Back to Intelligence

BADBOX Botnet Returns: 40 Billion Daily Ad Requests — Detection and Containment Guide for Android TV Box Infections

SA
Security Arsenal Team
November 22, 2026
12 min read

Introduction

The BADBOX botnet — the Android-based malware operation first disrupted in late 2024 and again in 2025 — has returned at a scale that should concern every network defender. According to reporting covered by BleepingComputer, the operators behind BADBOX have rebuilt their infrastructure and are now generating a record-breaking 40 billion ad requests per day through a massive ad-fraud operation running on compromised generic Android TV boxes and other low-cost Android devices.

This matters far beyond the ad-tech industry. BADBOX is not just an ad-fraud nuisance — infected devices ship (or are backdoored post-sale) with a persistent implant that gives operators remote control of the device: residential proxy services, account takeover tooling, credential theft, and a foothold inside your network perimeter. If your organization allows staff to connect cheap Android TV boxes, streaming sticks, digital signage players, or white-label tablets to corporate or guest networks, you may already be hosting a node of this botnet. For home users — including your remote workforce — these devices sit on the same LAN as corporate laptops, creating lateral-movement and data-exposure risk.

This post breaks down how the revived BADBOX operation works, how to hunt for it on your network, and how to remediate and prevent reinfection.

Technical Analysis

What Is BADBOX and What's New in the Current Campaign

BADBOX is a firmware-level and supply-chain-delivered backdoor family that primarily affects generic, no-name Android TV boxes and streaming devices — typically AOSP-based devices sold under dozens of rotating brand names on Amazon, AliExpress, and similar marketplaces. The original operation was documented by HUMAN Security (Satori team) and partners, with a coordinated disruption seizing C2 domains in 2024 and a follow-on "BADBOX 2.0" takedown in mid-2025. The current reporting confirms the operators have reconstituted the botnet and scaled ad-fraud operations to roughly 40 billion fraudulent ad requests per day — the largest volume observed from this actor to date.

Affected Products and Platforms

  • Generic Android TV boxes / streaming sticks running AOSP builds (often Android 9–12), frequently with misleading "certified" branding. Devices typically ship from Chinese manufacturers with the backdoor pre-installed in the firmware image or injected via a malicious OTA/update mechanism.
  • White-label tablets, digital photo frames, car head units, and projector devices built on the same ODM platforms.
  • Devices that are not Google Play Protect certified — a key tell. The backdoored builds cannot pass Play Integrity certification.

There is no CVE associated with BADBOX because it is not a vulnerability — it is a supply-chain implant. The devices are compromised before the customer ever powers them on, or shortly after first boot via malicious app-store/update components.

How the Attack Works (Defender's View)

The infection chain observed across BADBOX variants follows a consistent pattern:

  1. Pre-installed backdoor: A core library (historically observed as com.cyber.admin, com.hs.su, or embedded in the launcher/system apps) provides persistent remote access. In earlier variants, the implant modified libandroid_runtime.so and used the mex/syliva-style loaders to fetch modules.
  2. C2 check-in: On first network connection, the device beacons to hardcoded and DGA-generated domains to register itself and pull down operational modules.
  3. Modular payload delivery: Modules delivered over the C2 channel include:
    • Ad-fraud module: Loads hidden WebViews and generates massive volumes of fake ad impressions/clicks — the engine behind the current 40-billion-request/day operation.
    • Residential proxy module: The device is enrolled into a proxy network (sold commercially), meaning third-party traffic — potentially including criminal traffic — egresses from your IP address.
    • Account/credential modules: Historically used for creating fraudulent accounts and harvesting data.
  4. Persistence: The implant survives factory resets in many cases because components live in the system partition or in a backdoored recovery/update path. Simply "resetting" the box is not reliable remediation.

Exploitation Status

  • Confirmed active, at record scale: This is not theoretical. Tens of thousands of devices (prior disruptions identified hundreds of thousands of IPs) are actively participating, and the ad-fraud volume reported — 40 billion requests/day — indicates successful re-seeding of C2 infrastructure after the 2025 takedown.
  • No CVE / no KEV entry applies — this is a supply-chain compromise, not an exploitable software flaw. Defense is detection, containment, and removal.
  • Egress risk: Because of the residential-proxy module, infected devices on your network mean arbitrary third-party traffic exits through your public IP. For enterprises, this creates abuse-desk, blocklist, and potential liability exposure.

