"At its height, BADBOX 3.0 accounted for roughly 40 billion bid requests a day and more than 4% of all CTV traffic observed by the Human Defense Platform," reports HUMAN.
HUMAN's Satori researchers map a record-breaking return
HUMAN's Satori team, which uncovered and disrupted the latest BADBOX variant, reports that BADBOX 3.0 has re-emerged and grown into the largest documented ad-fraud operation by daily bid request volume. The researchers say the botnet generated roughly 40 billion fraudulent advertising bid requests per day, a peak the report compares to — and surpasses — prior large operations: BADBOX 3.0 produced more than three times the volume of the previous record holder, VASTFLUX, and represents a larger peak than the last five years' worth of Satori investigations combined.
The Satori researchers also say BADBOX 3.0 accounted for more than 4% of all connected TV (CTV) traffic observed by the Human Defense Platform at its height, underscoring the scale and the platform-level visibility of the operation.
How BADBOX 3.0 conducts ad fraud and proxying
The botnet's functionality follows established BADBOX tactics. BADBOX infects inexpensive Android-based phones, tablets, and streaming TV boxes running the Android Open Source Project (AOSP). Once installed, the malware performs ad fraud by playing video advertisements in hidden windows and impersonating legitimate streaming applications so that requests appear genuine to ad networks and anti-abuse systems.
HUMAN also observed fraudulent server-side ad insertion (SSAI) requests that made advertisers believe ads were played on streaming devices even when no device or video ad was involved. In addition to ad fraud, BADBOX 3.0 includes a proxying module that enrolled infected devices into one of 11 residential proxy networks without owners' knowledge or consent; SOME of the ad-fraud traffic was routed through these residential proxies to appear legitimate and evade anti-abuse filters, the researchers say.

The cyber insurance questionnaire just landed. Now what?
SOC 2, HIPAA, insurance renewals - someone has to own security strategy. Nubivance provides fractional CISO leadership without the full-time salary.
Get a security leadPXYEDGE, IpMoYu, and the proxy infrastructure
HUMAN identified one of the residential networks, PXYEDGE, as the rebranded successor to IpMoYu, a residential proxy service previously linked to BADBOX 2.0. According to the researchers, PXYEDGE advertised roughly 200,000 proxy nodes, while monitoring showed unique connections peaking at about 340,000 — a discrepancy that illustrates both the scale and the fluidity of the proxy layer BADBOX operators used.
HUMAN further notes that infected devices fetched malware modules from public code repositories and that the same payloads were mirrored across multiple hosting platforms for redundancy and to resist takedown efforts. That mirrored hosting, together with the proxying infrastructure, helped BADBOX 3.0 generate and hide massive volumes of fraudulent traffic.
Links to device supply chain and prior disruptions
HUMAN's investigation linked BADBOX to a specific Chinese electronics manufacturer through multiple signals: domains, device-specific checks in the code, a factory-installed TV launcher, and a repository maintainer. BADBOX was first discovered in early 2023 on a T95 Android TV box purchased on Amazon, and Germany's cybersecurity agencies warned in December 2024 that the botnet had already infected tens of thousands of devices.
Previous disruption efforts reduced BADBOX's reach but did not eliminate it. In March 2025, HUMAN and Shadowserver cut off more than 500,000 BADBOX 2.0 devices from their control servers, while Google removed 24 apps from Google Play that had spread the malware. Google later sued the botnet's operators, saying BADBOX 2.0 had infected over 10 million Android-based devices by April 2025. HUMAN also participated in Google's disruption of the related XCPE operation earlier this month, taking down multiple command-and-control domains used by BADBOX 3.0; one repository used to host malware modules was taken down as well, and HUMAN says its systems now filter fraudulent advertising traffic associated with BADBOX 3.0.
What this means for advertisers, device manufacturers, and end users
- Advertisers and ad platforms: The scale — tens of billions of daily bid requests — inflates impression counts and can distort programmatic bidding and attribution. HUMAN's finding that some traffic was routed through residential proxies to evade anti-abuse filters suggests buyers and platforms will need to monitor for SSAI anomalies and proxy-origin signals specifically tied to CTV inventory.
- Device manufacturers and supply-chain managers: The researchers' link between BADBOX and a named Chinese electronics manufacturer, plus the discovery of factory-installed launchers and device-specific checks, highlights risks in preinstalled software and vendor-controlled app ecosystems. Those responsible for device provisioning and factory images will face pressure to audit images and hosting repositories associated with their hardware.
- End users and device purchasers: HUMAN recommends buying certified Android devices from reputable manufacturers and ensuring Play Protect is available and active. Because BADBOX has infected inexpensive phones, tablets, and streaming boxes, consumers of low-cost AOSP devices are explicitly identified as the population most at risk in the researchers' account.
HUMAN's Satori team says it uncovered and disrupted the latest BADBOX variant and that its systems filter associated fraudulent advertising traffic, but it also warns that the botnet operators are likely to adapt to the disruptions, as they have in the past. The record-setting 40 billion daily bid request peak — plus the reuse of proxy networks and mirrored hosting — leaves a clear operational challenge for ad platforms, device makers, and defenders: takedowns and filtering can blunt an operation, but BADBOX's design and supply-chain signals make persistent reinvention possible.




