Join our Newsletter — 33% off our NHI Course
Home› Guides› HL7 v2 and Clinical Interface Engine Security Guide:…
Guide OT & IoT Security

HL7 v2 and Clinical Interface Engine Security Guide: Feeds, Accounts and Encryption

PDF for newsletter subscribers · free to read online

← All guides
By NHI Mgmt Group, edited by Lalit Choda Updated 27 October 2026 13 min read
On this page

Most hospitals still move admissions, orders, results and billing data between systems as HL7 v2 messages over plain TCP, routed by an interface engine that holds credentials to almost every clinical database in the building. HL7 v2 remains, in HL7 International's own words, the workhorse of clinical data exchange, and the engine in the middle (Mirth Connect, Rhapsody, Cloverleaf, Corepoint, Iguana and others) becomes a single point where protected health information (PHI) is decoded, transformed and stored. This guide covers how to secure the message transport, the engine itself, its service accounts and stored messages, and how to monitor it, mapped to the HIPAA Security Rule technical safeguards. It is written for integration engineers, healthcare security architects and HIPAA security officers. FHIR APIs are covered separately in API security guides.

Key takeaways

  • HL7 v2 has no built-in security. The common MLLP transport is framing over TCP with no authentication or encryption, so confidentiality and sender trust must come from TLS, VPN tunnels and network controls.
  • The interface engine is a privileged hub. It holds database, SFTP, web service and EHR credentials for every connected system, so it deserves the same protection as a domain controller.
  • Default settings are not secure. The open source Mirth Connect configuration ships with an admin/admin first login, password policy values of zero and an embedded web server on port 8080; all need changing at install.
  • Unsupported engines are a live risk. CVE-2023-43208, unauthenticated remote code execution in Mirth Connect before 4.4.1, is in the CISA KEV catalogue, and NextGen states open source releases 4.5.2 and lower are end of life with no patches.
  • HIPAA expects access control for software, not only people. 45 CFR 164.312(a) requires access only for persons or software programs granted access rights, which puts every engine service account in scope.

How HL7 v2 traffic moves and where it is exposed

An HL7 v2 message is a pipe-delimited set of segments: the MSH header identifies sending and receiving application and facility, message type and control ID; PID carries patient identity; other segments carry visits, orders, observations and charges. Receivers reply with an acknowledgement (ACK) carrying an MSA segment. Messages are typically sent over the Minimal Lower Layer Protocol (MLLP), which wraps each message in start and end bytes on a long-lived TCP connection. Nothing in that stack authenticates the sender or encrypts the content.

Exposure pointWhat can go wrongPrimary control
MLLP listener on the engine or on a clinical systemAny host that can reach the port can inject messages (false results, merged patients, fake orders)Source IP allowlists, TLS with client certificates where both ends support it, network segmentation
Message in transit on the LAN or WANInterception of PHI in cleartextTLS-wrapped MLLP, IPsec or VPN tunnels for site-to-site and third-party links
Engine administration (web launcher, API, client)Takeover of all channels and stored credentialsHTTPS only, strong unique admin accounts, MFA via SSO where supported, admin network only
Engine message store and logsBulk PHI exposure; months of messages retained by default in some configurationsRetention limits, encryption at rest, database access restricted to the engine
Outbound connectors (database, SFTP, SOAP/REST, file shares)Credential theft gives direct access to downstream systemsLeast-privilege service accounts, vaulted secrets, per-connector credentials
Third-party feeds (labs, imaging centres, payers, HIEs)A partner compromise becomes an inbound attack pathDedicated tunnels, partner-specific channels, input validation and alerting

Integrity matters as much as confidentiality. An injected ORU result or an ADT merge against the wrong patient is a patient safety event, not just a data breach. HIPAA 164.312(c) requires policies to protect ePHI from improper alteration or destruction, and 164.312(e) requires protection against unauthorised access to ePHI in transit.

Securing the transport

  1. Inventory every interface. List each channel with source, destination, port, message types, partner and whether it crosses a network boundary. The engine's channel list is the starting point; firewall flows confirm it.
  2. Remove direct reachability. Place engines in a dedicated integration zone; allow MLLP ports only from named source systems. A listener reachable from the user VLAN is the most common gap.
  3. Encrypt boundary-crossing links first. Third-party and inter-site feeds should run inside IPsec or VPN tunnels or TLS. Then prioritise internal links that cross shared networks.
  4. Use TLS for MLLP where both ends support it. Many engines can wrap MLLP in TLS; many older clinical systems cannot. Where an endpoint cannot, terminate TLS on a proxy or tunnel next to it and keep the cleartext leg on a short, isolated segment.
  5. Authenticate peers. Mutual TLS with client certificates gives cryptographic sender identity. Without it, combine IP allowlists with MSH-3/MSH-4 sending application and facility checks in the channel filter, recognising that MSH fields are easily forged and only catch mistakes.
  6. Test and roll back. Change one interface at a time in a test environment with replayed de-identified messages, confirm ACK behaviour and queue drain, then cut over in a window with the old listener available for rapid fallback.

