BADBOX 3.0: Cheap Android TV Boxes Ran the Biggest Ad Fraud Ever Disclosed. Check Your Box, Then Check Your Guest Wi-Fi.

If you own an off-brand Android TV box, check it tonight. If you run a network, look for those boxes on it this week. HUMAN Security's Satori team says BADBOX 3.0 is the largest ad fraud operation ever disclosed by bid volume, and it lives inside cheap streaming boxes sold around the world.
Do this first
At home
- Open the Google Play Store on the box, tap your profile, then Settings, then About, and read the line "Play Protect certification". HUMAN's advice, as reported by CyberInsider, is to check this. If it says the device is not certified, or the box has no Play Store at all, treat it as suspect.
- Compare the brand and model against Google's list of certified partners at android.com/certified/partners. An unknown brand, a box that came with its own app store or a pile of preinstalled apps you never chose, or a "fully loaded" box sold with free channels are all warning signs.
- If it fails, unplug it. A factory reset will not help: the coverage describes the backdoor as a system app shipped on the device, which is the part a reset restores. Replace it with a certified box from an established brand.
On a network you run
- Look for TV boxes, projectors and tablets on corporate and guest Wi-Fi. DHCP leases, MAC vendor prefixes and hostnames are the cheapest inventory. These devices should not be on a corporate network at all.
- Put anything you must keep on an isolated segment with no route to internal systems.
- Watch egress from those segments: a streaming box that holds long-lived connections to proxy services, or sends steady traffic to many unrelated hosts while nobody is watching anything, may be serving as a residential proxy exit. HUMAN says the malware enrolled boxes in one of 11 residential proxy networks.
- Block the command servers you can. The five ThreatFox-labeled BADBOX addresses we hold are listed below. They are a third-party list, not HUMAN's BADBOX 3.0 indicators, which we could not read.
Am I affected?
You are likely affected if all three hold: the box runs Android but fails the Play Protect certification check, it came from a brand you cannot find on Google's certified list, and it shipped with an app store or launcher other than Google's. On a network, you are affected if any such device is online and has a path to the internet that you do not monitor. Its traffic leaves from your public address, so ad fraud and other people's proxied traffic will look like it came from you.
What is confirmed
These come from HUMAN's report as described by BleepingComputer (Bill Toulas, November 21) and CyberInsider (Bill Mann, November 20). HUMAN's own blog post returned HTTP 403 to us, and in a browser it served a bot challenge, which we did not attempt. So every fact below is HUMAN's research as relayed by two outlets. Treat it as one source.
- Scale. HUMAN, as quoted by BleepingComputer: "At its height, BADBOX 3.0 accounted for roughly 40 billion bid requests a day." That is more than three times the previous record holder, VASTFLUX, and over 4% of the connected-TV traffic HUMAN's platform saw. CyberInsider reports the surge was visible by mid-2026.
- Devices. Inexpensive, off-brand streaming boxes made in China, plus phones and tablets, running the Android Open Source Project without Play Protect certification.
- Infection route. Preinstalled, not downloaded. The backdoor is packaged as a system app, and BleepingComputer describes a factory-installed TV launcher. The device then pulls fraud modules from public code repositories, with copies mirrored across hosting platforms.
- Who built it. HUMAN linked the module catalog to the Chinese electronics maker TXCZ, citing a company domain that hosted payloads and supplied hidden-ad settings, modules that check for TXCZ-specific device properties, and a repository maintainer.
- What it does. Plays video ads in hidden windows while impersonating real streaming apps. Sends fake server-side ad insertion requests claiming an ad played in a stream to a device that did not exist. Enrolls the box in one of 11 residential proxy networks without the owner's consent; 10 are commercial services and the eleventh, PXYEDGE, is according to HUMAN a rebrand of IpMoYu, the proxy network from BADBOX 2.0. PXYEDGE advertised about 200,000 nodes; HUMAN saw unique connections peak near 340,000. For one impersonated app, about 80% of US requests came from addresses flagged as residential proxy exits.
