SSH Hardening Baseline for Admin Teams

Review authorized keys, stale accounts, bastion access, and exposed ports.

Test the operational dependency after remediation, including service status, ports, storage, scheduled work, monitoring, and cluster membership. Security completion includes healthy production behavior.

Maintenance-window runbook

Group systems by role and redundancy, update one safe target first, validate the running state, and then continue through the fleet. Keep non-rebooted or pinned systems visible as exceptions.

• Capture package, kernel, process, and service versions before change.
• Apply the distribution-supported update with rollback and capacity prepared.
• Verify logs, listening services, monitoring, and workload health after restart.

Rank hosts by service impact

Patch public services, bastions, orchestration nodes, authentication infrastructure, and hosts with privileged workloads before low-impact internal systems. Use the distribution advisory for the fixed package, but confirm which process or kernel is active on each host. For clustered services, sequence nodes around capacity and failover tests. For immutable images and containers, rebuild and redeploy from a fixed base rather than changing a short-lived instance that will be replaced by the vulnerable image.

Avoid package-installed false confidence

Installing an update is not the same as loading it. Kernels await reboot, long-running processes retain old libraries, and containers may continue from stale layers. Verify the runtime and service state after the maintenance window. Also make sure rollback packages, snapshots, and golden images do not silently reintroduce the vulnerable version during recovery.

Running-state verification

Use uname, package-manager output, process maps, service state, and application checks to prove the fixed code is active. A downloaded package or pending reboot does not reduce the live exposure.

Record host groups completed, hosts deferred, the reason for each exception, and the next maintenance date. Include any temporary network restriction or live-patching control that remains in place.

Linux operations action

Document the approved access path so emergency changes do not quietly bypass the baseline.