Zigbee has carried AES-128-CCM encryption since its earliest security model and the cipher itself has never been the documented weak point in any research this guide could find — the join moment still is, where a well-known default key has let researchers recover a network key since at least 2015, a gap the Dynamic Link Key negotiation introduced in Zigbee PRO 2023 (Revision 23) is designed to close. The recurring, real-world failure, confirmed across a decade of independent academic work, a maintained attack toolkit, and real CISA/NVD findings, is how the key that feeds that cipher actually gets onto a joining device in the first place: a well-known default key called "ZigBeeAlliance09," a network key that research going back to 2013 has shown can be sniffed unencrypted or only weakly protected during commissioning, and implementation bugs — replay-counter checks, buffer overflows triggered at join time — layered on top of an otherwise sound cipher. Zigbee PRO 2023 targets exactly this join-time weakness with dynamic key negotiation and Device Interview, and CSA says Zigbee 4.0, announced in November 2025, "delivers modern cryptography, secure device authentication, and complete lifecycle security."
Key takeaways
- Zigbee is a low-power wireless mesh protocol maintained by the Connectivity Standards Alliance (CSA), formerly the Zigbee Alliance, on top of the IEEE 802.15.4 radio and MAC layer.
- Zigbee protects traffic with two kinds of AES-128 key: a network key shared by every device, and link keys between two devices.
- Install codes give each device its own pre-shared secret.
- Touchlink pairs a lamp with a remote by proximity instead of through the trust center, and its key is shared by every product.
What is Zigbee, and how is its mesh built?
Zigbee is a low-power wireless mesh protocol maintained by the Connectivity Standards Alliance (CSA), formerly the Zigbee Alliance, on top of the IEEE 802.15.4 radio and MAC layer. CSA's Zigbee page cites "over a billion Zigbee devices already deployed worldwide" on a "self-organizing, self-healing mesh topology" that is "scalable to thousands of nodes": bulbs, locks, sensors, thermostats, smart plugs and the hubs that tie them together.
A Zigbee network has one coordinator that forms it (usually also the trust center), routers that relay traffic, and sleepy end devices that do not. IEEE 802.15.4 supplies AES-based frame protection but leaves key management to higher layers ("Higher-layer processes may specify keys"), so how a device gets trusted keys is a Zigbee design choice. The same radio also carries Thread, which made different choices.
What is Zigbee's security model, and what role does the trust center play?
Zigbee protects traffic with two kinds of AES-128 key: a network key shared by every device, and link keys between two devices. Silicon Labs' Zigbee security concepts put it plainly: "Zigbee uses a 128-bit symmetric key to encrypt all transmissions at the network layer using AES-128."
The trust center decides who joins and hands out the network key, encrypted with a trust center link key. The Zigbee specification (Revision 23, 05-3474-23) lists the static options: "Default Global Trust Center link key (ZigBeeAlliance09); Install code derived pre-configured link key; Distributed Security Global Link Key; Touchlink Preconfigured Link Key". Its default is public: "The default value for the centralized security global trust center link key SHALL have a value of 5A 69 67 42 65 65 41 6C 6C 69 61 6E 63 65 30 39". Anyone who records a join that uses it can decrypt the network key.
Two later rules narrow that window. Since Revision 21 (2015), "the joiner is required to replace its initial trust center link key with an updated key". And per Silicon Labs' standard security guide, a trust center holding only a well-known key "should silently ignore the rejoin request"; such a key "should only be used for initial joining and even then it should allowed only for a brief period of time." Distributed security, where any router admits devices, skips the link key update entirely.
How do install codes and Dynamic Link Key change the join?
Install codes give each device its own pre-shared secret. Under the Base Device Behavior specification v3.1 (November 2025), an install code is "a 128-bit number and a 16-bit CRC", and "the pre-configured link key is derived from the Install Code using the Matyas-Meyer-Oseas hash function". The specification's test vector: install code 83FE D340 7A93 9723 A5C6 39B2 6916 D505 C3B5 yields the key 66B6900981E1EE3CA4206B6B861C02BB. The code travels out of band (label, QR code, app) to the trust center, so a sniffer without it learns nothing from the join.
