Two operators at a control-room desk with a row of monitors, a laptop and keyboards

OT/ICS security

KNX protocol security: what KNX Secure fixes, and why classic KNX cannot be patched

Last updated:

KNX, the dominant open protocol behind commercial building automation, was built with no authentication or encryption at all, and ABB's 2026 vulnerability disclosure for a legacy KNX tool, republished by CISA, states flatly that for classic KNX products "there are no options to resolve the vulnerability with a software update on a technical level." KNX Secure — split into KNX Data Secure for the bus and the now internationally standardized KNX IP Secure for IP backbones — closes that gap with AES-128 CCM encryption, but only for the devices and segments actually configured to use it, and two separate, documented vulnerabilities in KNX's own engineering tooling and authorization model show that real-world KNX failures mostly happen around the cryptography, not inside it.

Key takeaways

  • On IP, KNXnet/IP uses port 3671, which IANA's registry still lists under its older name "e Field Control (EIBnet)", and routing multicast goes to 224.0.23.12, registered as "KNXnet/IP".
  • They add authentication, freshness and encryption to devices configured to use them.
  • For the legacy part, yes.
  • They showed that an unset BCU key is enough to brick a building remotely.

Why does classic KNX run with no security at all?

KNX, standardized as EN 50090 and the KNXnet/IP variant as ISO 22510, grew out of the European Installation Bus and was designed for a network model that assumed physical isolation: a trusted bus running lighting, HVAC, blinds, and access-control messages inside a building, with no expectation that an outside party could reach it. The base protocol reflects that assumption directly — there is no field in a classic KNX telegram for a credential, session key, or integrity check, so any device able to transmit on a TP (twisted-pair) bus segment, or reach a KNXnet/IP gateway over the network, can both read and inject commands.

On IP, KNXnet/IP uses port 3671, which IANA's registry still lists under its older name "e Field Control (EIBnet)", and routing multicast goes to 224.0.23.12, registered as "KNXnet/IP". Anyone on that network segment can see and send that traffic unless KNX IP Secure is in use.

ABB's 2026 advisory for its KNX Update Tool, republished by CISA as ICSA-26-209-07 (CVE-2026-12705, CWE-353 missing integrity check, CVSS v3.1 6.4 as scored by ABB; NVD had no score of its own as of 18 November 2026), captures why this can't simply be patched: "Due to the nature of the classic KNX protocol stack and security concept, there are no options to resolve the vulnerability with a software update on a technical level," and "Legacy KNX standards were never designed to meet state of the art security standards like introduced with KNX Data Secure published in 2017." ABB adds that exploitation needs physical access to the bus, and that legacy KNX devices should not control sensitive functions such as hotel-room access. That is a statement about the protocol's foundational design, not about one unpatched firmware version.

What do KNX Data Secure and KNX IP Secure actually change?

They add authentication, freshness and encryption to devices configured to use them. KNX Secure was, per the KNX Association, "already created in 2015", with KNX Data Secure published in 2017 per ABB's advisory. It comes as two complementary, additive mechanisms rather than a single switch:

KNX Data Secure KNX IP Secure
Protects Telegrams on the TP (twisted-pair) or RF bus media KNXnet/IP traffic on a building's IP backbone
Mechanism Authentication code + sequence number + AES-128 CCM encryption Full encryption of transferred telegrams and data over IP
Standardized as EN 50090-3-4 EN ISO 22510 (published November 2019)
Who needs it most, per KNX's own guidance Wireless (RF) segments — "the number 1 attack target, as communication between devices is completely out in the open" Any installation reachable from the internet, where KNX's own guidance says a VPN tunnel "is an absolute MUST" in addition to IP Secure

The cryptography itself is conventional and reviewed: AES-128 CCM, compliant with the ISO 18033-3 cipher standard, with the authentication code giving integrity (detecting a tampered or forged telegram) and the sequence number giving freshness (rejecting a captured-and-replayed telegram). KNX IP Secure's path to EN ISO 22510 matters beyond branding: the KNX Association describes it as the first vendor- and application-independent security standard for building automation.

How are the keys handled?

