Skip to content

XCPE botnet: Google ties 1.3M TV boxes to Videostrong

Filip LuchianencoUpdated 7 min read

Last updated: 2026-11-17

On November 2, 2026, Google Threat Intelligence Group (GTIG) published "Shutting the Door on the XCPE Botnet", its report on a backdoor preloaded in the firmware of Android TV set-top boxes. From the report: "Last week, Google and industry partners dismantled the XCPE backdoor which enrolled more than 1.3 million set-top boxes into a botnet. We assess with high confidence that the backdoor and botnet were intentionally developed, deployed, and operated for financial gain by Videostrong Technology Co., an Android Original Equipment Manufacturer (OEM) and Original Design Manufacturer (ODM)." Per GTIG, the backdoor turned off Google Play Protect app verification and installed modules developed by other companies, most often residential proxy SDKs, through which the owner's IP address is rented out as an exit node. Eight residential proxy and bandwidth-sharing modules are named in the report, among them Honeygain's AppSalt, GaGaNode, NetNut's Popanet, Traffmonetizer and EarnFM's Fleet SDK.

The attribution to Videostrong is GTIG's assessment. Per The Record's November 3 article, Videostrong did not respond to requests for comment. No public statement from Videostrong was found on 2026-11-17.

What is the XCPE backdoor?

Per GTIG, the preinstalled malware has had a "modular, three-stage architecture" since at least late 2024; earlier versions were single-stage droppers:

  • Backdoor. The package x.cpe (com.oemservice in some builds) is preloaded in the firmware and runs with the android.uid.system privilege. After a user uninstalls it, it is reinstalled from the /oem/ partition at the next reboot.
  • Loader. com.abc.nexus receives instructions from x.cpe and starts the monetization modules. Many variants extract libseccomp_guard.so, an anti-analysis library, before the modules start.
  • Monetization modules. Modules developed by companies other than Videostrong are run as native ELF daemons, as SDKs compiled into the com.abc.nexus code, or as secondary APKs downloaded from the operator's servers.

Since August 2026, single-module droppers named com.android.system.ntp and com.android.sso.auth.handler have been installed over the command channel; per GTIG, these installs "appear to be synchronized with the mass-uninstallation of com.abc.nexus across the botnet." In multiple variants, libexploit.so restarts adbd as root, and the dropper copies itself into /system/priv-app/ with the INSTALL_PACKAGES and REBOOT permissions.

At the first connection, the serial number, MAC address, model and firmware version of the box are sent in an encrypted HTTP POST to a bootstrap host under anyevonline[.]com (earlier builds used tms, ott and ota hosts under szbboys[.]com). The bootstrap routine checks the Android system property ro.mecool.device to choose the host. The response contains the MQTT broker details; every sample GTIG identified connects to one broker, mqtt.szbboys[.]com, on ports 3885 to 3887.

The MQTT command set includes silent app installs, shell commands as the system user, key injection and screen capture. GTIG writes: "We also observed Videostrong operators executing sensitive actions on the device, including capturing live screenshots; injecting keystrokes; exfiltrating sensitive data including account information, device information, package inventories, and system logs".

Most infected devices were Android TV boxes; GTIG also found the backdoor on "telehealth devices, karaoke machines, and digital signage devices." TrendAI, one of the partners, counted 1,472,554 infected devices across 206 models in more than 170 countries between June and October 2026. Synthient, another partner, published the 182 device model strings it observed in the campaign, among them MECOOL, KM9PRO and KM7SE_GTV, with the caveat: "A model name alone is not sufficient to confirm compromise."

How did XCPE turn off Google Play Protect?

AppUtils.disablePackageVerify() is called on an MQTT command; per the report, it "targets the global setting package_verifier_user_consent via Settings.Global.putInt and sets it to -1, disabling Google Play Protect app verification." From the report: "The backdoor is able to silently download additional files to the device and execute them without user consent because it first wipes user consent configuration on the device and then disables Google Play Protect".

GTIG writes: "Videostrong employed obfuscation techniques to exploit a gap in our tests and load malicious code after the device's first boot. As a result, Google has implemented changes to its existing review process to better safeguard against similar activity in the future." The droppers were installed over MQTT and then copied into the system partition through a root ADB shell. Android developer verification, which began on September 30, 2026 in four countries, applies to installs from participating stores on certified devices. ADB installs are outside it.

Which proxy and ad fraud modules did XCPE load?

From the report: "Most often the modules are associated with residential proxy services and bandwidth sharing, but we've also observed ad fraud modules," with "as many as three residential proxy SDKs installed on a single device."

Module Provider, per the report Identifier in the report How it was started on XCPE devices
AppSalt Honeygain, owned by Oxylabs API key from libnative_keys.so (getAppsaltKey()) com.abc.nexus CoreService; optIn() called with no consent screen
GaGaNode IPCola com.gaganode.sdk embedded in com.abc.nexus; starts Popanet and Fleet SDK
Popanet NetNut io.popanet.Popa started by GaGaNode
Traffmonetizer Traffmonetizer API key from libnative_keys.so (getTraffmonetizerKey()) com.abc.nexus CoreService
Fleet SDK EarnFM libfleetshare.so started by GaGaNode
BWM / IPWeb not stated BwmService (SN4Service) com.abc.nexus service
Linkfog Beijing Linkfog Technology Go binary lw_android_agent com.abc.nexus LinkfogService
Jaguar not stated libjaguar.so com.abc.nexus AdtimeService

Source: GTIG report, "Monetization" and "Indicators of Compromise". Backdoor and ad fraud modules are also named in the report: Spirit, Moyu and Bird (Moyu and Bird are linked in the report to the BadBox enterprise), Fission (found inside the video app com.newtv.cboxtv) and HiddenView (com.ota.service).