Detection & Response

This is a technical threat, and the most reliable detections are network-centric — you cannot easily run EDR on a compromised Android TV box, but you absolutely control (or can observe) the network it talks on.

Hunting Strategy

Prioritize these signals:

  1. DNS queries to known BADBOX-associated C2 domains. Public reporting (HUMAN Security, and prior takedown notices) has associated domains including flyermobi.com, aituyou.com, logc.binicko.com, yczm.me, and related infrastructure with the operation. Historical indicators decay fast after takedowns — treat these as starting points and pivot on behavior.
  2. Ad-fraud traffic profile: sustained, high-volume HTTP/HTTPS to dozens or hundreds of distinct ad-exchange domains from a single IoT-class device, 24/7 (real users sleep; ad-fraud bots don't).
  3. Residential-proxy indicators: an IoT device making long-lived outbound connections to many distinct IPs on non-standard ports, and — critically — receiving inbound connections it then relays.
  4. Non-Play-Protect-certified Android devices on your network: fingerprint via DHCP fingerprinting, mDNS, or MDM inventory.

SIGMA Rules

YAML
---
title: BADBOX Android Botnet C2 DNS Resolution
id: 3f8c2e14-9a6b-4d71-b8e2-5c1a7f94d203
status: experimental
description: Detects DNS queries to domains publicly associated with the BADBOX Android botnet C2 infrastructure. Indicators decay after takedowns; supplement with passive DNS and threat-intel feeds for current domains.
references:
  - https://www.bleepingcomputer.com/news/security/badbox-30-botnet-returns-with-record-breaking-40-billion-daily-ad-requests/
  - https://www.humansecurity.com/learn/blog/badbox-2-0
  - https://attack.mitre.org/techniques/T1071/001/
author: Security Arsenal
date: 2026/04/06
tags:
  - attack.command_and_control
  - attack.t1071.001
logsource:
  category: dns
  product: windows
detection:
  selection:
    query|contains:
      - 'flyermobi'
      - 'aituyou'
      - 'binicko'
      - 'yczm.me'
  condition: selection
falsepositives:
  - Rare; these domains have no legitimate enterprise use
level: high
---
title: IoT Device Sustained High-Volume Outbound Ad Traffic
id: 8b4d6f21-2c7a-4e93-a1d5-9f3b8c02e714
status: experimental
description: Detects hosts initiating an abnormally high volume of distinct outbound web connections consistent with BADBOX hidden-WebView ad-fraud behavior. Correlate with IoT/device inventory to focus on Android-class endpoints.
references:
  - https://www.bleepingcomputer.com/news/security/badbox-30-botnet-returns-with-record-breaking-40-billion-daily-ad-requests/
  - https://attack.mitre.org/techniques/T1649/
author: Security Arsenal
date: 2026/04/06
tags:
  - attack.command_and_control
  - attack.t1071.001
logsource:
  category: firewall
  product: windows
detection:
  selection:
    dst_port:
      - 80
      - 443
  aggregation:
    - aggregate: count
      group-by:
        - src_ip
      condition:
        gte: 2000
        time-frame: 1h
  condition: selection and aggregation
falsepositives:
  - NAT gateways and proxy servers aggregating many users; scope the rule to IoT VLANs or exclude known proxies
level: medium
---
title: IoT Device Inbound Connection Consistent With Residential Proxy Enrollment
id: c17e9a34-5b2d-4f86-91a3-6d0c4e87b529
status: experimental
description: Detects inbound connections to IoT-segment devices, a hallmark of residential-proxy modules used by BADBOX to relay third-party traffic through infected Android TV boxes.
references:
  - https://www.bleepingcomputer.com/news/security/badbox-30-botnet-returns-with-record-breaking-40-billion-daily-ad-requests/
  - https://attack.mitre.org/techniques/T1090/
author: Security Arsenal
date: 2026/04/06
tags:
  - attack.command_and_control
  - attack.t1090
logsource:
  category: firewall
  product: windows
detection:
  selection:
    direction: inbound
    dst_ip|cidr:
      - '10.0.0.0/8'
      - '172.16.0.0/12'
      - '192.168.0.0/16'
  filter_iot_scope:
    src_ip|cidr:
      - '10.0.0.0/8'
      - '172.16.0.0/12'
      - '192.168.0.0/16'
  condition: selection and not filter_iot_scope
falsepositives:
  - Legitimate inbound services on IoT segments (rare); validate destination device identity
level: medium

Note on scope: adjust the CIDR filters to your actual IoT/guest segments before deploying, and run the aggregation rule in hunt mode for a week to baseline before alerting.

