Realtek SDK vulnerabilities in embedded firmware spread because hundreds of router, repeater and camera makers build on the same Realtek reference code, and as of 2026-11-17 two of its flaws are in CISA's KEV catalog: CVE-2021-35394 in the UDPServer diagnostic daemon and CVE-2021-35395 in the boa and webs management server. Find the SDK branch in your image, check for Realtek's patches and your own copies of the vulnerable code, then ship a fixed build for every affected model and say which ones.
Key takeaways
- Realtek sells Wi-Fi and router chips, and with them it gives customers a software development kit: a Linux or eCos system, drivers, and ready-made daemons for the web interface, WPS, UPnP and factory testing.
- Five CVEs from two Realtek advisories cover most of what a review of a Realtek-based image turns up.
- CVE-2021-35394 became one of the most heavily exploited IoT flaws of 2022.
- CVE-2022-27255 is in the cheaper eCos branch of the SDK, used in low-cost routers and repeaters.
Why does one Realtek SDK flaw reach hundreds of OEM products?
Realtek sells Wi-Fi and router chips, and with them it gives customers a software development kit: a Linux or eCos system, drivers, and ready-made daemons for the web interface, WPS, UPnP and factory testing. Device makers rebrand the web pages, add features and ship. The daemons rarely get a second look, so a bug in Realtek's code lands in every product built on that SDK branch, under hundreds of brand names.
ONEKEY's researchers, who reported the 2021 flaws, describe exactly this. Their advisory counts "at least 65 different affected vendors with close to 200 unique fingerprints", across residential gateways, travel routers, Wi-Fi repeaters, IP cameras, smart lighting gateways and connected toys. They also found the reverse problem: vendors that had received reports of these bugs "fixed them in their own branch but did not notify Realtek, leaving others exposed." A fix that stays in one OEM's tree protects nobody else.
The end user cannot see any of this. Unit 42's analysis of the attacks puts it plainly: "the chipset may not be visible from the exterior of the device", so owners cannot tell whether a given model is affected, and only the device maker's model list can answer that.
This guide covers the SDK as a supply-chain component. For the web servers themselves, Boa, GoAhead, lighttpd and others, see our embedded web server vulnerabilities guide. For UPnP stacks, including Realtek's older miniigd daemon, see our MiniUPnPd and UPnP guide.
Which Realtek SDK CVEs matter most for device makers?
Five CVEs from two Realtek advisories cover most of what a review of a Realtek-based image turns up. Realtek's 2021 AP-Router SDK advisory, released 2021-08-15, lists rtl819x-SDK-v3.2.x, v3.4.x, v3.4T and v3.4T-CT, plus rtl819x-eCos-v1.5.x, as affected by the first four. NVD files them under one record for the "Realtek Jungle SDK", versions 2.0 through 3.4.14B. Scores below are NVD's CVSS 3.1, EPSS values are FIRST's for 2026-11-16, and KEV status is from CISA's catalog version 2026.11.13.
| CVE | Component | What it is | NVD CVSS | EPSS | KEV |
|---|---|---|---|---|---|
| CVE-2021-35394 | UDPServer (MP Daemon) | Command injection and overflows in a diagnostic tool on UDP 9034, no authentication | 9.8 | 0.999 | Yes, added 2021-12-10 |
| CVE-2021-35395 | boa / webs web server | Stack overflows in form handlers, command execution in formSysCmd |
9.8 | 0.981 | Yes, added 2021-11-03 |
| CVE-2021-35393 | wscd (WiFi Simple Config) | Stack overflow parsing the UPnP SUBSCRIBE Callback header | 9.8 | 0.664 | No |
| CVE-2021-35392 | wscd (WiFi Simple Config) | Heap overflow building SSDP NOTIFY replies from the M-SEARCH ST header | 7.5 | 0.821 | No |
| CVE-2022-27255 | eCos SDK SIP ALG | Stack overflow rewriting SIP/SDP data, reachable from the WAN | 9.8 | 0.367 | No |
One numbering trap: Realtek's advisory and NVD give the stack overflow to CVE-2021-35393 and the heap overflow to CVE-2021-35392, while ONEKEY's original advisory lists them the other way round. Quote the descriptions from Realtek or NVD when you map findings, not the researcher's summary.
