HART (Highway Addressable Remote Transducer) is the protocol most process-plant transmitters, positioners and valve controllers speak over a 4-20mA current loop, and it has no authentication at all — anyone electrically connected to the loop can read and write a device's configuration — while its wireless sibling WirelessHART, standardized as IEC 62591, was built with mandatory AES-128 encryption and key management from the start, which means the real security gap in a HART deployment usually sits in physical loop access, gateway configuration and asset-management software, not in the wireless protocol's cryptography itself.
Key takeaways
- Because it was designed for a technician standing at the loop.
- WirelessHART extends the same HART command set over a wireless mesh instead of a wired loop, aimed at process measurement points where running new cable is expensive or impractical.
- Three kinds of 128-bit symmetric key, and only one of them is handled by people.
- With standard TLS and DTLS, since mid-2020.
What is HART, and why does the wired version have no security layer?
Because it was designed for a technician standing at the loop. HART started in 1986 when Rosemount (now part of Emerson) found a way to add two-way digital communication on top of the existing 4-20mA analog current loop that process instrumentation already used, by superimposing a low-amplitude frequency-shift-keyed (FSK) digital signal that doesn't disturb the analog reading. FieldComm Group, the industry body that now owns and maintains the protocol, still registers HART-compliant devices today, and the protocol remains the dominant way to configure and diagnose process transmitters, positioners and valve controllers without pulling them off the loop. Because the entire design goal was convenient configuration and diagnostics for a technician standing at a terminal box or holding a handheld communicator, classic wired HART has no field for a credential, token or signature: any HART master — a legitimate handheld, asset-management software, or a device clipped onto the loop by someone who shouldn't be there — can issue read and write commands a device will accept. The only real control over classic wired HART is physical: who can reach the loop wiring and junction boxes.
How is WirelessHART (IEC 62591) different?
WirelessHART extends the same HART command set over a wireless mesh instead of a wired loop, aimed at process measurement points where running new cable is expensive or impractical. The HART Communication Foundation released it in September 2007, and it was standardized internationally as IEC 62591 — first edition 2010, current second edition IEC 62591:2016, maintained by IEC Technical Committee 65/Subcommittee 65C. Unlike classic wired HART, FieldComm Group's own published security materials describe WirelessHART security as mandatory and always-on rather than a configuration option: a two-layer defense pairs hop-by-hop authentication at the data-link layer with end-to-end encryption and authentication at the network layer, both built on AES-128. A dedicated Security Manager role tracks keys and network security status across the mesh, and the join key a new device uses to authenticate to the network — provisioned out of band with a handheld or commissioning tool before the device is ever powered up at the plant — is treated as the single most important credential in the system. FieldComm Group's 2017 white paper describes unique per-device join keys as a device access control list that blocks unknown nodes, and says a common join key, where used, "should be changed periodically."
| Property | Wired HART (4-20mA loop) | WirelessHART (IEC 62591) |
|---|---|---|
| Authentication | None — any loop-connected master is trusted | Mandatory; join key + per-session keys managed by a Security Manager |
| Encryption | None | AES-128, hop-by-hop (data-link) and end-to-end (network layer) |
| Can security be disabled? | N/A — none exists to disable | No, per FieldComm Group's own description — always-on by design |
| Primary residual risk | Physical access to loop wiring and junction boxes | Gateway/firmware implementation bugs, join-key provisioning and management practice |
| Governing body | FieldComm Group (HART protocol owner) | FieldComm Group; standardized as IEC 62591 |
How do WirelessHART keys work in practice?
Three kinds of 128-bit symmetric key, and only one of them is handled by people. The HART Communication Foundation's WirelessHART Security Overview (HCF_LIT-114, March 2010) and FieldComm Group's 2017 white paper describe them:
| Key | Purpose | Who handles it |
|---|---|---|
| Join key | Keeps the joining exchange private and proves to the Security Manager that the device belongs to this network | Chosen by the user in some systems, generated automatically in others; unique per device or common per network |
| Network key | Shared by all devices on the network "to facilitate broadcast activity" | Generated by the Security Manager |
| Session keys | End-to-end CCM* mode AES-128 encryption between data source and consumer | Generated by the Security Manager |
Every packet also carries a message integrity check that covers routing information, so a node in the mesh cannot rewrite where a packet goes. Field devices route traffic but "are not authorized to be network key servers or wireless network managers", and the field network does not run a TCP/IP stack. The Security Manager "is usually embedded within the WirelessHART Gateway and automatically handles all keys except for the join key."