KQL — Microsoft Sentinel / Defender

This hunts DNS-based C2 check-ins and ad-fraud egress behavior using firewall/syslog ingestion (CommonSecurityLog) plus DNS analytics. If you ingest Android or proxy logs differently, adapt the table accordingly.

KQL — Microsoft Sentinel / Defender
// Hunt 1: DNS lookups to known BADBOX-associated C2 infrastructure
let BadboxDomains = dynamic(["flyermobi", "aituyou", "binicko", "yczm.me"]);
union isfuzzy=true
    (DnsEvents | where Name has_any (BadboxDomains) | project TimeGenerated, ClientIP, Name, Source="DnsEvents"),
    (CommonSecurityLog | where DeviceVendor =~ "Palo Alto Networks" and RequestURL has_any (BadboxDomains)
        | project TimeGenerated, ClientIP=SourceIP, Name=RequestURL, Source="CommonSecurityLog"),
    (Syslog | where SyslogMessage has_any (BadboxDomains)
        | project TimeGenerated, ClientIP=HostIP, Name=SyslogMessage, Source="Syslog")
| summarize Count=count(), FirstSeen=min(TimeGenerated), LastSeen=max(TimeGenerated) by ClientIP, Name, Source
| order by Count desc;

// Hunt 2: IoT-class devices generating ad-fraud-scale outbound volume
// Baseline: >1000 distinct destination FQDNs in 24h from a single internal host
CommonSecurityLog
| where TimeGenerated > ago(24h)
| where DestinationPort in (80, 443)
| where ipv4_is_private(SourceIP) and not(ipv4_is_private(DestinationIP))
| summarize DistinctDests=dcount(DestinationHostName), TotalConns=count(),
            FirstSeen=min(TimeGenerated), LastSeen=max(TimeGenerated)
  by SourceIP, DeviceAction
| where DistinctDests > 1000
| order by DistinctDests desc;

// Hunt 3: Inbound connections to internal devices (residential-proxy indicator)
CommonSecurityLog
| where TimeGenerated > ago(24h)
| where not(ipv4_is_private(SourceIP)) and ipv4_is_private(DestinationIP)
| summarize InboundConns=count(), DistinctExternalSources=dcount(SourceIP)
  by DestinationIP, DestinationPort
| where InboundConns > 50
| order by InboundConns desc;

Velociraptor VQL

You won't deploy Velociraptor to the TV box itself — but if BADBOX-adjacent tooling has touched a Windows endpoint (e.g., ADB sideloading, proxy clients, or users interacting with the box's management interface), or you want to inventory which endpoints have communicated with suspect IoT devices, hunt the managed fleet:

VQL — Velociraptor
-- Inventory endpoints with active or recent connections to hosts on IoT segments
-- and flag any endpoint running ADB (used to sideload/interact with Android boxes)
SELECT Pid, Name, CommandLine, Exe, Username, CreateTime
FROM pslist()
WHERE Name =~ '(?i)adb|scrcpy|android'
   OR CommandLine =~ '(?i)adb (connect|shell|install|push)'
VQL — Velociraptor
-- Map live network connections from managed endpoints to suspicious destinations
SELECT Pid, Name, Status, LocalAddr, LocalPort, RemoteAddr, RemotePort
FROM netstat()
WHERE Status =~ 'ESTABLISHED'
  AND NOT RemoteAddr =~ '^(10\.|172\.(1[6-9]|2[0-9]|3[01])\.|192\.168\.|127\.)'
  AND RemotePort IN (80, 443, 8080, 8888)

Remediation Script

The realistic enterprise remediation for a BADBOX-infected device is identify → isolate → remove from network → replace. The following Bash script helps network/IR teams sweep for indicators across DNS logs and block known C2 domains at the resolver/firewall level. Run it against exported DNS query logs or pipe live resolver logs to it.