KNX's own technical article describes the chain:

  1. Factory keyEach device ships with "an individual FDSK (Factory Default Setup Key)", printed with the serial number as a QR code on the device.
  2. Tool keyBecause the FDSK "may be known by others," ETS replaces it with a tool key "generated individually for each device."
  3. Group keys"For maximum security, a separate key is used for each group address."
  4. Replay protectionEach telegram carries a 6-byte sequence number, and a receiver only accepts numbers higher than the last one from that sender.
  5. IP sideSecure routing uses "a common key" (the backbone key) for all multicast routing traffic, with a time-based counter. Secure tunnelling first sets up a channel with Diffie-Hellman, then sends the user ID and password.

In ETS, per a second KNX article, you need a project password before adding secure devices, plus a device authentication code and a password for each tunnelling connection. The FDSK printout and the ETS project file are therefore the secrets to protect.

Does mixing secure and legacy KNX devices undo the protection?

For the legacy part, yes. KNX Secure was deliberately designed to coexist with legacy hardware rather than force a full replacement. Couplers and line devices can bridge secure and non-secure segments, which is exactly what lets a building keep decades-old, non-upgradeable field devices running alongside new KNX Secure ones. KNX Secure protects only the segments and devices actually configured to use it. A legacy device on a bridged segment still transmits in the clear, and an attacker reaching that segment sees and can inject telegrams on it exactly as on a fully classic installation.

Upgrading in place rarely works either: KNX's own article says adding KNX Secure to existing devices "will, in most cases, not be possible without changing the hardware." CISA's 2026 republication scopes ABB's finding the same way: it "applies exclusively to legacy KNX products," which in most commercial buildings is a meaningful, often unaudited share of the installed base.

What happened in the KNXlock attacks?

They showed that an unset BCU key is enough to brick a building remotely. Limes Security reported that in October 2021 a German engineering company asked for help after "Someone had gotten access to their building automation control system over the internet and locked the owners out of it." The building contained several hundred KNX components, and "¾ of them were no longer operational due to the attack." The attackers "firstly unloaded the devices and then set the BCU key on those devices." Limes notes the BCU key is a 4-byte password and "most devices will not allow a reset of the password, even with physical access to the device," and estimated replacement at "over hundred thousand euros." It also cites scans by Alpha Strike Labs showing "more than 16,000 potentially vulnerable systems" in Germany, Austria and Switzerland.

In January 2024, Limes reported a new twist: attackers setting random keys with messages such as "system hacked bcu key xx1234xx," still without ransom demands. Limes released KNXunlocker, which can guess the key "within a few hours using a regular PC" thanks to a reduced key space. The protocol weakness behind this is CVE-2023-4346, published by CISA as ICSA-23-236-01 (CVSS 7.5) with a note that CISA "has received reports of this vulnerability being actively exploited." CISA added it to the Known Exploited Vulnerabilities catalog on 15 July 2026.

Which KNX vulnerabilities should device makers and integrators know?

Consistently, not breaks of the AES-128 cryptography itself, but failures in tooling, configuration and device implementation:

CVE Product Issue Score
CVE-2021-36799 ETS5 through 5.7.6 Hard-coded password "ETS5Password" protects project data; local users can read project information CVSS 3.1 8.8 (NVD)
CVE-2023-4346 Devices with Connection Authorization Option 1 and no BCU key Purge and set BCU key, locking the owner out; in KEV 7.5
CVE-2026-12705 ABB/BJE KNX Update Tool ≤ 2.0.175 No integrity protection of firmware images for classic KNX products 6.4 (ABB; no NVD score)
CVE-2024-4008 ABB/Busch-Jaeger KNX Secure displays and RoomTouch panels (10 product IDs) on 1.00, fixed in 1.02; BCU KNX bus coupling units (3 product IDs) on 1.3.0.33, fixed in 1.3.1_63 (ABB 9AKK108464A0803) "FDSK Leak … allows attacker to take control via access to local KNX Bus-System" 8.8 (NVD)
CVE-2024-4009 Same 13 products and fixed versions Replay of captured KNX telegrams 7.8 (NVD)
CVE-2021-37740 MDT KNXnet/IP Secure router SCN-IP100.03 and KNX IP interface SCN-IP000.03 before v3.0.4 Malformed SESSION_REQUEST makes the secure layer unresponsive until reboot 7.5 (NVD)
CVE-2023-33277 Gira KNX/IP-Router 3.1.3683.0, 3.3.8.0 Directory traversal in the web interface 7.5 (NVD)

ETS project-password obfuscation (CVE-2021-36799). ETS, the KNX Association's configuration tool, holds and exports the keyring material KNX Secure devices rely on. Through version 5.7.6, ETS5 protected the project password with a hard-coded password string ("ETS5Password") and a hard-coded salt value, so a local user with the project file could recover project information. The weakness sits in the engineering workstation, not on the wire.

