Start with named roles and authorised tasks.
List administrators, suppliers and automated processes. Assign permissions to the tasks each performs and separate ordinary application use from privileged administration. Use appropriate strong authentication, including multi-factor authentication where the access path supports it.
Avoid making one shared credential the only operational identity. Define an approval process for new access, periodic review and prompt removal when responsibilities end. Record who can grant privileges and how important changes can be traced to an authorised actor.
Manage keys as operational credentials.
For SSH access, identify the owner and purpose of each authorised key. Protect private keys, avoid transferring them through ordinary enquiries and remove access that is no longer required. Keep automation credentials distinct from a person’s interactive access where practical.
Document an emergency access route with controlled availability and an accountable owner. Test that authorised recovery is possible without depending on a departed employee, an unavailable laptop or a failed identity service. Review and close emergency access after use.
Coordinate access with the maintenance process.
Patching needs an inventory, a decision owner and a way to test application behaviour after change. Identify maintenance windows, restart requirements and the response to an urgent issue. Record exceptions with a reason and review date rather than allowing them to disappear from view.
Keep configuration changes reviewable and retain operational logs appropriate to the service. When a provider and customer both administer a server, agree how changes are communicated. This helps the next person investigate a problem without guessing what changed.
Define the management boundary separately from public traffic.
Restrict administrative interfaces to the approved access paths and networks. Separate management access from the public application where the architecture permits it, and review exposed services. Confirm that monitoring covers failed access attempts and important privilege changes.
DDoS planning addresses a different part of availability. Identify the public services, expected traffic and escalation route; discuss upstream protection and application-layer controls within a scoped security review. Include loss of connectivity in recovery planning so the team knows how administration will work during an incident.