Bash / Shell
#!/usr/bin/env bash
# BADBOX network sweep & containment helper
# Usage: ./badbox_sweep.sh /path/to/dns_queries.log

LOGFILE="${1:-/var/log/dns_queries.log}"
BLOCKLIST="badbox_blocklist.txt"

# Known BADBOX-associated C2 patterns (update from current threat intel)
cat > "$BLOCKLIST" <<'EOF'
flyermobi
aituyou
binicko
yczm.me
EOF

echo "[+] Scanning $LOGFILE for BADBOX C2 indicators..."
grep -iEf "$BLOCKLIST" "$LOGFILE" | awk '{print $NF, $0}' | sort | uniq -c | sort -rn > badbox_hits.txt

if [ -s badbox_hits.txt ]; then
  echo "[!] POTENTIAL BADBOX INFECTIONS FOUND — source IPs hitting C2 domains:"
  cat badbox_hits.txt
  echo
  echo "[+] Next steps:"
  echo "    1. Isolate source IPs at the switch/NAC immediately (quarantine VLAN)."
  echo "    2. Add domains to your DNS sinkhole / protective resolver blocklist."
  echo "    3. Physically remove and replace the offending device — do NOT trust factory reset."
else
  echo "[+] No matches found for current indicators. Continue behavioral monitoring (ad-fraud egress volume)."
fi

# Optional: apply sinkhole via dnsmasq (uncomment and adjust)
# while read -r d; do echo "address=/$d/0.0.0.0" >> /etc/dnsmasq.d/badbox_sinkhole.conf; done < "$BLOCKLIST"
# systemctl reload dnsmasq

echo "[+] Done."

For organizations using PowerShell on Windows-based management hosts to audit DHCP leases for suspicious Android IoT devices:

PowerShell
# Enumerate DHCP leases and flag likely generic Android TV boxes by MAC vendor / hostname patterns
$scopes = Get-DhcpServerv4Scope
$suspectPatterns = '(?i)android|tvbox|tv_box|h96|x96|t95|mxq|allwinner|amlogic|rockchip'

foreach ($scope in $scopes) {
    Get-DhcpServerv4Lease -ScopeId $scope.ScopeId |
        Where-Object { $_.HostName -match $suspectPatterns } |
        Select-Object IPAddress, HostName, ClientId, @{n='ScopeId';e={$scope.ScopeId}} |
        ForEach-Object {
            Write-Host "[SUSPECT] $($_.IPAddress) - $($_.HostName) - MAC $($_.ClientId)" -ForegroundColor Yellow
        }
}
Write-Host "Review flagged devices: verify Play Protect certification; isolate anything uncertified."

Remediation

Given this is a supply-chain implant and not a patchable CVE, remediation is operational:

  1. Quarantine immediately. Move any suspected device to an isolated VLAN with no LAN access and no (or fully logged) internet access. Do this at the switch/NAC layer — the device itself cannot be trusted.
  2. Replace the hardware. Do not rely on factory resets or firmware "cleaning." BADBOX components have demonstrated persistence in system partitions and backdoored update paths. The only reliable fix is physical removal and replacement with a certified device.
  3. Procurement controls (the real fix): Mandate Google Play Protect certified Android devices in purchasing policy. Block procurement of generic/no-name Android TV boxes, tablets, and signage players. For digital signage and kiosk use cases, use managed enterprise devices with attested firmware.
  4. Network segmentation: Place all IoT/media devices on a dedicated VLAN with: no access to internal corporate segments, egress restricted to required services only, and DNS forced through your protective resolver (block DNS bypass to 8.8.8.8/1.1.1.1).
  5. Block known C2 and monitor for re-seeding: Add current BADBOX domains to DNS sinkholes and egress blocklists, and subscribe to HUMAN Security / vendor threat-intel updates — C2 infrastructure rotates after each disruption, as this 40-billion-request comeback demonstrates.
  6. Egress abuse monitoring: If a device was enrolled in the residential proxy network, your public IP may have been used to relay third-party traffic. Check your IPs against abuse blocklists (Spamhaus, etc.) and review outbound traffic volume from the affected segment for the exposure window.
  7. Remote-workforce guidance: Employees' home networks are your extended perimeter. Publish guidance warning against cheap streaming boxes; infected devices share LANs with corporate laptops. Consider this in your home-network security awareness content and VPN/split-tunnel policy.
  8. Verify with hunting: Run the detections above across DNS and firewall logs going back at least 90 days — given the botnet's re-seeding, historical check-ins identify devices that may attempt reconnection to new C2 infrastructure.

There are no vendor patch URLs for this threat; the authoritative references are the BleepingComputer report and HUMAN Security's ongoing BADBOX research.

Related Resources

Security Arsenal Incident Response Services AlertMonitor Platform Book a SOC Assessment incident-response Intel Hub

Is your security operations ready?

Get a free SOC assessment or see how AlertMonitor cuts through alert noise with automated triage.