BCU-key lockout (CVE-2023-4346). Devices implementing Connection Authorization Option 1 without a BCU (Bus Coupling Unit) key let an attacker with network access purge every device without additional security options, then set their own BCU key. Recovery without the key is not guaranteed. The fix is to configure a BCU key; the default unconfigured state is the exposure.

The FDSK leak and replay CVEs are implementation flaws in specific products rather than in the KNX Secure design, but they show that "supports KNX Secure" is a claim to test, not to assume.

What should a KNX device maker build in?

The table above reads as a requirements list:

  • Ship KNX Data Secure or IP Secure on new products, with a unique FDSK per device and no way to read it back over the bus.
  • Make BCU-key handling recoverable by the owner, for example through a documented, physical-access reset, so a lockout does not mean replacing hardware.
  • Sign firmware images and verify them on the device before flashing; ABB's advisory shows what happens when classic products cannot.
  • Fuzz the KNXnet/IP and secure-session parsers, including length fields, before release.
  • Treat the IP router's web interface as an attack surface: authentication, path handling and update channel.
  • Keep an SBOM per firmware release so you can answer customer questions when a library CVE lands.

Worked example: how would an integrator audit a mixed KNX building?

Synthetic example. The building, devices and findings are invented to show the method; they describe no real site.

An integrator takes over a four-floor office with about 300 KNX devices installed between 2009 and 2024:

  1. InventoryThe ETS project shows 40 KNX Secure devices on two new floors, the rest classic TP devices behind line couplers.
  2. Network pathsThe KNXnet/IP router sits on the office LAN with a port forward for UDP 3671, left by a former service contractor. The forward is closed and remote access moves to a VPN.
  3. BCU keysNo BCU keys are set on the classic devices. The integrator sets them in ETS and hands the key list to the owner with the project documentation.
  4. Secure sideThe router supports KNX IP Secure but runs in plain mode; it is switched to secure tunnelling with ETS-generated passwords.
  5. RecordsThe ETS project and FDSK sheets move to a dedicated commissioning laptop and an offline safe.
  6. Residual riskClassic devices that control door access are listed for replacement, following ABB's advice not to use legacy KNX for sensitive functions.

Can you monitor KNX traffic on the network?

Partly. Wireshark's KNXnet/IP dissector decodes KNXnet/IP, including the secure wrapper and session frames, and can decrypt KNX Secure traffic when given the keys. Bus telegrams on TP never reach an IP network unless a gateway forwards them. CISA's ICSNPP set of Zeek parsers has no KNX parser, and SoftWSP's OT Snapshot does not decode KNX either: KNXnet/IP shows up only as UDP 3671 flows between hosts in its asset inventory. For a known list of KNX routers and gateways, SoftWSP can match vendor and product against NVD, CISA KEV, FIRST EPSS and CISA CSAF ICS advisories, with a cited report reviewed by an engineer. See OT Snapshot.

What are the common pitfalls?

  • Port-forwarding 3671 "temporarily" for remote programming and never closing it.
  • Leaving BCU keys unset because ETS does not force them.
  • Keeping the ETS project and FDSK printouts on a shared drive; they are the keys to the building.
  • Assuming the multicast default is harmless. KNX's checklist asks whether 224.0.23.12 was changed for IP multicast.
  • Calling a building "KNX Secure" when only new floors are.

What should owners and integrators check first?

  1. Inventory which lines and devices actually negotiate KNX Secure today, versus which are legacy devices bridged in through a coupler.
  2. Treat the KNXnet/IP router and any internet-facing access path as the top-priority segment, per KNX's own guidance: VPN tunneling for remote access, VLAN separation and firewalling between the KNX IP network and everything else.
  3. Prioritize RF (wireless) KNX segments for KNX Data Secure, since KNX's own materials flag wireless as the most exposed attack surface by design.
  4. Update ETS and any KNX engineering workstations to current versions, and don't treat an ETS5 project password as a meaningful secret on unpatched installs given CVE-2021-36799.
  5. Set a BCU key on every device that supports oneand hand it to the owner, as KNX Association recommends in ICSA-23-236-01.
  6. Use the KNX Secure Checklist: the KNX Association checklist asks whether router ports are closed towards the internet, whether Point-to-Point and Broadcast communication is blocked across couplers, and whether the PID_Download_Counter shows downloads after your own configuration.
  7. Don't assume legacy KNX devices can be remediated by a firmware update— segmentation and physical bus access control are the compensating controls for that hardware.

