Threat reportMalwareTL-2026-2312
StreamRat Android Banking Trojan Spreads via Fake Streaming-Service Ads on Meta and TikTok
StreamRat, a newly discovered Android banking Trojan and infostealer built on a Malware-as-a-Service (MaaS) backend, is being distributed through fake free-TV-streaming-service ads on Meta (Facebook/Instagram) and TikTok, primarily targeting Spanish-speaking users in Spain. A Meta campaign named "Steamtv Esp." ran June 11 to July 3, 2026 and reached roughly 570,000-570,950 users; once sideloaded, StreamRat abuses Accessibility Services, MediaProjection, and Device Administrator privileges to deliver credential-stealing overlays, keylogging, HVNC/VNC-style screen streaming, and full remote device control.
Key facts
| Identifiers | TL-2026-2312 |
|---|---|
| Also known as | Steamtv Esp. |
| Severity | HIGH |
| Affected products | Google Android |
| Malware | Godfather, Mirax, StreamRat |
| Tools | Android/PUP.Agent.ACR02DB0614H7 |
| MITRE ATT&CK techniques | T1407 Download New Code at Runtime, T1417 Input Capture, T1417.001 Keylogging, T1417.002 GUI Input Capture, T1424 Process Discovery, T1437 Application Layer Protocol, T1437.001 Web Protocols, T1456 Drive-By Compromise, T1513 Screen Capture, T1516 Input Injection, T1544 Ingress Tool Transfer, T1626 Abuse Elevation Control Mechanism and 4 more in the attack flow |
| Motivation | FINANCIAL |
| Target sectors | financial services, consumer retail banking customers |
| Target regions | spain, European Union |
| Detection rules | 9 (1 shown free) |
| Sources | 6 cited sources |
Detection coverage
As of 2026-09-03, Threadlinqs Intelligence publishes 9 detection rule(s) for TL-2026-2312 across Splunk SPL, Microsoft KQL and Sigma, covering 18 indicator(s) of compromise. The whole corpus is readable without an account. Every threat comes with one complete detection in Splunk SPL, Microsoft KQL and Sigma, free for everyone; a free account adds daily TL Creds to unlock the rest. Paid plans add a larger daily allowance, correlation and the MCP server. Threadlinqs MCP server · View plans.
Free detection: StreamRat WebSocket Binary-RPC C2 Beacon and Post-Compromise Capability Abuse
Detects outbound WebSocket-upgrade connections from Android clients to StreamRat's hardcoded C2 servers (45.147.28.59, 193.32.2.245) carrying the custom X-Device-Id/X-Device-Model/X-Api-Level bot-registration headers, or periodic low-and-slow beaconing consistent with the maintenance_check/injection_check foreground-app polling loop (Process Discovery) that drives HTML overlay delivery for credential theft (GUI Input Capture), Accessibility-based keylogging, HVNC/VNC-mode screen capture, and remote input injection/device unlock. Every StreamRat post-compromise capability (37+ bot commands) rides this single C2 channel.
- Languages: Splunk SPL, Microsoft KQL, Sigma
- ATT&CK: T1437, T1437.001, T1513, T1417, T1417.001, T1417.002, T1424, T1516
- Data sources: firewall / next-gen firewall traffic logs (Palo Alto, generic), network stream/proxy logs (stream:http), Microsoft Defender for Endpoint DeviceNetworkEvents
- Severity: critical
- Confidence: high
Splunk SPL
index=network (sourcetype=pan:traffic OR sourcetype=stream:http OR sourcetype=firewall)
dest_ip IN ("45.147.28.59","193.32.2.245")
| eval is_websocket=if(match(http_header,"(?i)upgrade:\s*websocket") OR dest_port IN (80,443,8080), 1, 0)
| eval has_device_headers=if(match(http_header,"(?i)x-device-id") OR match(http_header,"(?i)x-device-model") OR match(http_header,"(?i)x-api-level"), 1, 0)
| bin _time span=5m
| stats count as conn_count, values(dest_port) as ports, max(has_device_headers) as device_headers_seen, max(is_websocket) as websocket_seen by src_ip, dest_ip, _time
| eventstats count as beacon_windows by src_ip, dest_ip
| where beacon_windows >= 3 OR device_headers_seen=1
| sort - beacon_windowsMicrosoft KQL
let C2Ips = dynamic(["45.147.28.59","193.32.2.245"]);
DeviceNetworkEvents
| where Timestamp > ago(24h)
| where RemoteIP in (C2Ips)
| summarize BeaconCount = count(), Ports = make_set(RemotePort), FirstSeen = min(Timestamp), LastSeen = max(Timestamp) by DeviceId, DeviceName, RemoteIP, InitiatingProcessFileName
| extend AvgIntervalMin = datetime_diff('minute', LastSeen, FirstSeen) / iif(BeaconCount > 1, BeaconCount - 1, 1)
| where BeaconCount >= 3
| order by BeaconCount descSigma
title: StreamRat C2 Beacon to Hardcoded IPs over WebSocket
id: 5f2a9c6e-4b1d-4e7a-9c3f-8a1b2c3d4e5f
status: experimental
description: Detects network connections to StreamRat's hardcoded C2 infrastructure (45.147.28.59, 193.32.2.245) established via a WebSocket upgrade carrying the bot's custom device-fingerprint headers, or repeated low-and-slow polling consistent with the maintenance_check/injection_check overlay-delivery loop that backs screen capture, keylogging, and input injection.
