Aventura, Florida · serving clients across the United States (305) 404-9404Toll-free (800) 767-8566Emergency? Get a call within 1 hour
Healthcare Illustrative engagement

A hospital network that keeps clinical systems, medical devices and guests apart, and logs everything

A regional hospital group in the southeastern United States, a leader in its segment.

Scale
6 campuses, 40 outpatient clinics, 12,000+ connected devices, 20+ segments
Engagement
Assessment, redesign and phased rollout over 9 months

The challenge

Infusion pumps, imaging systems, badge readers, building controls and guest Wi-Fi were all on networks that could reach the electronic health record. Many medical devices cannot run security software or be patched on demand, so the network itself had to contain them.

Hospitals are unusual networks: life-critical devices, decades-old equipment and some of the most sensitive data there is, all sharing cabling. The group's networks had grown campus by campus, and a flat design meant an infusion pump and a records server could, with the right mistake, reach each other. Many devices cannot run security software or be patched on demand, and replacing them was not an option, so the design had to put the control in the network.

Each class of device now has a home. Clinical workstations, medical equipment, building systems, administration, guests and card payments each live in a separate VLAN and routing domain, and traffic between them passes through a firewall that permits only named flows: imaging to the archive over DICOM, interfaces to the records system over HL7, nothing else. Access switches authenticate with IEEE 802.1X where a device supports it. For the many that do not, device profiling and MAC authentication bypass place them in a restricted segment, and an alert fires if a pump shows up on a workstation port. RADIUS change-of-authorization (RFC 5176) moves a device to a quarantine VLAN the moment its profile changes.

Cutover happened ward by ward in agreed windows, with clinical engineering present. Each ward was first mirrored: the new segments were built beside the old network, devices were profiled in monitor-only mode for two weeks, and the allow-lists were tuned against real traffic before anything was enforced. On the night, ports were moved in groups, a clinical engineer confirmed each device class still reported to its server, and the old flat network was retired segment by segment. Devices that failed profiling stayed on a legacy VLAN with tight firewall rules until their owners could be reached.

Testing was part of the design, not an afterthought. Before each wave we ran flows that must work (DICOM, HL7, badge and nurse-call traffic) and flows that must fail (a guest laptop reaching a records server, a pump reaching the internet). Core and distribution switches were failed by pulling power and links at the busiest hour of a test day, and traffic converged in well under a second with no dropped sessions on monitored devices. Logging was tested the same way: an investigator was asked to trace a planted test device from the moment it joined the network to the last packet it sent.

Logging was designed alongside segmentation rather than after it. Switch, firewall, wireless and authentication events flow to a central collector with synchronized time, so a team can follow a device from first connection onward. This supports the audit-controls requirement of the HIPAA Security Rule, 45 CFR 164.312(b); compliance itself is a program for the organization to run, not something a network can promise. The result is medical devices contained in their own segments, a network that identifies what is plugged in before it grants access, and one searchable audit trail that answers who and what touched which system, and when.

What our team delivered

  • Clinical, medical-device (IoT), building-systems, administrative, guest and payment VLANs, each in its own routing domain behind firewall policy
  • IEEE 802.1X authentication for staff devices and MAC-authentication bypass with device profiling for equipment that cannot authenticate
  • Dynamic VLAN assignment so a device lands in the right segment wherever it is plugged in
  • Allow-lists between segments for the protocols actually needed, such as DICOM and HL7 to named servers only
  • Central, tamper-evident log collection of network, authentication and access events with defined retention, supporting the HIPAA Security Rule audit-controls requirement (45 CFR 164.312(b))
  • Redundant core and distribution switching with tested failover

Result

Medical devices contained in their own segments, a network that identifies what is plugged in before it grants access, and one searchable audit trail that answers who and what touched which system and when. In the modeled engagement, the 12,000+ devices were profiled and placed into segments over nine months, clinic-by-clinic cutovers ran in overnight windows with a rollback plan, and unmanaged devices that could not authenticate were contained in restricted segments with a written exception and an owner. Failover of the redundant core was tested during rollout, and the audit trail retained events long enough to answer an investigation after the fact.

Separate by region and by business unit

Each business unit lives in its own VRF and VLAN in every region, so a problem in retail cannot reach finance. Traffic between units has to pass a firewall that denies by default and allows only what is written down.

