Network Segmentation for Medical IT
Network segmentation is the practice of dividing a network into separate parts so that a problem in one cannot spread freely into the others. On a hospital network it is one of the most effective controls available, because it limits how far a single compromised system can reach.
What a Flat Network Gets Wrong
When everything shares one network, every device can in principle talk to every other. That convenience is also the weakness: a compromise anywhere gives a foothold that can be used to reach everywhere else.
Medical equipment often sits on such a network precisely because it was easy to connect. Over time the estate accumulates devices with widely different security postures, all able to see one another, which is the opposite of what good design would suggest.
Physical separation and logical separation both have a place. Putting equipment on its own network hardware is a strong control where it is practical, while logical rules can achieve a similar effect where cabling cannot be changed, provided the rules are carefully written.
What Segmentation Achieves
Segmentation limits the blast radius. If a device in one group is compromised, the attacker's reach is bounded by the rules between groups rather than by the size of the whole network.
It also clarifies responsibility. When devices are grouped deliberately, it becomes possible to say who owns each group, what it may talk to and what should be treated as abnormal traffic within it.
It is worth documenting each group in plain language, describing what it contains and what it may talk to. A rule set that only a specialist can interpret tends to be treated as a hindrance by everyone else and eroded one exception at a time.
How a Sensible Grouping Looks
- Devices grouped by clinical function and similar technical risk.
- Explicit allowances for the traffic clinical work genuinely requires.
- A default that denies everything not on the allowed list.
- Named owners for each group, so changes have an obvious route.
- Periodic checks that actual traffic still matches the design.
Grouping Devices Sensibly
Groups should follow clinical and technical logic rather than convenience of location. Devices with similar risk, similar software and similar support needs belong together, and clinical systems that must exchange data belong in a group that permits that exchange explicitly.
It is worth resisting the temptation to create a group per department. Too many groups become unmanageable, and rules that nobody can explain are rules that will eventually be loosened without thinking.
Remote access deserves specific attention. Support arrangements that let a vendor reach a device are convenient and are also a route into the network, so the access should be granted for the period it is needed and withdrawn afterwards rather than left open.
Permitting What Is Needed
Segmentation is only useful if the traffic that is genuinely required can still flow. A clinical system that must reach a device has to be allowed to do so, and those allowances should be defined deliberately and recorded.
The default should be to deny, with explicit exceptions for what clinical work requires. That arrangement is more work to set up and far easier to defend than a permissive network patched with blocks wherever a problem appears.
Wireless adds a further dimension. Medical devices on a wireless network share the same risks as any connected equipment, and the ease of joining makes a deliberate enrolment process and periodic review of what is associated rather more important, not less.
Design Choices and Their Effect
| Choice | Benefit | Cost to manage |
|---|---|---|
| Many small groups | Tighter containment | More rules to maintain |
| Few broad groups | Simpler to run | A wider reach if breached |
| Default deny | Predictable posture | More setup and exceptions |
| Default allow | Quick to introduce | Weakest protection |
Testing and Maintaining the Design
A segmentation scheme drifts as devices are added and systems are replaced. Occasionally checking what actually talks to what, and comparing it with the intended design, catches the slow accumulation of exceptions.
Testing should include the failures as well as the normal path. Confirming that an isolated device really cannot reach a system it should not reach is the only way to know the control is doing what it was intended to do.
Changes to the network should follow the same route as changes to clinical equipment. A new rule or a new connection is a modification with consequences, and treating it as a routine technical adjustment is how unintended reachability appears without anyone intending it.
Working With the Teams Involved
Segmentation touches clinical engineering, IT, networking and the clinical departments that use the devices. All of them have to agree the groups and the allowances, or the design will be bypassed as soon as it obstructs routine work.
Explaining the purpose makes the restrictions easier to accept. Staff who understand that a rule exists to contain a compromise are far less likely to press for it to be relaxed for the sake of a small convenience.
Reviewing the arrangement with the people it affects is worth the time. Clinical teams notice when something can no longer reach a system it used to reach, and they are the quickest route to discovering that an intended allowance was set too tightly.
Segmentation is not a product to install but a design to maintain. Group devices deliberately, permit only what clinical work needs, check that reality still matches the plan and keep the teams affected by it involved.
Read next
Securing Connected Medical Devices
The baseline controls that protect networked equipment without getting in the way of clinical work.
Vulnerability Management in Hospitals
Turning vulnerability information into a plan a clinical engineering team can actually carry out.
Incident Response for Connected Devices
What to do in the first hours of a security incident involving equipment in clinical use.