For the wider OT picture, see NIST SP 800-82, the Purdue model and BACnet security.

This guide draws on the KNX Association's own published materials on KNX Secure, ABB's 2026 advisory as republished by CISA, CISA advisory ICSA-23-236-01, Limes Security's KNXlock reports, and NVD's published CVE records. Treat the checklist above as a starting point, not a substitute for a direct read of a given building's actual KNX configuration.

Put this guide to work

Frequently asked questions

Does classic KNX have any built-in authentication or encryption?

No. KNX (standardized as EN 50090 / ISO 22510, and the dominant open protocol for European and global commercial building automation — lighting, HVAC, blinds, access control) was designed as a trusted-bus protocol, and its original specification carries no cryptographic protection at all: any device on a classic KNX TP (twisted-pair) bus segment, or any attacker who can reach a KNXnet/IP gateway, can read and inject telegrams without authentication. This is not a bug that was later discovered and left unfixed — it is a structural property of the original design that vendors themselves now state plainly in vulnerability disclosures: a 2026 CISA ICS advisory covering an ABB/Busch-Jaeger KNX update tool (CVE-2026-12705, missing firmware-image integrity protection) quotes the vendor's own position that 'legacy KNX standards were never designed to meet state of the art security standards' and that, for classic KNX products, 'there are no options to resolve the vulnerability with a software update on a technical level.'

What do KNX Data Secure and KNX IP Secure actually add, and when is each used?

KNX Secure, which the KNX Association says it created in 2015, is two complementary mechanisms rather than one umbrella fix. KNX Data Secure protects telegrams directly on the TP/RF bus media, adding data integrity (an authentication code appended to every message, so a receiver can detect a modified or forged frame), freshness (a sequence number that defeats simple replay of a captured telegram), and confidentiality via AES-128 CCM encryption, compliant with the ISO 18033-3 cipher standard. KNX IP Secure does the equivalent job for IP-based installations, encrypting the KNXnet/IP traffic that runs over a building's backbone or management network. KNX IP Secure went further than a vendor spec: it was published as the international standard EN ISO 22510 in November 2019, making it, per the KNX Association's own description, the first vendor- and application-independent security standard for building automation. Both mechanisms are additive to the base protocol, not a replacement for it — a device has to explicitly support and be configured for KNX Secure to get any of this.

Can a KNX installation mix secure and non-secure devices, and what happens if it does?

Yes, and this is the central practical risk in any real KNX deployment, because KNX Secure was designed to be backward-compatible rather than to force a rip-and-replace. A coupler or line device can bridge secure and non-secure segments so that older, non-upgradeable field devices keep working alongside newer KNX Secure ones. The security consequence is straightforward: KNX Secure only protects the segments and devices actually configured to use it. A non-secure device on a bridged segment still communicates in the clear, and an attacker who can reach that segment (physically, or through a weak IP gateway) sees and can inject telegrams on it exactly as they could on a fully classic installation, regardless of how well the secure side of the same building is configured. The CISA/ABB advisory above makes the same point from the vendor side: the integrity gap it describes 'applies exclusively to legacy KNX products' still in the field, which in most real buildings means a meaningful fraction of the installed base, not a hypothetical edge case.

What do real, documented KNX tooling vulnerabilities look like, beyond the bus protocol itself?

The most concrete publicly tracked one sits in ETS (Engineering Tool Software), the KNX Association's own configuration tool, not in the bus protocol: CVE-2021-36799 (CVSS v3.1 8.8) found that ETS5 through version 5.7.6 stores project passwords by obfuscating them with a hard-coded password, 'ETS5Password,' and a hard-coded salt value, letting any local user who can read the project file recover the actual credential protecting that project's KNX Secure key material. A separate, protocol-level finding, CVE-2023-4346 (CVSS v3.1 7.5), affects devices implementing KNX Connection Authorization Option 1 with no BCU key configured: an attacker with network access to the installation can purge every device that has no additional security option enabled and then set their own BCU key, locking the legitimate operator out of devices they previously controlled with no guaranteed software-only recovery path. Neither of these is a break of KNX Secure's AES-128 cryptography itself — both are tooling and configuration-authorization gaps around it, which is consistently where documented KNX incidents actually occur.