That design holds as long as no node is compromised. A 2022 formal analysis with the Tamarin prover by Martin Gunnarsson of RISE (ACSAC ICSS workshop paper) confirmed previously published attacks and found a new one in which "an insider attacker" can "maliciously re-key devices", a denial of service on large parts of the network. The weak points are therefore a leaked join key and a compromised device, not the cipher.
How is HART-IP secured?
With standard TLS and DTLS, since mid-2020. HART-IP carries the same HART commands over IP. FieldComm Group says it was first deployed in 2009, mostly in I/O systems and WirelessHART gateways, and uses port 5094. Its Introduction to HART-IP Security (FCG_AG10197, Edition 2.0, 15 March 2023) states that security "has always been mandatory", but "the initial security choices were left to the vendor" until the minimum requirements were standardized in mid-2020. The minimum for a HART-IP server (version 2):
- TLS/DTLS 1.2 or later, with four required cipher suites and AES-128 for traffic once the session is up.
- Pre-Shared Key (PSK) and Secure Remote Password (SRP) authentication; servers must support both, and the site chooses.
- At least 5 user credential sets. Slot 0 belongs to the Security Manager, the only user that can change credentials. Users can be read-only; a read-only client that attempts a change is disconnected.
- Audit Log of at least 128 client sessions (who, when, whether the configuration change counter moved). It is held in volatile memory, so a power cycle clears it; the last power-up time exposes that.
- syslog events for start-up, credential changes, configuration changes, session open and close, and failed connection attempts.
The commissioning rule is the trap. If a new or factory-reset server's first session comes from a version 1 client, the server "is locked into Version 1 and no security credential can be entered", and later v2 clients are refused until a factory reset. The paper's advice is to provision on the bench, on an isolated network, where a fresh device answers at 192.168.0.42.
What must a HART-IP device maker build in?
Section 7 of the same paper turns the protocol requirements into product requirements:
Field-upgradable firmwareAll HART-IP servers must support field firmware upgrades, because cipher suites will change; the device needs memory for both the update and the running image. A root of trust and secure bootFor example by booting from read-only code with JTAG disabled on production units, a coprocessor that validates the main code, or protected boot storage. "If the firmware module does not authenticate then it is not executed." Secure credential storagePSKs and passwords are confidential data in secure memory; the Unique ID and MAC address are protected but not secret. A physical factory resetthat returns slot 0 credentials to defaults, resets the tag to the MAC address printed on the device and returns ports to 5094. A security development lifecycle, from threat modelling to fuzz testing and an incident response plan. The paper notes this lifecycle is assessed when a device goes for security certification; see IEC 62443-4-1 and IEC 62443-4-2.
What do real advisories against HART products actually show?
The documented failures sit exactly where the comparison above suggests: in implementation code and configuration, not in WirelessHART's AES-128 design.