Separate by region and by business unitA grid with three regions, Americas, EMEA and APAC, and three business units, Finance, Retail and R&D. Each cell is a separate VRF with its own VLAN and VNI. Each business unit has one overlay running across its three regional cells, shown as a row. On the right a firewall zone enforces a zero-trust policy that denies by default. A red packet from Retail trying to reach Finance is blocked at the firewall. A teal packet from R&D to the shared DNS and identity services is allowed. AMERICASEMEAAPACFinanceVRF fin · VLAN 110VNI 10110VRF fin · VLAN 110VNI 10110VRF fin · VLAN 110VNI 10110RetailVRF retail · VLAN 120VNI 10120VRF retail · VLAN 120VNI 10120VRF retail · VLAN 120VNI 10120R&DVRF rnd · VLAN 130VNI 10130VRF rnd · VLAN 130VNI 10130VRF rnd · VLAN 130VNI 10130Firewallzoneszero trustdeny bydefaultSharedDNS · IdPdenyallowVRF · L3 isolationVLAN to VNI · L2Default-deny policyEVPN-VXLAN fabricWireGuard + IPsec WANLogs to one SIEM

Every sign-in, verified

Staff sign in once, but behind the scenes the device, the identity, a phone approval and your access policy are all checked before any app opens, and strangers are turned away.

Every sign-in, verifiedDiagram of user authentication. A user signs in; the device is checked for MDM and Intune compliance; the identity provider, Active Directory or Entra ID, verifies the account; an MFA push is approved on the user's phone; conditional access evaluates the policy; and access is granted to email, ERP and files. On a second pass, a sign-in from an unknown device in a new location is blocked by the conditional access policy. VERIFIED SIGN-INUserSigns inDevice checkMDM · IntuneIdentityAD · Entra IDMFA pushPhone approvalConditionalaccessPolicy checkEmailERPFilesUnknown deviceNew locationApprove?Blocked

Roaming without dropping the call

Time from leaving one access point to the first data frame on the next, split by phase. A full WPA2/WPA3‑Enterprise re-authentication repeats the 802.1X exchange; 802.11r (Fast BSS Transition) reuses cached keys, 802.11k (neighbour reports) removes the blind scan, and 802.11v (BSS transition management) lets the network steer the client. Illustrative figures.

Roaming without dropping the callStacked horizontal bars of handoff time in milliseconds. Full re-authentication 580: scan 180, authentication and association 10, 802.1X 350, four-way handshake 40. PMK caching 230. 802.11r only 200. 802.11r with k and v 40, under the 50 millisecond voice target.Handoff time by roaming method (ms)Channel scanAuth + association802.1X / EAP4-way handshake0100200300400500600ms50 ms: a voice call does not noticeFull re-authenticationno roaming aids580 msPMK caching / OKCkeys reused, still scans230 ms802.11r onlyfast transition, still scans200 ms802.11r + k + vtargeted scan, fast transition40 msIllustrative: scan 20 to 250 ms depending on channel count; EAP depends on the RADIUS round trip.

Twelve months of availability

Measured availability per site over twelve months against a 99.99% target, which allows 52.6 minutes of downtime a year. Six of seven sites meet it; Singapore misses and is shown in red. Illustrative figures.

Twelve months of availabilityDot plot of twelve-month availability. Miami 99.998, Frankfurt 99.997, New York 99.995, London 99.993, Tokyo 99.992, São Paulo 99.991, all at or above the 99.99 percent target. Singapore 99.987 is below target.Availability per site, last 12 monthsMeets 99.99% targetBelow target99.99% target99.98%99.985%99.990%99.995%100.00%Miami99.998% · 10.5 min downNew York99.995% · 26.3 min downFrankfurt99.997% · 15.8 min downLondon99.993% · 36.8 min downTokyo99.992% · 42.1 min downSão Paulo99.991% · 47.3 min downSingapore99.987% · 68.4 min down · below target

Frequently asked questions

Direct answers to the questions we hear most. Something else on your mind?

Send us a message
How are medical devices that cannot be patched protected?

Many medical devices cannot run security software or be patched on demand, so the network contains them. They sit in their own IoT segment, with allow-lists permitting only needed protocols, such as DICOM and HL7 to named servers. Devices that cannot authenticate use MAC-authentication bypass with device profiling.

How does a device land in the right segment?

Staff devices authenticate with IEEE 802.1X, and dynamic VLAN assignment places each device in the right segment wherever it is plugged in. The network identifies what is connected before it grants access. Clinical, building-systems, administrative, guest and payment networks each have their own routing domain behind firewall policy.

How does the design support HIPAA audit controls?

Network, authentication and access events go to central, tamper-evident log collection with defined retention. This supports the HIPAA Security Rule audit-controls requirement (45 CFR 164.312(b)). The result is one searchable trail showing who and what touched which system and when.

Do you know if your company is safe?

Speak with a security engineer today. The first assessment call is free and confidential.

CallFree assessment