- Disruption. HUMAN says it disrupted the operation and now filters its ad traffic. One malware repository was taken down. Google's takedown of the related XCPE botnet earlier this month also removed several domains that controlled parts of BADBOX 3.0.
What is not known
- Device models. Neither outlet names a model or retail brand. The only brand in the coverage is TXCZ, the manufacturer. The original 2023 BADBOX was found on a T95 box; that tells you nothing about which boxes carry 3.0.
- Indicators. Neither outlet published a command-server domain, IP or hash. HUMAN's report may list them; we could not read it.
- Account takeover. The 3.0 coverage describes ad fraud and proxy enrollment. It does not describe account takeover, so we do not claim it.
- Device count. No infected-device total was reported for 3.0. The 10 million figure belongs to BADBOX 2.0 (Google's lawsuit, April 2025) and the 1.3 million figure to XCPE.
What our data shows
- Indicators held. Our corpus holds five IP:port entries labeled BADBOX, all ingested July 19, 2026 from the ThreatFox feed at confidence 70: 165.154.202[.]29:9999, 144.217.243[.]201:80, 172.96.114[.]6:80, 98.98.71[.]19:80 and 98.98.71[.]20:80. Our exact-match lookup confirms each one as held, in a single lane. This is coverage via a redistributed third-party feed, not a detection and not a lead. ThreatFox's label does not say which BADBOX generation they belong to, and we cannot tie them to 3.0.
- They do not ship in our default blocklist. We pulled the default ips.csv today (5,917 rows, full download) and none of the five appear. Holding a record is not the same as publishing it. The reason is by design: the default file carries bulk third-party feed records for seven days after they arrive, and these came in from ThreatFox in July, so they have aged out. Only campaign indicators that we curate and attribute ourselves stay in past that window. We will not relabel ThreatFox's records as our own to get them back in. If HUMAN's report publishes BADBOX 3.0 command servers, we will add those under this campaign.
- The proxy brand. ipmoyu.com is held since July 17, 2025 (tweetfeed) and again from ThreatFox, plus two ipmoyu.xyz hostnames from ThreatFox since March 20, 2026. All third-party, all coverage. We hold nothing for PXYEDGE.
- No first-party sighting. We checked all five addresses against our edge since July 1: 764,931 edge blocks, 200,633 edge canary hits and 7,660 origin honeypot captures, zero matches. That is expected, not reassuring: an infected box calls out to its command server, so these addresses would show up in your egress logs, not at our edge.
- Nothing new fed. We feed only indicators a primary publishes, and none were published in the sources we could read. We will not feed the home addresses of infected boxes: residential proxy exits are victims, not threats.
Sources
- HUMAN Security, Satori Threat Intelligence: "BADBOX 3.0 Hides the Largest Ad Fraud Operation Ever Disclosed Inside Infected TV Boxes" (returned HTTP 403 to us): https://www.humansecurity.com/learn/blog/satori-threat-intelligence-disruption-badbox-3-0-hides-the-largest-ad-fraud-operation-ever-disclosed-inside-infected-tv-boxes/
- BleepingComputer, November 21, 2026: https://www.bleepingcomputer.com/news/security/badbox-30-botnet-returns-with-record-breaking-40-billion-daily-ad-requests/
- CyberInsider, November 20, 2026: https://cyberinsider.com/badbox-3-0-malware-used-android-tv-boxes-to-generate-40-billion-daily-bid-requests/
- Google certified Android TV partners: https://www.android.com/certified/partners/
Read next: Google Took Down a 1.3 Million-Box Android TV Botnet That Shipped in the Firmware, the XCPE takedown that also hit BADBOX 3.0 infrastructure, with the package names and settings to check on a box.
Five percent of this is wrong somewhere. It always is.
Was this useful? Rate this post. The widget is at the bottom of the page, and we read every response.
How do AI models see YOUR brand?
AIPM has audited 250+ domains. 15 seconds. Free while still in beta.
Audit your site now → aipmsec.com
Was this useful? Thirty seconds, no cookies, no tracking, no third parties, your address hashed and never stored. The question lives here: https://analytics.dugganusa.com/nps.html#post=badbox-3-tv-box-ad-fraud-40-billion