Validate what arrives: transport security proves where a message came from, not that it is sane. Channel filters should reject messages whose type, trigger event, sending application or facility is not expected on that interface, enforce field lengths and character sets before data reaches downstream databases, and route rejects to a monitored error queue rather than silently dropping them. Negative acknowledgements returned to the sender should not echo internal error detail. These checks also catch the far more common problem of a misconfigured upstream system sending the wrong feed, which is itself a patient safety risk.

HIPAA currently treats transmission encryption (164.312(e)(2)(ii)) as addressable. The HHS Security Rule notice of proposed rulemaking issued on 27 December 2024 would require encryption of ePHI in transit with limited exceptions, require MFA and network segmentation, and remove the addressable distinction. As of 27 October 2026 the HHS fact sheet still describes the rule as proposed and states the current Security Rule remains in effect. See the HHS HIPAA Security Rule Update (2025 NPRM) Guide.

Hardening the interface engine

Engine products differ, but the hardening areas are the same. The examples use the open source Mirth Connect defaults from its public repository because they are visible to anyone; check your own product's documentation for the equivalents.

AreaExample default (Mirth Connect open source mirth.properties and README)Hardened setting
First loginDefault username and password admin/admin, which the README says to change immediatelyDisable or rename the default account; named admin accounts; SSO with MFA where the product supports it
Password policypassword.minlength = 0, password.retrylimit = 0, password.lockoutperiod = 0Set length, lockout and reuse limits in line with your identity policy
Web serverhttp.port = 8080 serving the launcher page; administrator on https.port = 8443 with a self-signed certificateRestrict 8080 or disable HTTP; replace the certificate; allow admin ports only from an admin network
KeystoreKeystore path and store password set in the properties fileChange keystore passwords from shipped values and protect the file with operating system permissions
DatabaseEmbedded Derby by default; external database credentials stored in database.username and database.passwordExternal database with a dedicated least-privilege account; protect or externalise the credential
TLS for the APIhttps.client.protocols = TLSv1.3,TLSv1.2 and a cipher suite listKeep to TLS 1.2 and 1.3 and remove suites you do not need
Custom codeJavaScript transformers (Rhino engine) and a custom library directoryCode review and change control for channel scripts; restrict who can deploy channels and libraries
  • Patch or replace: CVE-2023-43208 (CVSS 3.1 9.8) is an unauthenticated RCE in Mirth Connect before 4.4.1, an incomplete fix for CVE-2023-37679, and CISA added it to KEV on 20 May 2024. NextGen's security policy states that from 4.6.0 Mirth Connect is commercial and that open source versions 4.5.2 and lower are end of life with no evaluation, backports or advisories. Organisations on open source builds need a supported path.
  • Run as a non-privileged service account: the engine process should not run as root or a local administrator, and should only write to its own directories.
  • Separate environments: test and production engines should not share credentials, and test should use de-identified data.
  • Restrict channel deployment rights: anyone who can deploy a channel can read every message it touches and send data anywhere the engine can reach.

Where identity fits: service accounts and connector credentials

An interface engine is one of the densest concentrations of non-human identities in a hospital. A typical engine holds database logins for the EHR, lab, radiology and billing systems, SFTP keys for payers and reference labs, API credentials for HIEs and cloud services, and its own credentials to its message store. HIPAA's access control standard covers "software programs" as well as persons, and 164.312(a)(2)(i) requires unique identification for tracking identity.

  • One credential per connector and destination: never reuse the same database account across channels; it makes revocation and audit impossible.
  • Least privilege: a results feed needs insert rights on specific tables, not database owner.
  • Vault the secrets: where the engine supports external secret references or configuration maps, keep credentials out of channel exports and source control. Channel export files and configuration maps often contain cleartext credentials.
  • Rotate on staff and vendor changes: integration contractors frequently know every connector password.
  • Inventory owners: each account should have a named application owner and a review date.

See the Healthcare Identity Security Guide, the Service Account Security Guide and the Secrets Management Guide.

Message stores, retention and data minimisation

  • Set retention per channel: engines often keep raw and transformed messages for troubleshooting. Keep the minimum period operations need (days, not months) and prune automatically.
  • Encrypt stored messages: use database or volume encryption, and the engine's own content encryption where available.
  • Restrict message browsing: viewing message content is access to PHI; limit it to support roles and log it.
  • Minimise payloads: filter segments and fields a destination does not need, for example removing identifiers from feeds to analytics systems.
  • Protect error queues and dead letters: failed messages are often written to file shares with broad access.

Monitoring and measurement

HIPAA 164.312(b) requires mechanisms that record and examine activity in systems that contain or use ePHI. For an engine, that means three log streams forwarded to the SIEM: engine audit events (logins, failed logins, user changes, channel deploys and edits, message views and reprocessing), operating system and database logs, and network flow logs for MLLP ports.