Revision 23 (Zigbee PRO 2023) adds "a mechanism to negotiate a dynamic link key before the device joins the network", using ECDHE or SPEKE over Curve25519. BDB 3.1 restricts it: Dynamic Link Key "SHALL be used only if an install code derived link key or unique passphrase exist as pre-shared secret", and the trust center "SHALL reject the join" for an unknown device. Revision 23 also brought Device Interview (the trust center can query a joiner before admitting it), APS frame counter synchronization against replay, Restricted Mode and Trust Center Swap-out. CSA's Zigbee 4.0 one-pager describes the 4.0 release in general terms only; the security mechanisms are those of the Revision 23 line.
Why is Touchlink commissioning a risk?
Touchlink pairs a lamp with a remote by proximity instead of through the trust center, and its key is shared by every product. BDB 3.1 says the network key in a Touchlink exchange is "encrypted with the touchlink preconfigured link key", key index "0x04 (master key)" outside certification. Tobias Zillner's Black Hat USA 2015 white paper noted that "the ZLL master key has supposedly been leaked in the Internet", that Touchlink worked "from several meters away" with hobby radios, and that "if an attacker is able to sniff a device join using the default TC link key, the active network key is compromised".
Ronen, Shamir, Weingarten and O'Flynn's "IoT Goes Nuclear" (IEEE S&P 2017) used "a major bug in the implementation of the Touchlink part of the ZigBee Light Link protocol" to pull Philips Hue lamps off their networks, plus a side channel to recover the firmware-update key, for a worm design. In 2023 Silicon Labs' EmberZNet 7.1.3–7.3.0 had CVE-2023-41094: Touchlink packets processed after timeout "may allow a device to be added outside of valid TouchLink range or pairing duration". If a product does not need Touchlink, compile it out.
What does independent security research show about attacking a real Zigbee network?
The research converges on key handling, not the cipher. Vidgren et al. (HICSS 2013) and Olawumi et al., "Three Practical Attacks Against ZigBee Security" (IEEE HIS 2014), showed the network key sent in the clear in Standard Security and protected only by a shared master key in High Security, a mode removed in Revision 21. River Loop Security's KillerBee, started by Joshua Wright in 2009, turns this into tooling: zbdsniff scans captures for "NWK frames and over-the-air key provisioning" and prints keys it finds.
Wang, Cicala, Hussain, Bertino and Li (ACM WiSec 2020) argued that a pre-shared secret installed in millions of devices "will be inevitably leaked", and proposed a "certificate-less Zigbee joining protocol" with ECDH key exchange; they note their own ECDH-only interim step "is still vulnerable to an adversary that can carry out a man-in-the-middle attack", which is why their design also protects install codes with public keys.
Which CVEs and advisories name Zigbee?
Published Zigbee CVEs cluster at commissioning and join-time parsing:
| CVE | Product and versions | Issue | Source |
|---|---|---|---|
| CVE-2024-6350, -6351, -6352, -10106, -7322 | 34 Schneider Electric Wiser products on Silicon Labs stacks | Denial of service "when a malicious device joins the network"; CISA ICSA-26-027-03 (27 January 2026) lists mitigations only, including "Use install codes and avoid the well-known key" | CISA CSAF record |
| CVE-2025-8414 | Silicon Labs Zigbee EZSP hosts (Simplicity SDK, Gecko SDK) | Green Power buffer overflow, CVSS 4.0 9.4; access to a network key required | NVD |
| CVE-2020-6007 | Philips Hue Bridge 2.X up to 1935144020 | Heap overflow on a long ZCL string "during the commissioning phase"; fix 1935144040 deployed 13 January 2020 | Check Point |
| CVE-2026-3555 | Philips Hue Bridge v2 before 1975170000 | Heap overflow in custom ZCL handling during pairing, CVSS 8.0 | ZDI-26-153 |
| CVE-2022-39065 | IKEA TRÅDFRI gateway before 1.19.26 | One malformed broadcast frame makes the gateway unresponsive | NVD |
| CVE-2021-27289 | Ksix Zigbee kit | Weak frame-counter check allows replay | NVD |
None of these breaks AES-128-CCM, and as of the 13 November 2026 edition the CISA KEV catalog lists no Zigbee entry; absence from KEV means no confirmed exploitation reported, not safety.