| Advisory | Product | CVE | CVSS | What it actually allowed |
|---|---|---|---|---|
| ICSA-15-008-01A (2015) | CodeWrights HART DTM library, used across Emerson Rosemount/Fisher HART field devices | CVE-2014-9191 | 2.1 (v2, NVD); 1.8 (v2, CISA) | Crafted response over the 4-20mA loop crashed the DTM/FDT Frame software (no control disruption) |
| ICSA-20-287-04 (2020) | FieldComm Group HART-IP Developer Kit and hipserver reference software | CVE-2020-16209 | 9.8 (v3) | Oversized HART-IP message caused a stack-based buffer overflow; could crash the device or allow code execution |
| ICSA-20-135-02 (2020) | Emerson Wireless 1410/1420/1552WU WirelessHART Gateways | CVE-2020-12030 | 10.0 (v3) | Enabling the VLAN feature silently disabled the gateway's internal firewall, exposing all ports to connected wireless field devices |
| NVD, 31 Aug 2021 | Pepperl+Fuchs WirelessHART-Gateway 3.0.7 to 3.0.9 | CVE-2021-34565 | 9.8 (v3.1) | SSH and Telnet services active with hard-coded credentials |
| NVD, 31 Aug 2021 | Pepperl+Fuchs WirelessHART-Gateway up to 3.0.7 | CVE-2021-33555 | 7.5 (v3.1) | Unauthenticated path traversal reading arbitrary files |
The 2015 Emerson finding is a parsing bug reachable over the physical wired loop itself — a reminder that "no authentication" cuts both ways, since a compromised or rogue field device can target the engineering software reading it, not just the other way around. The 2020 HART-IP finding is in FieldComm Group's own reference implementation of the Ethernet-based HART-IP extension, the layer that lets HART traffic ride a plant's IP network rather than a dedicated loop. The Emerson gateway finding is the most serious for WirelessHART specifically: it didn't touch the AES-128 security model at all, it broke the network-level firewall that was supposed to be a second layer of defense around a WirelessHART gateway, through an unrelated feature interaction. The Pepperl+Fuchs findings sit in the same place: the gateway's own management services and web server. An earlier gateway directory traversal, CVE-2018-16059 in the fcgi-bin/wgsetcgi filename parameter, carries two vendor names: NVD's description names Endress+Hauser's WirelessHART Fieldgate SWG70 3.x, while CISA's ICSA-19-073-03 and CERT@VDE's VDE-2019-002 assign it to Pepperl+Fuchs WHA-GW WirelessHART gateways, fixed in firmware 03.00.08 (Modbus) and 02.00.01 (EtherNet/IP). NVD also lists a cross-site scripting flaw in Siemens' IE/WSN-PA Link gateway (CVE-2019-13923). None required breaking WirelessHART's cryptography to succeed.
Can you see HART traffic in a network capture?
Partly. Wired HART lives on the 4-20mA loop and WirelessHART on an IEEE 802.15.4 radio mesh, so neither appears on an Ethernet mirror port. What a capture does show is the IP side: HART-IP sessions on TCP or UDP 5094 between the gateway or remote I/O and its clients, plus the gateway's web, SSH and Telnet services. CISA's ICSNPP HART-IP parser for Zeek decodes HART-IP and calls itself "a minimal release" in which not all commands are implemented. Worth alerting on: a HART-IP client that is not the asset-management or control system, HART-IP crossing a zone boundary, and any connection to a gateway's Telnet or SSH port from outside the engineering network.
Worked example: what does a HART review of one process unit find?
Synthetic example, not a real plant. A unit has 40 wired HART transmitters on remote I/O, one WirelessHART gateway with 25 wireless devices, and an asset-management PC. The review finds:
| Observation | Why it matters | Action |
|---|---|---|
| The gateway uses one common join key, unchanged since commissioning | One leaked key admits any device | Move to per-device join keys, or rotate on a schedule with every device online |
| The asset-management PC reaches the gateway over HART-IP version 1 | No TLS/DTLS on that session | Upgrade the client, factory-reset, re-provision as v2 on the bench |
| Gateway Telnet answers from the office VLAN | The CVE-2021-34565 pattern | Disable Telnet, firewall the management ports |
| DTM library version predates the CVE-2014-9191 fix | Known, low-severity crash | Update the DTM |
What should an asset owner check first?