MetricTarget
Percentage of boundary-crossing interfaces encrypted (TLS, VPN or IPsec)100 percent
Percentage of MLLP listeners with source allowlists100 percent
Connector credentials that are unique, owned and vaultedRising to 100 percent
Engine version currency: days on a release with an open Critical advisoryZero beyond your patch SLA
Message retention compliance per channelAll channels within the approved period
Channel changes deployed without an approved changeZero

Decision guidance

  • TLS on the link versus a tunnel: choose TLS (ideally mutual) when both endpoints support it, because it authenticates the peer. Choose IPsec or a VPN when an endpoint cannot do TLS or when many interfaces cross the same boundary, accepting that it authenticates sites, not applications.
  • Central engine versus point-to-point: a central engine concentrates risk but also concentrates control and monitoring. Point-to-point links multiply unmanaged listeners. For most organisations the engine is the better place to enforce security, provided it is hardened as a tier zero system.
  • Small practices versus health systems: a small practice may have only a handful of vendor-managed interfaces; the priorities are vendor remote access with MFA, a supported engine version and encrypted partner links. A health system needs an interface inventory, change control for channels, a secrets vault and SIEM monitoring. See the Cyber Security for Dental and GP Practices Guide.

Framework mapping

Control areaHIPAA Security RuleNIST SP 800-53 Rev. 5NIST CSF 2.0
Service account access and least privilege164.312(a)(1), (a)(2)(i)AC-2, AC-6, IA-9PR.AA-05
Transport encryption and integrity164.312(e)(1), (e)(2)(i), (e)(2)(ii)SC-8, SC-13, SC-23PR.DS-02
Message integrity and validation164.312(c)(1), (c)(2)SI-10PR.DS-02
Audit logging164.312(b)AU-2, AU-12DE.CM-01
Admin authentication164.312(d)IA-2, IA-5PR.AA-03
Engine configuration and patching164.308 risk management (administrative safeguards)CM-6, SI-2PR.PS-02

Common failure patterns

  • MLLP listeners open to the whole network: any workstation can send a message that a clinical system will accept and file.
  • Cleartext partner feeds across the internet or a shared WAN: justified because "the vendor does not support TLS", with no tunnel in front.
  • Default and shared engine accounts: the first-install admin account still in use, shared by the integration team and contractors.
  • End-of-life engine left running: open source builds with known KEV-listed flaws reachable from user networks.
  • Credentials in channel exports: exported channels and configuration maps checked into shared folders or repositories with cleartext connector passwords.
  • Remote access to integration infrastructure without MFA: the Change Healthcare breach shows what happens when a healthcare data exchange intermediary exposes a remote access portal without MFA: attackers moved through the network for nine days before ransomware disrupted claims and pharmacy processing nationwide.

Practitioner checklist

  • Build and maintain an interface inventory with ports, partners, message types and encryption status.
  • Restrict every MLLP listener to named source systems and place engines in a dedicated integration zone.
  • Encrypt all boundary-crossing feeds with TLS or tunnels; use mutual TLS where both ends support it.
  • Replace default engine accounts, enforce password and lockout policy, and require MFA for admin access.
  • Run a supported engine release and track KEV-listed flaws such as CVE-2023-43208.
  • Give each connector its own least-privilege credential, vault it and assign an owner.
  • Set per-channel message retention, encrypt the message store and log message views.
  • Forward engine audit, OS, database and flow logs to the SIEM with alerts for new sources and off-window deployments.
  • Put channel changes and custom scripts under change control with peer review.

Standards and references

Related NHI Mgmt Group resources: Healthcare Identity Security Guide · HIPAA Security Rule Guide · Hospital Clinical Network Segmentation Guide · Threat Actors Targeting Healthcare Guide

PDF for subscribers

Take this guide with you

Every section, table and checklist in one PDF, with links to the sources and related guides. Subscribe to our weekly NHI & AI Identity Journal and we’ll email you a download link.

Free weekly newsletter

Subscribe to the NHI & AI Identity Journal

The latest on NHI and Agentic AI security – articles, research, breaches, news and events every week.

Bonus 33% off our NHI Course when you subscribe.

NHIMG Editorial Note
Researched and written by the NHI Mgmt Group editorial team using AI tools, under the editorial direction of Lalit Choda. Facts are drawn from the sources listed on this page and checked by our automated quality process. Last updated 27 October 2026. Spotted an error? Use our contact page.
NHI Mgmt Group, the independent authority on Non-Human Identity, IAM and Agentic AI security. nhimg.org

PDF Downloads for Newsletter Subscribers

Get the guide

Enter your details and we’ll email you a download link.

The full guide is free to read online. The PDF is a thank you for subscribing.

Fill in all fields marked * and tick the PDF box to continue.