references:
- https://www.threatfabric.com/blogs/from-meta-ads-to-full-device-takeover-uncovering-streamrat
author: Threadlinqs Team
date: 2026-09-03
tags:
- attack.command_and_control
- attack.t1437
- attack.t1437.001
- attack.discovery
- attack.t1424
- attack.collection
- attack.t1513
- attack.credential_access
- attack.t1417
- attack.t1417.001
- attack.t1417.002
- attack.impact
- attack.t1516
detection:
selection_ip:
dst_ip:
- '45.147.28.59'
- '193.32.2.245'
selection_headers:
http_header|contains:
- 'X-Device-Id'
- 'X-Device-Model'
- 'X-Api-Level'
condition: selection_ip or selection_headers
logsource:
category: firewall
product: generic
fields:
- src_ip
- dst_ip
- dst_port
- http_header
falsepositives:
- None expected; these IPs are exclusively reported as StreamRat C2 infrastructure at time of writing
level: criticalKnown false positives:
- None expected for the hardcoded-IP match; the beaconing-interval heuristic alone can false-positive on legitimate polling apps (push-notification services, VoIP keep-alives) — require the IP match or device-fingerprint headers before high-confidence alerting
Tuning notes: Retire or re-verify the hardcoded C2 IPs periodically — StreamRat's MaaS backend can be reconfigured with new C2 infrastructure per customer/build; pivot detection to the X-Device-Id/X-Device-Model/X-Api-Level header pattern and the 2-byte-opcode/4-byte-command-ID WebSocket framing if the IPs rotate.
- StreamRat Malvertising Lure — Referrer-Fingerprinted Dropper Delivery ([indicator] / [indicator]) · unlock with TL Creds
- StreamRat Escalating Permission-Grant Chain — Default Launcher + Device Admin + VPN Block + Accessibility · unlock with TL Creds
- StreamRat WebSocket Binary-RPC C2 Beacon and Post-Compromise Capability Abuse · unlock with TL Creds
- StreamRat Malvertising Lure — Referrer-Fingerprinted Dropper Delivery ([indicator] / [indicator]) · unlock with TL Creds
- StreamRat Escalating Permission-Grant Chain — Default Launcher + Device Admin + VPN Block + Accessibility · unlock with TL Creds
- StreamRat WebSocket Binary-RPC C2 Beacon and Post-Compromise Capability Abuse · unlock with TL Creds
- StreamRat Malvertising Lure — Referrer-Fingerprinted Dropper Delivery ([indicator] / [indicator]) · unlock with TL Creds
- StreamRat Escalating Permission-Grant Chain — Default Launcher + Device Admin + VPN Block + Accessibility · unlock with TL Creds
Indicators of compromise
As of 2026-09-03, Threadlinqs Intelligence tracks 18 indicators of compromise for TL-2026-2312.
Sample indicators (3 of 18)
- IP address:
45.147.28[.]59 - File hash:
e0714788b4e2518b0d9d4cbf18c7217bb97718e01689d77338f1cc4a230fcb6c - IP address:
193.32.2[.]245
Defanged, TLP:CLEAR. Full set of 18 indicators: unlock with TL Creds.