Inventory every HART-capable device, including transmitters, positioners and valve controllers, plus the handhelds, asset-management software (DTM/FDT-based tools) and WirelessHART gateways that talk to them — none of the advisories above, or any future one, can be assessed without model and firmware/version detail. Treat wired HART loop wiring as the actual security boundarySince classic HART has no authentication, physical access to loop terminals and junction boxes is the control that exists; restrict and monitor who can reach them. Verify WirelessHART security is configured as intended, not just presentConfirm whether join keys are unique per device or shared and rotated, and that a Security Manager is actively monitoring network security status, per FieldComm Group's own material. Patch gateways and asset-management/DTM softwareagainst the CVEs above and any newer CISA ICS advisory matching the exact vendor and model in use — the documented failures are in this layer, not the wireless protocol's encryption. Keep WirelessHART gateways off the internet and segmented from business networks, and re-verify firewall and VLAN configuration after any update — the Emerson gateway case shows a single feature interaction can silently disable a gateway's protections. The Purdue Model and NIST SP 800-82 give the zoning.
A passive capture is a practical way to start on steps 1 and 5 without touching a running loop or wireless network. SoftWSP's OT Snapshot reads Zeek and ICSNPP logs from a mirror-port recording and builds a host-by-port inventory, including hosts talking on port 5094, and lists remote-access sessions (SSH, Telnet, RDP, VNC) that cross from outside the private address range. It has no HART-IP decoder, so HART-IP shows at connection level, and it sees nothing of the loop or the radio mesh. For gateway and instrument makers, Device & Firmware Security matches a component list or firmware SBOM against NVD, CISA KEV, FIRST EPSS and CISA ICS advisories. An engineer reviews every AI-drafted, cited report before it is sent.
What are the common pitfalls?
- Commissioning a HART-IP device with an old client, which locks it into unsecured version 1.
- One common join key across the plant, never changed.
- Leaving default credentials on gateway web, SSH or Telnet services.
- Assuming WirelessHART encryption protects the gateway's Ethernet side.
- Treating a HART-IP Audit Log as a durable record; it is volatile, so send syslog to a server.
This guide draws on FieldComm Group's HART-IP security paper (FCG_AG10197) and 2017 WirelessHART white paper, the HART Communication Foundation's WirelessHART Security Overview, the IEC Webstore's catalogue record for IEC 62591:2016, CISA ICS advisories, NVD records, the ICSNPP HART-IP repository and NIST SP 800-82 Revision 3. It is general information, not a substitute for FieldComm Group's own specifications, a vendor's product-specific security guidance, or a current CISA ICS advisory search for your exact hardware and firmware version.
Put this guide to work
Frequently asked questions
Does the wired HART protocol have any authentication or encryption?
No. Classic wired HART — the Highway Addressable Remote Transducer protocol that Rosemount introduced in 1986 and that FieldComm Group now maintains — layers a low-rate digital FSK signal on top of the same two wires carrying a 4-20mA analog current loop, and it was designed purely for device configuration and diagnostics in an era when the only way to reach a transmitter was to physically connect a handheld or modem to that loop. The protocol has no concept of a credential, session key or message signature: any HART master (a handheld communicator, asset management software, or a rogue device clipped onto the loop) that can electrically reach a field device can read and write its configuration. Physical access to the loop wiring is effectively the only control that exists for classic wired HART.
Is WirelessHART (IEC 62591) actually more secure than wired HART?
Yes, by design, and this is the one part of the HART family where security was built in from the start rather than bolted on. WirelessHART was released by the HART Communication Foundation in 2007 and standardized as IEC 62591 (first edition 2010, current second edition IEC 62591:2016), and FieldComm Group's own published security materials describe its security as always-on and not something a user can disable: data is authenticated hop-by-hop at the data-link layer and encrypted and authenticated end-to-end at the network layer, both using AES-128. A dedicated Security Manager role in the network tracks keys and network security status, and a device's join key — provisioned out of band with a handheld or commissioning tool before the device is powered up at site — is treated as the most security-critical credential in the system. The mechanism-level description above is drawn from FieldComm Group's own public security overview page and summaries of its WirelessHART security white paper, and should be checked against the primary PDF for exact wording if you need it for a compliance document.
What have real CISA advisories found in HART-speaking products?