Older flaws still matter too. CVE-2014-8361, command execution in Realtek's earlier miniigd UPnP daemon, was added to KEV on 2023-09-18, and ONEKEY notes that Realtek replaced miniigd with mini_upnpd and wscd in response.
How was CVE-2021-35394 used by botnets?
CVE-2021-35394 became one of the most heavily exploited IoT flaws of 2022. The UDPServer daemon listens on UDP port 9034 and, per NVD, has "multiple memory corruption vulnerabilities and an arbitrary command injection vulnerability that can be exploited by remote unauthenticated attackers." ONEKEY's write-up shows why: the daemon appends > /tmp/MP.txt to the received buffer after checking only its first three bytes and passes it to system(). The same injection had been reported in 2015 without a CVE.
Unit 42 found that between August and October 2022, attacks on this one CVE made up "more than 40% of the total number of attacks" it recorded, with 134 million exploit attempts by December 2022. Most payloads delivered Mirai, Gafgyt and Mozi, plus a new Golang DDoS bot it named RedGoBot. Unit 42 counted "almost 190 models of devices from 66 different manufacturers", including 31 D-Link, 8 LG, 6 Belkin, 6 Zyxel, 4 Asus and 1 Netgear model in enterprise networks it monitors.
Singapore's Cyber Security Agency followed with alert AL-2023-012 on 2 February 2023, reporting "an increase in exploitation activities" against CVE-2021-35395. It told users to install vendor firmware and, if a device was slow or sending heavy traffic to unknown domains, to factory-reset it, set a strong administrator password and apply all available updates. For a device maker, that advice only works if a fixed image exists for each model.
What is the eCos SIP ALG flaw (CVE-2022-27255)?
CVE-2022-27255 is in the cheaper eCos branch of the SDK, used in low-cost routers and repeaters. Realtek's advisory, released 2022-03-25, says the SIP ALG module "calls strcpy to copy some contents of SIP packets to a predefined fixed buffer and does not check the length", and that an attacker can exploit it "through a WAN interface", causing a crash or remote code execution. It lists rtl819x-eCos-v0.x and v1.x series as affected and names one patch, 20220314_ecos_fix_crash_caused_by_vulnerability_of_sip_alg.rar. NVD's record names eCos RSDK 1.5.7p1 and MSDK 4.9.4p1.
The WAN angle is what makes it serious. The SIP ALG rewrites SIP traffic passing through the router for NAT, so the vulnerable code parses packets from the internet without any service being exposed for management. Faraday Security presented the research at DEF CON 30, and its public materials list 30 vulnerable devices from brands including Tenda, Zyxel, D-Link, Intelbras and Hikvision. Faraday's repository also shows the identification problem on eCos: its checker uses binwalk to find the MIPS endianness and load address, then runs a Ghidra script over the raw image, because there is no file system to list.
How do you identify a Realtek SDK build in a firmware image?
Start with the indicators the SDK leaves in Linux-based images, then confirm with the binaries. ONEKEY's advisory lists the easiest ones:
- rlx-linux. The
/etc/motdbanner shows "RLX Linux", and the init script usually sets the hostname torlx-linux, Realtek's stripped-down Linux for its RTL chips. - /etc/version. When present, it states the chip and SDK release, for example "The SDK version is: Realtek SDK v3.2.3-r14377", along with Ethernet and wireless driver version lines. This is the most direct way to place an image against Realtek's affected-series list.
- Realtek daemons.
/bin/boaor/bin/websfor the web interface,wscdandmini_upnpdfor WiFi Simple Config and SSDP, often with asimplecfg.xmldescription served on port 52881, andUDPServerfor the MP tool. - Handler strings.
formWsc,formSysCmd,formRebootCheck,formWlSiteSurveyandformStaticDHCPin the web server binary are the SDK's form handlers named in the CVE records. - Chip and driver strings. RTL819x part numbers and Realtek wireless driver version strings in the kernel and its modules show which chip family the image targets.
eCos images need a different approach: they are one monolithic RTOS binary, so look for the SIP ALG and other SDK code with a disassembler, as Faraday's tooling does. Many Realtek-based Linux images also carry BusyBox as the userland, which our BusyBox vulnerabilities guide covers separately.
One caution on version strings: they record the SDK the vendor started from, not the patches it applied afterwards. An image that says v3.4.11 may or may not include Realtek's 2021-06-08 boa patch, so confirm the fix in the binary or the build tree.
