Build a runbook for unmanaged device containment
An unmanaged device is any endpoint that the organisation cannot reliably identify, monitor, update or control. It may be a contractor laptop, forgotten company phone, personal device used for work, rogue IoT sensor or shadow-IT laptop. The device is not necessarily compromised; the risk comes from uncertainty about who owns it, what data it holds and whether it can reach business systems.
A containment runbook turns that uncertainty into rehearsed decisions. It defines what triggers isolation, who may authorise it, which actions are safe, how evidence is protected and how normal access is restored. Without this discipline, responders may wipe a useful source of evidence, shut down equipment with safety implications or leave stolen credentials active after removing the device.
Those decisions must work across Australian operating models: a Sydney head office supporting a Brisbane branch, a Perth team covering regional sites, or field staff using devices across mine and agricultural locations. Privacy obligations under the Privacy Act 1988 and the Notifiable Data Breaches scheme, where applicable, can make containment a legal event as well as a technical one.
Map the unmanaged device estate
Begin with a clear definition. A device is unmanaged when its owner, purpose, management status or risk is unknown to the security team. Unknown status should be recorded separately from confirmed compromise. This distinction prevents every forgotten charger or newly connected sensor from triggering the same emergency response as a device communicating with an attacker’s infrastructure.
Build the estate picture by reconciling endpoint management, EDR, network access control, DHCP, IP address management, DNS, VPN, identity and proxy logs. Service desk tickets, procurement records, visitor registers and physical device checks can reveal endpoints missing from those systems. Network connections are especially useful because they show where an unmonitored device has reached, even when nobody can confirm its identity.
Assign each unmanaged endpoint a preliminary risk level based on exposure, data sensitivity, connectivity, device criticality and suspected behaviour. A personal phone accessing email through a managed application has a different profile from an unknown laptop holding customer records on a shared network. This risk rating should determine investigation effort and the urgency of containment.
Define triggers, authority and safety limits
Set objective triggers before an incident occurs. Examples include confirmed malware command-and-control traffic, signs of credential theft, unauthorised access to sensitive data, repeated policy bypass or an unknown device with unrestricted access to critical systems. A credible trigger should permit immediate action without waiting for complete attribution.
Authority must also be predefined. The runbook should identify the security lead, incident commander, system owner, privacy contact and legal representative, with deputies for out-of-hours cover. Reversible actions such as network quarantine may be pre-authorised, while destructive actions such as reimaging or factory reset should require approval from the relevant system owner.
Containment can affect availability, privacy and safety. A device supporting clinical equipment, a mining control system or agricultural automation may need an operational safe state before it is isolated. The Privacy Act, the Notifiable Data Breaches scheme, contractual obligations and relevant state or sector requirements should be assessed in parallel with the response, not after containment has begun.
Prepare evidence, tools and communications
Only prescribe tools that are already deployed and supported. Depending on the environment, useful controls may include endpoint detection and response, network access control, mobile device management, identity governance, VPN revocation, firewall rules, secure web gateways, DHCP controls and SIEM alerts. If an endpoint has no management agent, containment may require an out-of-band action such as disabling a switch port or retrieving the device.
Treat the first containment action as evidence-sensitive. Record the device identifier, serial number, network address, last known user, time and method of detection, observed connections and the person authorising action. Capture logs, images and relevant configuration where feasible, and maintain a chain-of-custody record. A factory reset performed too early can erase the path used to scope the incident.
Prepare a communication card, escalation paths and short message templates before they are needed. Australian teams often work across states, daylight-saving changes and regional time zones, so use 24-hour times and identify the relevant office explicitly. Plain, direct wording is preferable: state what is known, what action is being taken, who owns it and when the next update will arrive.
Execute containment without losing visibility
Follow a consistent sequence: identify the endpoint, verify the alert, establish its role, contain it, address associated identities and then monitor for secondary activity. This sequence stops responders from isolating the wrong device while giving them a repeatable account of what happened. Each action and result should be recorded in the incident log.
Match the response to the device class. A managed company phone can be locked or wiped remotely. An unmanaged contractor laptop may require switch-port shutdown, quarantine VLAN assignment or physical collection. An IoT or operational technology device should generally be segmented rather than powered down, with the operational owner controlling any transition to a safe state. For a personal device, apply proportionate controls and consider the user’s data and ongoing access needs.
Removing an endpoint does not remove compromised credentials. Revoke active sessions, suspend the associated account where justified, rotate exposed secrets or tokens, and block confirmed command-and-control indicators across identity, email, cloud and network systems. Monitor those systems for unusual behaviour while the endpoint remains isolated. Release the device only after ownership, integrity, required security controls and the incident’s recovery criteria have been confirmed.
Coordinate enterprise and external responders
Containment rarely sits within one security team. A managed service provider may operate the network, a telco may provide circuit details, a forensic specialist may collect images, and a business owner may understand the affected workflow. The runbook should establish a simple responsibility matrix and show when an external incident retainer, insurer, regulator or law-enforcement agency enters the process.
CARM Security, developed by Exclusive Networks, is suited to coordinating a response that spans multiple security technologies and specialist services. By giving responders a shared operational view, the platform can support decisions across unmanaged endpoint discovery, containment, investigation and post-breach remediation. The runbook should still identify the human owner of every decision; technology should accelerate the response rather than obscure responsibility.
Australian organisations often combine internal security teams with managed service providers, regional IT partners and specialist incident responders. Their contracts should define escalation times, evidence access, notification duties and authority to isolate systems. Many buyers also align their programmes with ASD guidance, including the Essential Eight, so containment records can support control assurance as well as immediate operational needs.
Test, measure and sustain the response
Exercise the runbook against realistic scenarios. A useful tabletop might involve a contractor laptop found on a corporate network, a personal phone used to access a shared inbox, or an unknown IoT sensor communicating with a production environment. Record every ambiguous instruction, missing contact and dependency on an unavailable tool. Update the runbook after each exercise and whenever the endpoint estate or supporting technology changes.
| Device class | Common Australian context | Initial containment | Evidence and follow-up |
|---|---|---|---|
| Personal or BYOD device | Staff accessing email or collaboration tools | Quarantine managed applications, revoke sessions and protect linked accounts | Retain relevant identity and network logs; avoid unnecessary removal of personal data |
| Contractor laptop | Temporary project or remote worker | Suspend access, isolate the network connection and confirm asset ownership | Image or collect logs where an incident is suspected |
| Unknown company device | Lost, replaced or never registered | Lock, segregate and verify its management history | Preserve device identifiers, user details and connection history |
| Rogue IoT or OT equipment | Building, agriculture, utilities or industrial site | Apply network segmentation while the operational owner establishes a safe state | Preserve telemetry and configuration; coordinate shutdown carefully |
| Former employee endpoint | Offboarding or credential misuse | Disable the account, revoke certificates and isolate the endpoint | Review access history and investigate possible data exposure |
Measure containment time, scope accuracy, evidence completeness, escalation speed and the number of manual actions. Record whether isolation affected critical operations and whether identity controls caught activity that endpoint tools missed. These measures show whether the runbook works in practice and identify where integrations, training or supplier agreements need attention. The next concrete step is a booked 90-minute tabletop exercise using the contractor-laptop scenario, with one owner assigned to close every identified gap before the session ends.