What should a device maker or integrator check first?
Ship install codes or a unique passphrase, printed as a QR code, and enable the trust center policy that requires them. Confirm the device performs the Trust Center Link Key update after joining, and that the hub ignores rejoins under the well-known key. Open the join window only while pairing, and close it automatically. Disable Touchlink unless the product needs it; if kept, use a fixed stack (EmberZNet 7.1.6, 7.2.4, 7.3.1 or later for CVE-2023-41094). Move to a Revision 23 stackfor Dynamic Link Key, Device Interview and frame counter synchronization. Track the Zigbee stack version in the firmware SBOM(Silicon Labs, TI, NXP, Espressif), because advisories name stack versions; see the firmware SBOM guide. Fuzz ZCL and commissioning frames, where the Hue and Schneider bugs lived.
What does a join-security review of a smart plug look like?
The following is a synthetic, labelled example, not a real product: a Zigbee 3.0 smart plug and its hub.
| Check | Finding | Fix |
|---|---|---|
| Join key | Plug joins with ZigBeeAlliance09; install code printed but hub never asks for it | Hub policy: require install code; scan QR at setup |
| Link key update | Plug skips the update and keeps the default key | Stack setting; retest rejoin, which the hub must ignore |
| Join window | Hub permits joining for 254 s after any app launch | Open only from "Add device", 60 s |
| Touchlink | Enabled by SDK default | Removed from the build |
| Stack version | SBOM shows a Zigbee stack inside the CVE-2024-6350 range | Upgrade SDK; rerun fuzzing |
Common pitfalls: treating the "Zigbee 3.0" label as proof that the default key is refused; printing an install code nobody can enter; leaving distributed security on a hub product; and assuming the coordinator firmware is covered by the radio module vendor's updates.
Where does this connect to firmware security work?
Several of these checks depend on what is inside the firmware, starting with the Zigbee stack and its version. SoftWSP extracts a software bill of materials from a firmware image as CycloneDX 1.6 JSON, matches the components against NVD, the CISA KEV catalog, FIRST EPSS scores and CISA CSAF ICS advisories, and AI drafts a report in which each finding cites its record; an engineer reviews every finding before delivery. It does not test the join procedure over the air or check key configuration. See device firmware security, and the Matter guide for the IP-based alternative.
Put this guide to work
Frequently asked questions
Does Zigbee encrypt its traffic, and is that encryption actually strong?
Yes, Zigbee has carried AES-128 encryption since its earliest security model, and Silicon Labs' Zigbee security documentation states that 'Zigbee uses a 128-bit symmetric key to encrypt all transmissions at the network layer using AES-128', with link keys protecting application-support-layer (APS) traffic between two devices. AES-128-CCM itself is not the weak point — CCM (Counter with CBC-MAC) is a standard, well-studied authenticated-encryption mode, and nobody in the academic literature this guide reviewed claims to have broken the cipher. The documented weakness sits one layer up, in how the keys that feed that cipher get onto a device in the first place: IEEE 802.15.4, the radio/MAC standard Zigbee runs on, leaves key management to higher layers ('Higher-layer processes may specify keys'), meaning the strength of the encryption is only as good as whatever key-distribution scheme Zigbee (and the device vendor) layers on top of it. That is exactly where the historical and still-relevant vulnerabilities below live.
What is the 'ZigBeeAlliance09' default key, and is it still a real risk?
It is a real, publicly documented default value: the Zigbee specification (Connectivity Standards Alliance document 05-3474-23) states that 'the default value for the centralized security global trust center link key SHALL have a value of 5A 69 67 42 65 65 41 6C 6C 69 61 6E 63 65 30 39' — which spells out to the ASCII string 'ZigBeeAlliance09.' In Zigbee's Standard Security Mode, a joining device uses this well-known key (or a vendor default) to decrypt the network key sent to it by the trust center during the join handshake; because the key string is public and identical across every device that relies on it, anyone who captures that join exchange with an IEEE 802.15.4 sniffer can decrypt the network key and read or inject traffic on the rest of the network afterward. The Alliance has narrowed reliance on it over time — since Revision 21 'the joiner is required to replace its initial trust center link key with an updated key', and Silicon Labs' documentation says a trust center holding only a well-known key 'should silently ignore the rejoin request', pushing vendors toward Install-Code-derived keys instead — but the underlying design pattern (a key transported over the air, protected only by another key) is the same pattern that later motivated the Dynamic Link Key negotiation the Alliance added in Revision 23 (Zigbee PRO 2023).