How does a device maker get fixed SDK drops and backport patches?
Fixes come from Realtek as patch archives per SDK branch, through the chip customer's support channel, not as a new public release. The 2021 advisory names them: wscd and MP daemon patches dated 2021-06-22 for SDK 3.2.3 and 3.4.11E, 2021-07-05 and 2021-06-22 for two 3.4T pre-releases, boa patches dated 2021-06-08 for 3.2.3, 3.4.11 and 3.4T-CT, and a 2021-07-01 patch for eCos 1.5.3. ONEKEY adds that the Jungle SDK patches "need to be backported", that the 2.x branch was no longer supported by Realtek, and that the separate "Luna" SDK fixed these issues in version 1.3.2a.
A workable process for a device maker:
Inventory by SDK branchFor every shipped and supported model, record the chip, the SDK branch and revision, and the original design manufacturer if the board came from one. Many brands buy finished designs, so the SDK tree may sit with the supplier, not with you. Request the patch set in writingAsk Realtek or your supplier for the patch archives named in each advisory for your exact branch, and for the list of CVEs each one closes. Apply and diff your customisationsRealtek's patch fixes Realtek's handlers. NVD's CVE-2021-35395 record warns that "any binary based on Realtek SDK webserver will probably contains its own set of issues on top of the Realtek ones", because vendors copy the SDK's insecure examples. Grep your own handlers for unchecked strcpy,sprintfandsystem()calls; our memory corruption fix guide covers the replacements.Remove what production does not needUDPServer is a factory test tool. Strip it, formSysCmdand similar diagnostic paths from release builds, and turn the SIP ALG off by default where customers do not need it.Verify in the binaryRebuild, then confirm the fix in the shipped image, by re-testing the patched handlers or by diffing the old and new binaries, before you mark a model fixed.
How should you tell customers which models are affected?
Publish a model-level advisory that maps each product to the SDK CVEs and to a firmware version. Zyxel's advisory for CVE-2021-35392 and CVE-2021-35393 is a usable pattern. It names four supported models with a fixed firmware version and a target date for each, such as NBG6615 V1.00(ABMV.6)C0 in the middle of November 2021. It states that five older models from the researcher's report are end of life and will get no update, recommends replacing them, and keeps a revision history: first release 2021-08-16, then 2021-08-26 and 2021-09-16 as models were added.
That last point matters for SDK flaws. The affected list grows as you check more of your catalogue, so date each revision. Cite the CVE IDs, not only "Realtek SDK issue", so customers' scanners can match them, and say plainly which products will not be fixed. If you publish machine-readable advisories or VEX, an entry per model and firmware version lets customers filter out the products that are not affected.
How does AI-assisted analysis help with Realtek SDK findings?
SoftWSP's automated scan unpacks a firmware image, catalogs the software packages it can identify in the extracted file system with their versions (the inventory our firmware SBOM guide describes), and matches each one against CVE records, CISA's KEV catalog and FIRST's EPSS scores. This is deterministic code: the code scores, and AI narrates. You can inspect the resulting CycloneDX file in the free SBOM viewer.
That matching has a clear limit with Realtek SDKs. NVD files the 2021 CVEs under cpe:2.3:a:realtek:rtl819x_jungle_software_development_kit, but Realtek's daemons usually ship without a package record, so a package catalog sees BusyBox and the kernel and not "Realtek SDK 3.4.11". eCos images have no file system to catalog at all. Recognising the SDK branch from /etc/version, rlx-linux and the boa, wscd or UDPServer binaries, and deciding whether Realtek's patch and your own handler fixes are present, is the engineer's part of the review. AI writes the plain-language summary of each finding, citing the component, version and record behind every sentence, and an engineer checks it before the report is delivered.
To have a specific Realtek-based image reviewed, see our device and firmware security services.
This guide draws on Realtek's AP-Router SDK advisories for CVE-2021-35392 to CVE-2021-35395 and for CVE-2022-27255, ONEKEY's 2021 advisory, Unit 42's January 2023 analysis, CSA Singapore alert AL-2023-012, Faraday Security's CVE-2022-27255 materials, Zyxel's WiFi Simple Config advisory, NVD, CISA's KEV catalog (version 2026.11.13) and FIRST EPSS data dated 2026-11-16, all read directly on 2026-11-17.