Technical analysis
ThreatFabric researchers uncovered StreamRat, an Android banking Trojan/infostealer distributed through malvertising on Meta and TikTok impersonating a free TV-streaming service. The Meta campaign, internally named "Steamtv Esp.," ran from June 11 to July 3, 2026 and reached an estimated 570,000-570,950 accounts in the EU, overwhelmingly Spanish-speaking users in Spain; identical banner creative was also run on TikTok.
Victims who click the ad are sent to a lure site that fingerprints the visitor's OS (blocking non-Android devices) and referral source (Instagram, TikTok, Facebook, or generic browser), then serves referrer-tailored install instructions via a page named r1edmi.html. The victim is guided to enable installation from unknown sources and download an initial APK (app.apk). The APK is a dropper built from a set of interconnected HTML/JavaScript pages (Index.html, Set_launcher.html, Vpn_required.html) rather than compiled native obfuscation. The dropper coerces the victim into three escalating grants: (1) setting the dropper as the default/HOME launcher, so pressing the Home button re-displays the dropper's social-engineering flow; (2) establishing a Device VPN permission used to create a deliberately non-functional VPN that blocks Internet access for every other app on the device (excluding the dropper itself), degrading other security tools' ability to perform cloud/online-reputation checks during installation; and (3) downloading and installing a second-stage payload (named update_{timestamp}.apk) that requests Accessibility Services. Once Accessibility access is granted, the dropper uninstalls/removes itself, and the final payload becomes the persistent backdoor.
Two payload samples were identified: package io.base.one887 (app label "StrεαmTV Pro") and package io.meat.hint (app label "Sistema de vídeo"). Malwarebytes detects the family generically as Android/PUP.Agent.ACR02DB0614H7.
Once active, the bot communicates with hardcoded C2 servers (45.147.28.59 and 193.32.2.245) over a WebSocket-based, binary RPC-like protocol: the initial upgrade request carries custom X-Device-Id / X-Device-Model / X-Api-Level headers, and subsequent frames use a 2-byte opcode + 4-byte command ID + payload format (bot responses mirror with a 4-byte command ID + 1-byte status + 4-byte length + payload). The bot supports 37+ commands spanning shell execution, device lock/wake, VNC-mode and HVNC-mode screen capture, an Accessibility Node Viewer for remote UI inspection/control, overlay management, arbitrary app launching, PIN/pattern extraction and remote device unlock, Accessibility-based keylogging, and fake notification generation.
Credential theft is delivered through a foreground-app-monitoring loop: the bot sends maintenance_check and injection_check requests containing the current foreground package name, and the C2 responds with an HTML overlay (saved client-side as <package_name>.html) rendered in a WebView with a JavaScript bridge enabled. Overlay types include a black screen covering ~98% of the display to mask fraudulent activity from the victim, a fake system-update screen used as a distraction cover, and interactive injection overlays used to harvest typed data. To reduce backend load and limit detectable signal, the malware computes Adler-32 checksums over serialized AccessibilityNodeInfo UI-tree dumps and only reports changed states.
Control-panel code analysis indicates StreamRat was built as a Malware-as-a-Service offering: the backend implements user/supervisor/admin role separation for multiple simultaneous operators and includes a dedicated builder for customers to generate customized droppers and payloads. ThreatFabric ties the same threat actor to at least two prior Android financial-malware families -- GodFather and StreamRat's most direct predecessor, Mirax (an April 2026 campaign). The StreamRat payload was retrieved from a GitHub repository belonging to the same developer account previously used to distribute Mirax, with daily package/release updates, and the StreamRat dropper's design closely mirrors Mirax's dropper architecture, indicating iterative reuse rather than a clean-sheet build. ThreatFabric did not attach a named threat-actor label to the operator.
MITRE ATT&CK techniques by tactic:
Defense Evasion
T1407 Download New Code at Runtime; T1629 Impair Defenses; T1630 Indicator Removal on Host; T1655 Masquerading
Credential Access
T1417 Input Capture; T1417.001 Keylogging; T1417.002 GUI Input Capture
Discovery
Command and Control
T1437 Application Layer Protocol; T1437.001 Web Protocols; T1544 Ingress Tool Transfer
Initial Access
Collection
Impact
Privilege Escalation
T1626 Abuse Elevation Control Mechanism; T1626.001 Device Administrator Permissions
Remediation
Immediate actions
- Block outbound traffic to StreamRat C2 IPs 45.147.28.59 and 193.32.2.245 at perimeter/DNS/firewall layers.