What has independent academic research found by actually attacking Zigbee networks?
A consistent line of research spanning more than a decade, converging on the join/commissioning moment as the recurring soft spot. Vidgren, Haataja, Patino-Andres, Ramirez-Sanchis and Toivanen's 2013 paper at the Hawaii International Conference on System Sciences (HICSS), followed by Olawumi, Haataja, Asikainen, Vidgren and Toivanen's 2014 follow-up at the IEEE International Conference on Hybrid Intelligent Systems ('Three Practical Attacks Against ZigBee Security'), documented the network-key-sniffing attack directly: in Zigbee's Standard Security level the network key is transmitted unencrypted over the air during commissioning, and even in High Security mode (removed in Revision 21) it is only protected by a shared master key common to every node, so an attacker capturing that exchange — or later cracking the weaker protection — recovers the key for the whole network. Joshua Wright's KillerBee framework, first released in 2009 and still maintained, operationalizes this finding directly: its `zbdsniff` tool is built specifically to monitor captured IEEE 802.15.4 traffic for 'NWK frames and over-the-air key provisioning' and prints any network key it finds. More recently, Wang, Cicala, Hussain, Bertino and Li's 'Analyzing the Attack Landscape of Zigbee-enabled IoT Systems' (ACM WiSec 2020) argued that a pre-shared joining secret installed in millions of devices 'will be inevitably leaked' and proposed a certificate-less joining protocol based on Elliptic Curve Diffie-Hellman as a fix. None of this breaks AES-128 itself — every result is about compromising or intercepting the key before encryption is even relevant.
Has CISA published an ICS advisory that actually names Zigbee, and are there real CVEs?
Yes to both, and this guide verified real examples of each rather than assuming them. CISA's ICSA-26-027-03 ('Schneider Electric Zigbee Products,' released 27 January 2026) is a direct republication of a Schneider Electric CPCERT advisory covering more than 30 Zigbee-based Wiser smart-home products (thermostats, switches, sockets, EV chargers) built on Silicon Labs Zigbee processors, and it names five CVEs — CVE-2024-6350, CVE-2024-6351 and CVE-2024-6352 (CWE-120, classic buffer overflow, CVSS 6.5/4.3/4.3), CVE-2024-10106 (CWE-120, CVSS 3.7) and CVE-2024-7322 (CWE-400, uncontrolled resource consumption, CVSS 5.8) — all triggered, per the advisory, when 'a malicious device joins the network,' and notably shipped with no firmware patch available at publication, leaving network-level mitigations (install codes, restricting unknown-device joining) as the only remedy. Separately in NVD, CVE-2025-8414 ('Zigbee Green Power Host Buffer Overflow,' Silicon Labs Gecko/Simplicity SDK, published 17 October 2025) is a CVSS 9.4 stack-corruption flaw in Zigbee EZSP host applications that explicitly requires 'access to a network key' to exploit — tying it directly back to the key-exposure problem above. CVE-2022-39065 (IKEA TRÅDFRI gateway, CVSS 6.5) showed a single unauthenticated, malformed Zigbee broadcast frame could make the gateway unresponsive, and CVE-2021-27289 (Ksix Zigbee smart-home kit, CVSS 9.1) found the vendor's anti-replay frame-counter check was implemented so weakly that captured commands could be replayed with a bumped sequence number to spoof control messages. This guide found no CISA advisory or CVE claiming AES-128-CCM itself was broken — every real finding here is an implementation, configuration or key-handling failure sitting on top of the cipher.
What does Zigbee 4.0 actually change about security, and is it a fix for the known weaknesses?