Put this guide to work
Frequently asked questions
Which Realtek SDK vulnerabilities matter most for device makers?
As of 2026-11-17, two Realtek Jungle SDK flaws are in CISA's KEV catalog (version 2026.11.13): CVE-2021-35394, command injection and overflows in the UDPServer MP diagnostic daemon, added on 2021-12-10, and CVE-2021-35395, overflows and command execution in the boa and webs management web server, added on 2021-11-03. Both affect Jungle SDK v2.x up to v3.4.14B per NVD, score 9.8 and have FIRST EPSS scores above 0.98. CVE-2021-35392 and CVE-2021-35393 in the wscd WiFi Simple Config daemon and CVE-2022-27255 in the eCos SDK's SIP ALG follow.
How do I tell whether a firmware image is built on the Realtek SDK?
On Linux-based images, look for an /etc/motd or hostname that says rlx-linux, an /etc/version file that prints a line such as 'The SDK version is: Realtek SDK v3.2.3-r14377', and Realtek's daemons: boa or webs for the web interface, wscd and mini_upnpd for WiFi Simple Config and UPnP, and UDPServer for the MP diagnostic tool. Strings such as formWsc, formSysCmd and formRebootCheck in the web server binary point to the SDK's form handlers. eCos-based images are one monolithic RTOS binary with no file system, so they need disassembly with the right base address instead of a file listing.
Where does a device maker get fixes for Realtek SDK vulnerabilities?
From Realtek, through the SDK support channel the chip customer already has. Realtek's advisories name patch archives per SDK branch, for example 20210608 boa patches for SDK 3.2.3, 3.4.11 and 3.4T-CT and a 20210701 patch for eCos 1.5.3, and a 20220314 patch for the eCos SIP ALG flaw. Realtek's public PDFs do not include the patches themselves. The device maker has to apply them to its own SDK tree, check that its customised handlers carry the same fix, rebuild and ship firmware per model.
Is a Realtek SDK fix enough if the vendor customised the web server?
Often not. NVD's record for CVE-2021-35395 notes that exploitability depends on what each manufacturer did with the SDK web server, and that vendors who reuse its insecure example code will 'probably' carry their own issues on top of Realtek's. After applying Realtek's patch, audit every custom form handler for unchecked strcpy, sprintf and system calls, and remove diagnostic handlers such as formSysCmd and the UDPServer daemon from production builds.
Does SoftWSP detect the Realtek SDK in my firmware automatically?
SoftWSP's automated scan catalogs the software packages it can identify in an unpacked firmware file system, with versions, and matches them against CVE records, CISA's KEV catalog and FIRST EPSS scores. Realtek's SDK daemons usually carry no package record, so recognising the SDK branch from indicators such as /etc/version, rlx-linux and the boa, wscd or UDPServer binaries, and mapping it to NVD's Realtek SDK record, is part of the engineer review. AI writes the plain-language summary of each finding, citing the component, version and record, and an engineer checks it before delivery.
Sources
- Realtek AP-Router SDK Advisory (CVE-2021-35392/35393/35394/35395), 2021-08-15
- Realtek AP-Router SDK Advisory (CVE-2022-27255), 2022-03-25
- ONEKEY Research: Advisory: Multiple issues in Realtek SDK affect hundreds of thousands of devices down the supply chain
- Unit 42: Realtek SDK Vulnerability Attacks Highlight IoT Supply Chain Threats
- Cyber Security Agency of Singapore AL-2023-012: Critical Vulnerability In Realtek Jungle Software Development Kit (SDK)
- Faraday Security (infobyte): CVE-2022-27255 Realtek eCos SDK SIP ALG buffer overflow, DEF CON 30 materials
- Zyxel Security Advisory for WiFi Simple Config Buffer Overflow Vulnerabilities
- NIST NVD: CVE-2021-35392
- NIST NVD: CVE-2021-35393
- NIST NVD: CVE-2021-35394
- NIST NVD: CVE-2021-35395
- NIST NVD: CVE-2022-27255
- CISA: Known Exploited Vulnerabilities (KEV) Catalog
About this guide
SoftWSP Firmware Security Desk
Covers the Firmware vulnerabilities & fixes cluster: 129 guides.

In this cluster
Questions about this page?
