Tenant Settings > Infrastructure > Audit Trails |
| Non-FIPS packs, features, and imported clusters | Tenant Console Tenant Settings > Platform > Platform Settings |