Most of the join-time fixes predate Zigbee 4.0. The Connectivity Standards Alliance announced Zigbee 4.0 on 18 November 2025 as backward compatible with Zigbee 3.0 devices, and its one-pager says only that Zigbee 4.0 'delivers modern cryptography, secure device authentication, and complete lifecycle security.' The concrete mechanisms aimed at the weaknesses above are in the Zigbee PRO 2023 (Revision 23) specification: Dynamic Link Key negotiation (ECDHE or SPEKE over Curve25519, which the Base Device Behavior specification allows only with an install code or unique passphrase as the pre-shared secret), Device Interview (the trust center can query a joining device before authorizing it), APS frame counter synchronization against replays like CVE-2021-27289, Restricted Mode, and Trust Center Swap-out so a failed hub can be replaced without re-commissioning the mesh. Revision 23.2 of the specification and Base Device Behavior v3.1 are both dated 10 November 2025. These features only help if the product enables them: a Zigbee 4.0 label does not show that install codes are required or that Touchlink is disabled.
Sources
- Connectivity Standards Alliance: Zigbee Standard overview (architecture, AES-128-bit encryption claim, mesh topology)
- Connectivity Standards Alliance newsroom: "Zigbee Privacy & Security Fundamentals" (2026-04-01)
- Connectivity Standards Alliance: Zigbee specification (CSA document 05-3474-23) — trust center link key and ZigBeeAlliance09 default value
- Connectivity Standards Alliance newsroom: "The Connectivity Standards Alliance Announces Zigbee 4.0" (2025-11-18)
- River Loop Security / GitHub: KillerBee — IEEE 802.15.4/ZigBee Security Research Toolkit (originated by Joshua Wright, 2009)
- Wikipedia: IEEE 802.15.4 (PHY/MAC scope, security suites left to higher layers)
- Olawumi, Haataja, Asikainen, Vidgren, Toivanen: "Three Practical Attacks Against ZigBee Security," IEEE 14th International Conference on Hybrid Intelligent Systems, 2014
- Wang, Cicala, Hussain, Bertino, Li: "Analyzing the Attack Landscape of Zigbee-enabled IoT Systems and Reinstating Users' Privacy," ACM WiSec 2020
- CISA ICS Advisory ICSA-26-027-03: Schneider Electric Zigbee Products (released 27 January 2026; CVE-2024-6350, -6351, -6352, -10106, -7322)
- NVD: CVE-2025-8414 — Zigbee Green Power Host Buffer Overflow Vulnerability (Silicon Labs, published 17 October 2025)
- NVD: CVE-2022-39065 — IKEA TRÅDFRI Gateway E1526 malformed Zigbee frame denial of service
- NVD: CVE-2021-27289 — Ksix Zigbee smart-home kit replay-attack / frame-counter bypass
- Connectivity Standards Alliance: Zigbee PRO Base Device Behavior Specification v3.1 (22-65816-030, 10 November 2025): install codes, MMO hash, DLK, Touchlink keys
- Connectivity Standards Alliance: Zigbee 4.0 one-pager
- Connectivity Standards Alliance: Zigbee Security and Privacy Fundamentals (Zigbee PRO 2023 introduced Dynamic Link Key)
- Silicon Labs: Zigbee security concepts (network key, frame counter synchronization since R23)
- Silicon Labs: Zigbee standard security (well-known link key and rejoin handling)
- Tobias Zillner: "ZigBee Exploited: The Good, the Bad and the Ugly," Black Hat USA 2015 white paper
- Ronen, Shamir, Weingarten, O'Flynn: "IoT Goes Nuclear: Creating a ZigBee Chain Reaction," IEEE S&P 2017 (IACR ePrint 2016/1047)
- CISA ICSA-26-027-03 CSAF record (Schneider Electric Zigbee Products)
- NVD: CVE-2023-41094 — Silicon Labs EmberZNet Touchlink processing after timeout
- Check Point Research: "Don't be silly – it's only a lightbulb" (CVE-2020-6007, Philips Hue Bridge)
- Zero Day Initiative ZDI-26-153: Philips Hue Bridge Zigbee stack heap overflow (CVE-2026-3555)
About this guide
SoftWSP Firmware Security Desk
Covers the Firmware vulnerabilities & fixes cluster: 131 guides.

In this cluster
Questions about this page?
