HUMAN Security says it disrupted BADBOX 3.0, which sent roughly 40 billion fake ad requests a day at its peak.
The malware runs on cheap, uncertified Android devices built in mainland China, mainly low-cost streaming TV boxes.
Infected devices were also enrolled in 11 residential proxy networks without their owners knowing or agreeing.
Google urges people to check that a device is Play Protect certified and built with Android TV OS.
Quick read · 1 min
HUMAN Security says it has disrupted BADBOX 3.0, a botnet hidden inside cheap Android streaming boxes. At its peak it sent about 40 billion fake ad requests a day, more than three times the previous record for an ad fraud operation.
The boxes were infected before buyers plugged them in. They played video ads in hidden windows and were enrolled in 11 residential proxy networks, which rent out a home internet connection to paying strangers. Owners were never asked.
The devices are uncertified, built in mainland China and run the open-source version of Android.
They lack Play Protect certification, and they don't run Android TV OS.
FBI warning signs from the earlier wave include unfamiliar brands and boxes sold as unlocked for free content.
Google says to check that your device is Play Protect certified. HUMAN expects the operators to adapt and is still watching them.
BADBOX 3.0 turned cheap streaming boxes in people’s living rooms into a machine that fired off roughly 40 billion fake advertising requests a day. HUMAN Security, the company that tracked it down, says no ad fraud operation ever made public has been bigger by that measure.
If you own a low-cost, no-name Android TV box, this story is about your home network. Advertisers were the ones being robbed, but the same boxes also rented out their owners’ internet connections to strangers without asking, and the malware was in place before the box was ever plugged in. So what is BADBOX 3.0, what changed from the version before it, and how do you tell whether a device on your shelf is a risk? The FBI and Google have both given pointers, and they are below.
01
What is BADBOX 3.0?
BADBOX 3.0 is the third known generation of a botnet, which is a network of hijacked devices that criminals control from a distance. It lives inside cheap, uncertified Android gadgets. HUMAN’s Satori research team announced on Nov. 18 that it had disrupted the operation.
The devices carry new backdoors, hidden ways in that let someone else send them instructions. Once a box is online it contacts the operators’ control servers and pulls down fraud modules from a catalog that is actively kept up to date. That catalog sits in public code repositories, with copies in several backup locations so that one takedown doesn’t kill it. HUMAN’s researchers recognized some of the modules from the original BADBOX and from BADBOX 2.0.
The malware has two main jobs:
Hidden ads. It plays video ads in windows the owner can’t see. The ad requests pretend to come from legitimate streaming apps, so the traffic looks real to the companies paying for it.
Proxying. It signs the device up to residential proxy networks, services that let paying customers route their traffic through ordinary home connections. Owners were never told and never agreed.
HUMAN also saw a third trick called server-side ad insertion spoofing. In plain terms, advertisers were sent requests that made it look as if an ad had played on a streaming device when there was no device and no ad at all.
02
How big did BADBOX 3.0 get?
At its height, BADBOX 3.0 accounted for about 40 billion bid requests a day, according to HUMAN. A bid request is the automated message that offers an ad slot to advertisers, so each fake one is a chance to collect money for an ad nobody watched.
That volume was more than 4% of all the connected TV traffic HUMAN’s platform observed. It was also more than three times the size of the previous record holder, an operation known as VASTFLUX. HUMAN says the peak was larger than the past five years of its published Satori investigations put together.
The proxy side was large too. The infected devices were enrolled in 11 residential proxy networks. One of them, called PXYEDGE, advertised roughly 200,000 proxy nodes, while the researchers’ own monitoring showed unique connections peaking at around 340,000. HUMAN assesses PXYEDGE to be the renamed successor to IpMoYu, a proxy service tied to BADBOX 2.0, and says researchers at Nokia reached similar conclusions on their own.
The two halves fed each other. Some of the fake ad traffic was sent through those home connections so that it looked like ordinary households and slipped past anti-abuse filters.
03
Which TV boxes and devices are affected?
The affected devices are low-cost, uncertified consumer gadgets, with cheap streaming TV boxes at the center, along with budget phones and tablets. HUMAN describes them this way:
They are built in mainland China and shipped around the world.
They run the Android Open Source Project, the free base version of Android that anyone can put on hardware.
They are not Android TV OS devices.
They are not Play Protect certified, meaning they haven’t passed Google’s security testing.
HUMAN’s chief information security officer, Gavin Reid, put it bluntly: anyone who bought one of these boxes got a device that was already compromised when it came out of the packaging.