Is KNX IP Secure's ISO standardization the same thing as every KNX IP installation being secure?

No — standardization describes what the specification requires of a conformant implementation, not what any given deployed building actually runs. EN ISO 22510 means KNX IP Secure is now a vendor-neutral, internationally specified mechanism any manufacturer can implement and interoperate against; it does not retroactively enable itself on a building's existing KNXnet/IP router, which has to be a model that supports it, correctly configured with ETS, and actually running in secure mode rather than in a legacy-compatible plain mode most routers still ship with by default. The KNX Association's own guidance on when to apply which layer is explicit about risk, not blanket mandate: it calls out wireless (RF) installations as a particular priority because 'communication between devices is completely out in the open,' and states that for any installation linked to the internet, 'the use of a VPN tunnel to access it via the internet is an absolute MUST' — implying that IP Secure alone, without that network-layer control, is not treated by KNX's own guidance as sufficient for an internet-facing deployment.

What should a building owner or integrator actually check first on a KNX installation?

Start from the assumption that most of a mixed-age KNX installation is still running in the clear, and work outward. First, inventory which lines and devices actually negotiate KNX Data Secure or KNX IP Secure today versus which are classic/legacy devices bridged in — a datasheet claiming 'KNX Secure capable' is not the same claim as a device actually configured and running in secure mode in that specific building. Second, treat the KNXnet/IP router and any internet-facing access path as the highest-priority segment to lock down, per the KNX Association's own VPN-and-firewall guidance, since that is the one path an attacker doesn't need physical building access to reach. Third, patch and update ETS and any KNX engineering workstations to current versions, and don't assume the project password is a meaningful secret on older ETS5 installs given CVE-2021-36799's findings. Fourth, check which devices use KNX Connection Authorization Option 1 without a configured BCU key, since CVE-2023-4346's lockout risk applies specifically to that unconfigured state and is fixable by simply setting the key. None of this requires replacing classic KNX field devices overnight — it requires knowing, specifically, which parts of the installation the ABB disclosure's statement that legacy products cannot be fixed with a software update actually applies to.

Sources

  1. KNX Association: What makes KNX the secure solution for your smart building? (KNX Secure, KNX Data Secure, KNX IP Secure overview)
  2. KNX Association: KNX IP Secure becomes new ISO Standard (EN ISO 22510, November 2019)
  3. CISA ICS Advisory ICSA-26-209-07: ABB KNX Update Tool (CVE-2026-12705, missing firmware integrity protection on legacy KNX products)
  4. NVD: CVE-2021-36799 (KNX ETS5 hard-coded password exposing project/key material)
  5. NVD: CVE-2023-4346 (KNX Connection Authorization Option 1 BCU-key lockout)
  6. ABB PSIRT CSAF advisory 9AKK108472A9270: File integrity can be bypassed in KNX Update Tool for classic KNX products (CVE-2026-12705)
  7. CISA CSAF: ICSA-26-209-07 ABB KNX Update Tool (republication of ABB advisory)
  8. CISA CSAF: ICSA-23-236-01 KNX Protocol (CVE-2023-4346, Connection Authorization Option 1)
  9. KNX Association (KNX Today): KNX Secure - protecting installations inside and outside (FDSK, tool key, sequence counter, secure routing and tunnelling)
  10. KNX Association (KNX Today): Security - configuring KNX Secure systems in ETS
  11. KNX Association: KNX Secure Checklist (V11-21), copy hosted by KNX Portugal
  12. Limes Security: KNXlock - an attack campaign against KNX-based building automation systems
  13. Limes Security: A new Twist in the KNXlock Attack Campaign (January 2024)
  14. IANA IPv4 Multicast Address Space Registry (224.0.23.12 KNXnet/IP)
  15. Wireshark: KNXnet/IP dissector source (packet-knxip.c)
  16. NIST SP 800-38C: Recommendation for Block Cipher Modes of Operation: the CCM Mode
Was this guide helpful?

In this cluster

KNX protocol security: what KNX Secure fixes, and why classic KNX cannot be patched