Three, each a different kind of problem, and none of them break WirelessHART's cryptography. CISA's ICSA-15-008-01A (Update A, 2015) assigned CVE-2014-9191 to an improper-input-validation flaw in the CodeWrights HART Device Type Manager library Emerson used across dozens of Rosemount and Fisher HART field devices — a specially crafted response sent over the 4-20mA loop itself could crash the DTM software. CISA's ICSA-20-287-04 (last revised 13 October 2020) assigned CVE-2020-16209 (CVSS v3 9.8) to a stack-based buffer overflow in FieldComm Group's own HART-IP Developer Kit and hipserver reference software, discovered by Reid Wightman of Dragos — oversized HART-IP messages could crash the device or allow code execution. CISA's ICSA-20-135-02 (14 May 2020) assigned CVE-2020-12030 (CVSS v3 10.0) to an improper-access-control bug Emerson found internally in its Wireless 1410, 1420 and 1552WU WirelessHART gateways: enabling the VLAN feature on an affected gateway silently disabled its internal firewall, exposing every gateway port and allowing commands to reach connected wireless field devices. All three are implementation bugs in specific vendor software or reference code, not failures of the WirelessHART AES-128 security model itself.
What should a plant running HART or WirelessHART instruments check first?
Inventory every HART-capable transmitter, positioner and valve controller along with the handhelds, asset-management software and WirelessHART gateways that talk to them, since none of these advisories can be assessed without knowing model and firmware version. Physically secure 4-20mA loop wiring and junction boxes, because classic wired HART has no electronic substitute for that control. Confirm WirelessHART security is actually configured as intended — unique per-device join keys rather than one join key shared across a network, and a Security Manager that is actually monitoring network status — rather than assuming the protocol's mandatory encryption means nothing else needs checking. Patch gateways and asset-management software against the specific CVEs above and any newer CISA ICS advisory for the exact model in use, and keep WirelessHART gateways off the internet and segmented from business networks, the same standing recommendation CISA attaches to the Emerson gateway advisory and to OT protocols generally.
Sources
- FieldComm Group: WirelessHART Security overview
- IEC 62591:2016, Industrial networks — Wireless communication network and communication profiles — WirelessHART (IEC Webstore catalogue record)
- CISA ICS Advisory ICSA-15-008-01A: Emerson HART DTM Vulnerability, Update A (2015; CVE-2014-9191)
- NVD: CVE-2014-9191 (CodeWrights HART DTM library; NVD Primary CVSS v2 2.1)
- CISA ICS Advisory ICSA-19-073-03: PEPPERL+FUCHS WirelessHART-Gateways (14 March 2019; CVE-2018-16059)
- CERT@VDE VDE-2019-002: Pepperl+Fuchs: Path traversal in WirelessHART Gateway (6 March 2019; CVE-2018-16059)
- NVD: CVE-2018-16059 (description names Endress+Hauser WirelessHART Fieldgate SWG70 3.x)
- CISA ICS Advisory ICSA-20-287-04: FieldComm Group HART-IP and hipserver (last revised 13 October 2020; CVE-2020-16209)
- CISA ICS Advisory ICSA-20-135-02: Emerson WirelessHART Gateway (14 May 2020; CVE-2020-12030)
- NIST Special Publication 800-82 Revision 3, Guide to Operational Technology (OT) Security (September 2023)
- FieldComm Group: Introduction to HART-IP Security (FCG_AG10197, Edition 2.0, 15 March 2023, PDF)
- FieldComm Group: Intrinsically Secure WirelessHART Field Device Networks (June 2017, PDF)
- HART Communication Foundation: WirelessHART Security Overview (HCF_LIT-114, March 2010, PDF hosted by Emerson)
- Gunnarsson: Formal Verification of the WirelessHART Protocol (ACSAC 2022 ICSS workshop, PDF)
- CISA ICSNPP HART-IP: Zeek parser for HART-IP (GitHub)
- NVD: CVE-2021-34565 (Pepperl+Fuchs WirelessHART-Gateway hard-coded SSH/Telnet credentials)
About this guide
SoftWSP OT/ICS Desk
Covers the OT/ICS security cluster: 87 guides.

In this cluster
Questions about this page?