The researchers traced the BADBOX 3.0 module catalog to a Chinese electronics manufacturer they name as TXCZ. The links include preinstalled software such as a factory-installed TV launcher, checks in the code for specific devices, domains, control servers and a code repository hosted under an account name that includes TXCZ.
The earlier wave gives a sense of how wide the net can be. When the FBI warned about BADBOX 2.0, the devices tied to it included digital picture frames, projectors, aftermarket car infotainment systems, tablets and TV streaming devices. More than a third of those were in Brazil, and the United States was among the other countries with large numbers.
04
What changed since BADBOX 2.0?
The method is the same as before, but the scale is new. BADBOX first turned up in early 2023, on a T95 streaming box that had been ordered from Amazon. By December 2024, cybersecurity agencies in Germany were warning that tens of thousands of devices were infected.
Then came the big push against BADBOX 2.0. In March 2025, HUMAN and Shadowserver cut more than 500,000 infected devices off from their control servers, and Google pulled 24 malware-spreading apps out of its Play store. A lawsuit against the operators followed, with Google saying BADBOX 2.0 had infected over 10 million Android-based devices by April 2025. It also updated Play Protect to block known apps linked to the botnet automatically.
None of that stopped the people behind it from rebuilding. HUMAN warned at the time that its disruption could not dismantle the supply chain that lets the backdoor be planted in devices on their way to shoppers. With the original BADBOX, the backdoor sat in a part of the device’s built-in software that can’t be written to, so it could not be uninstalled.
This time, HUMAN says it took part in Google’s disruption of a related malware operation called XCPE earlier this month, which took down several control domains that BADBOX 3.0 used. One of the malware repositories has been removed, and HUMAN says its systems now filter out the fraudulent ad traffic.
The clearest test is certification. Google’s advice is to make sure your device is Play Protect certified and built with Android TV OS, and HUMAN adds that Google Play Protect should be turned on.
The FBI’s BADBOX 2.0 warning listed signs that a device may be infected. Treat a box as suspect if:
It was sold under a brand you don’t recognize.
It is a generic streaming device advertised as unlocked or able to get free content.
It asked you to switch off Google Play Protect, or it isn’t Play Protect certified.
It has suspicious apps or app marketplaces installed.
You have noticed unexplained or suspicious internet traffic at home.
If a device raises those flags, the FBI urged people to consider disconnecting it from their network. The bureau also told users to avoid downloading apps from unofficial marketplaces and to keep connected devices updated.
Why does it matter if the box still plays your shows? Because the proxy side puts strangers’ traffic on your home connection. In the BADBOX 2.0 case, the FBI said access to compromised devices was sold through residential proxy services, which make it easier to run account takeovers, create fake accounts, spread malware and launch attacks that knock websites offline. HUMAN’s researchers said then that a backdoored device could be told to carry out any attack its controllers came up with.
For future purchases, the security advice is simple: buy certified Android devices from reputable manufacturers, and check that Play Protect is available and active.
06
What happens next
HUMAN expects the operators to adapt to this disruption, as they did after the earlier ones. Its Satori researchers say they are continuing to watch the group for its next move.
07
Quick answers
What is BADBOX 3.0?
It is a botnet built into cheap, uncertified Android devices, mostly streaming TV boxes. It plays hidden video ads to defraud advertisers and enrolls the device in proxy networks that rent out the owner’s internet connection.
Are certified Android TV devices part of BADBOX 3.0?
No, according to HUMAN: the infected hardware doesn’t run Android TV OS and lacks Play Protect certification. Google encourages users to check that their device is certified and built with Android TV OS.
Can the BADBOX backdoor be removed?
With the original BADBOX, the backdoor was embedded in a non-writable part of the firmware and could not be uninstalled. The FBI’s advice for suspicious devices was to consider disconnecting them from your network.
Brussels says hackers used a publicly available AI tool to hit its internal systems in October. Here is what went down, what is back, and what nobody will say yet.
A report says intruders got into the database officers check at traffic stops. Here is what is confirmed, what the system holds and what nobody knows yet.
Hospital staff in England looked at patient files without good reason over 3,000 times in five years. Here is how to read your own record and push back.
6 min read
We use cookies to understand how readers use Talk With Tech, so we can make it better. Is that OK? Privacy policy
The Daily Brief
Today's biggest tech stories, in 5 minutes
Every morning, the news that matters from AI, phones, apps and the people shaping tech. Explained in plain English. Free.
One email a day. No spam, unsubscribe anytime. Privacy policy