Per GTIG, the Popanet module "is identical to the malware described in our previous disruption of NetNut", the July 2, 2026 action in which GTIG estimated the NetNut network at "at least 2 million devices".

From the AppSalt section: "Appsalt SDK is initialized without displaying a consent screen - instead it explicitly calls the optIn() method which sets a consent flag without displaying any user interface. During Appsalt SDK startup it checks the application's installation source and stops execution if the application was installed from the official Google Play Store."

From the report's conclusion: "In our analysis, these proxy operators only seek consent when they are forced due to platform policy and enforcement. Silent opt-in without user consent remains the most prevalent means by which devices are turned into exit nodes."

Oxylabs sent a statement to KrebsOnSecurity after its November 2 article was published; the statement was added in an update dated November 5. Per that update, Oxylabs said its system accepts a device as an exit node only after a consent signal, and that the developer, unnamed in the update, faked the signal. Quoted in the update: "As soon as we saw that, we suspended the relationship and sent them a formal legal inquiry," and "This supplier accounts for less than 1% of our IP address pool. We had a signed contract with them, they assured our QA team that users had given consent, and we saw the consent flow ourselves." Per the same update, Oxylabs has no record of Videostrong and identified com.abc[.]nexus among its suppliers.

On 2026-11-17, the Oxylabs statement was found only in the KrebsOnSecurity update; the Oxylabs and Honeygain websites had no copy of it. No statement from IPCola, NetNut, Traffmonetizer, EarnFM, Linkfog or the operators of BWM/IPWeb and Jaguar was found on the same date.

What did Google do?

GTIG lists four actions:

  • "Legal action to take down the domains and servers Videostrong used to control the botnet and install malware modules"
  • "Legal action to take down C2 domains used by over a dozen malware families loaded as modules via this backdoor"
  • "Banned Videostrong and any future subsidiaries or affiliates from building and selling certified Android devices in the future, or supplying software to other certified vendors"
  • "Implementing improvements to how we ensure the integrity of firmware on certified Android TV devices"

Per GTIG, the affiliates that supplied the modules "each have diversified distribution channels for their modules beyond XCPE", and GTIG assesses that taking down their command domains "will sever access to at least 10 million additional compromised devices". Partners named in the report are Japan's National Police Agency, Lumen, HUMAN Security, Plume, Synthient and TrendAI. No court, case number or order for the legal action is named in the report. Web searches for a "Videostrong" docket on 2026-11-17 found no court or case number; the PACER and CourtListener docket indexes, which need an account, were not searched. The court record, once found, would establish what was ordered and against whom.

What do an Android app scan and a traffic capture show for proxy SDKs?

The XCPE backdoor is in the box firmware, and its modules were loaded after first boot. A CanITrustThat scan inspects one app package at a time. For five of the proxy SDKs in the report, the scan records their presence in an Android app package:

  • GaGaNode: classes under com.gaganode.sdk. or the host gtxvdqvuweqs.com (the host is also in GTIG's indicator list).
  • NetNut Popanet: classes under io.popanet.
  • Honeygain: classes under com.honeygain or the string sdk.honeygain.com.
  • Traffmonetizer: the hosts blnc.traffmonetizer.com or srv.traffmonetizer.com (Synthient lists both hosts in the com.abc.nexus builds it examined).
  • EarnFM: api.earn.fm or classes under com.earn_fm.

The scan also records the Eclipse Paho MQTT client when it is in the same package as a proxy SDK, whether the text "Bandwidth Sharing Policy" is in the package, and the MoYu payload classes and hosts published for BadBox 2.0. The report gives no AppSalt class namespace, so whether the AppSalt build matches the Honeygain rule is not established; whether the XCPE Moyu build matches the BadBox 2.0 indicators is also open. Each match is presence in the build: the SDK code is in the app. Whether the SDK ran, and whether a consent screen was shown first, is established only by a capture.

In a network capture on a test phone, each connection from a proxy SDK to its servers is recorded with the host, the time, and whether the connection opened before or after the app's consent screen was accepted. A connection that cannot be decrypted is recorded with its server address only. Per the report, the AppSalt build on XCPE devices stops when its app was installed from Google Play, so the install source of the tested app is part of the test. The difference between presence and transmission and the limits of each method are on the methodology page. The same scan and capture questions are covered in earlier posts on Midnight Mimosa, firmware malware paid through an ad SDK, on the AdFection spyware ads shown in 75,373 apps and on an exported activity in the EngageLab push SDK.

I run these investigations for lawyers and for researchers working with lawyers, on one app or on a list of apps in bulk. Sign up for a CanITrustThat account to run your own research, let us run an investigation for you, or read the app teardowns.

App privacy law, applied to real apps.

Posts on new laws, fines and studies, and teardowns of the apps those laws apply to.

How we use your address: privacy notice. Prefer a feed reader? Use the RSS feed.