- For any device suspected of installing io.base.one887 ("StrεαmTV Pro") or io.meat.hint ("Sistema de vídeo"), disconnect it from Wi-Fi/mobile data immediately to cut C2 access.
- Uninstall the malicious app; if it was set as the default launcher, reset the default HOME app via Android Settings > Apps > Default apps before/while removing it.
- Revoke Accessibility Services, Device Administrator, and Default App permissions granted to the malicious package via Settings before uninstalling.
- Change passwords for any banking, email, and social-media accounts accessed on the compromised device, and contact affected financial institutions to flag potential account takeover/fraud.
Workarounds
- Only install Android apps from the Google Play Store; avoid APKs offered via advertisements, direct-download links, or messages from strangers.
- Before installing, verify the developer's identity, review history, and install counts; be suspicious of new/low-review-count "free streaming" apps.
- Treat any app requesting Accessibility Services, Device Administrator, MediaProjection (screen sharing), or default-launcher/HOME-app status without a clear functional need as a red flag and decline the request.
- Be suspicious of any installer flow that instructs disabling security warnings or granting a VPN permission as a precondition for using an unrelated streaming app.
Longer-term hardening
- Deploy a mobile threat defense (MTD) or EDR solution capable of behavioral detection of Accessibility-Service and MediaProjection abuse, not just signature matching.
- Enable and enforce Google Play Protect scanning and block sideloading (installation from unknown sources) via MDM/EMM policy on managed devices.
- Use DevicePolicyManager.setPermittedAccessibilityServices (or equivalent EMM control) to restrict which apps may register as Accessibility Services on managed fleets.
- Monitor ad-platform brand/keyword abuse (fake streaming-service creative) and report malicious Meta/TikTok ad campaigns for takedown.
Affected and fixed versions
- Google — Android
Vulnerable versions: Any Android version exposing the Accessibility Service, MediaProjection, Device Administrator, and default-launcher (HOME role) APIs abused by StreamRat
Timeline
- ThreatFabric-tracked Mirax Android banking-trojan campaign runs (April 2026); its dropper architecture and GitHub release infrastructure closely resemble and are directly reused by StreamRat's dropper.
- Meta advertising campaign "Steamtv Esp." begins, running fake free TV-streaming-service ads on Facebook and Instagram targeting Spanish-speaking users, primarily in Spain; identical banner creative also appears on TikTok.
- The "Steamtv Esp." Meta ad campaign concludes, having reached an estimated 570,000-570,950 Meta accounts across the EU.
- ThreatFabric researchers identify and begin technical analysis of the StreamRat campaign, malware, and MaaS backend (reported as occurring in late July 2026).
- The Hacker News, GBHackers, and CyberSecurityNews publish coverage summarizing ThreatFabric's StreamRat findings and IOCs for a broader security audience.
- ThreatFabric publishes "Uncovering StreamRat: From Meta Ads to Full Device Takeover," detailing the dropper chain, WebSocket C2 protocol, bot command set, MaaS backend, and IOCs.
- Malwarebytes Labs publishes coverage of the StreamRat campaign, detecting the payload generically as Android/PUP.Agent.ACR02DB0614H7 and issuing consumer-facing remediation guidance.
Sources cited
- Uncovering StreamRat: From Meta Ads to Full Device Takeover threatfabric.com · Vendor report · 2026-09-02
- StreamRat Android malware spreads through Meta and TikTok ads malwarebytes.com · Blog · 2026-09-03
- Meta Ads Push StreamRat Android Trojan That Can Gain Near-Complete Device Control The Hacker News · News · 2026-09-02
- StreamRAT Abuses Android Accessibility, MediaProjection and HVNC gbhackers.com · News · 2026-09-02
- New StreamRAT Android Trojan Gives Hackers Full Remote Control Through VNC and Accessibility cybersecuritynews.com · News · 2026-09-02
- This devious Android malware has returned disguised as TikTok or streaming apps techradar.com · News · 2026-09-02
Previous and next in malware
Drafted by the Threadlinqs research pipeline from 6 sources; published ; last updated —.