# AI-Assisted Phishing Campaign Exploits Browser Permissions via edgeone.app Infrastructure to Capture Biometric Data

> Cyble CRIL identified a widespread social engineering campaign hosted on Tencent EdgeOne Pages (edgeone.app) infrastructure. The campaign deploys fake verification pages impersonating TikTok, Telegram, Instagram, Chrome, and Flappy Bird to trick victims into granting browser permissions for camera, microphone, contacts, and geolocation. Captured multimedia and biometric data is exfiltrated via Telegram Bot API, enabling identity theft, deepfake attacks, and extortion.

- **Published:** 2026-03-16T12:00:00Z
- **Last reviewed:** 2026-03-16T12:00:00Z
- **Canonical:** https://intel.threadlinqs.com/threat/TL-2026-0233
- **ID:** TL-2026-0233
- **Severity:** HIGH
- **Category:** PHISHING
- **Status:** ACTIVE
- **Detections:** 9 · **IOCs:** 20 (full data via the Threadlinqs MCP server — Purple tier)

## Description

A highly active, AI-assisted phishing campaign has been identified by Cyble Research and Intelligence Labs (CRIL) targeting users worldwide through social engineering lures hosted on Tencent EdgeOne Pages (edgeone.app) infrastructure. The campaign represents a significant evolution in browser-based phishing, moving beyond traditional credential theft to exploit HTML5 browser permission APIs for real-time biometric data capture.

The attack chain begins with victims being directed to convincing fake verification pages that impersonate popular platforms including TikTok (fake ID scanner), Telegram (fake account freeze warnings), Instagram (fake verification), Google Chrome/Drive (fake file access), and Flappy Bird (fake gaming). These lure pages are professionally designed and show indicators of AI-assisted development in their code structure and multilingual capabilities.

Upon visiting a lure page, the malicious JavaScript requests browser permissions through legitimate HTML5 APIs. The getUserMedia API is invoked to access the victim's camera and microphone, the Contacts Picker API to harvest contact lists, and the Geolocation API to pinpoint the victim's physical location. Device fingerprinting is performed through navigator properties to collect detailed system information.

The camera capture mechanism is particularly sophisticated. JavaScript accesses the device camera via navigator.mediaDevices.getUserMedia(), renders video frames to an HTML5 canvas element using ctx.drawImage(), converts frames to JPEG blobs via canvas.toBlob(), and transmits the captured images to attacker-controlled infrastructure using the Telegram Bot API's sendPhoto endpoint. Audio is captured using the MediaRecorder API and similarly exfiltrated.

All captured data — live camera images, microphone audio, contact lists, geolocation coordinates, device metadata, and public IP addresses — is exfiltrated through the Telegram Bot API. The attackers use sendPhoto for image data, sendDocument for audio recordings and contact lists, and sendMessage for text-based telemetry including IP addresses (resolved via api.ipify.org) and geolocation enrichment (via ipapi.co).

The infrastructure leverages Tencent EdgeOne Pages (edgeone.app), a legitimate CDN and hosting platform, allowing the phishing pages to benefit from fast global delivery, valid SSL certificates issued by DigiCert, and the reputation of a major cloud provider. The edgeone.app domain resolves to IP 43.159.77.156, hosted by ACE in Los Angeles, with DNS handled by Tencent's name servers (ns-open1.qq.com through ns-open3.qq.com). Multiple security vendors including G-Data, ADMINUSLabs, and CyRadar have blacklisted the domain for phishing and malicious activity.

The captured biometric data poses severe downstream risks: live camera footage and audio can be used for deepfake generation with as little as 3 seconds of audio achieving 85% voice clone accuracy. Contact lists enable secondary targeting campaigns, and combined data packages support synthetic identity fraud, extortion, and account takeover attacks.

This campaign is notable for its AI-assisted development patterns, its abuse of legitimate browser APIs for malicious data collection, its use of reputable infrastructure to evade detection, and its focus on biometric data capture rather than traditional credential phishing.

## MITRE ATT&CK

- T1566 Phishing
- T1059 Command and Scripting Interpreter
- T1204 User Execution
- T1125 Video Capture
- T1123 Audio Capture
- T1005 Data from Local System
- T1056 Input Capture
- T1082 System Information Discovery
- T1614 System Location Discovery
- T1016 System Network Configuration Discovery
- T1071 Application Layer Protocol
- T1102 Web Service
- T1567 Exfiltration Over Web Service
- T1583 Acquire Infrastructure
- T1608 Stage Capabilities
- T1589 Gather Victim Identity Information
- T1036 Masquerading
- T1531 Account Access Removal

## Sources

- [Cyble CRIL: AI-Assisted Phishing Campaign Exploits Browser Permissions to Capture Victim Data](https://cyble.com/blog/ai-assisted-phishing-campaign/)
- [Gridinsoft: edgeone.app Security Analysis — Trust Score 6/100](https://gridinsoft.com/online-virus-scanner/url/edgeone-app)
- [Tencent EdgeOne Resource Abuse Prevention Documentation](https://www.tencentcloud.com/document/product/1145/63522)
- [Tencent Cloud Illegal Content Complaint Portal](https://www.tencentcloud.com/report-platform/illegalcontentcomplaint)
- [ScamAdviser: edgeone.app Subdomain Analysis](https://www.scamadviser.com/check-website/12345test-mt4c6mfuxg.edgeone.app)
- [Security Boulevard: Weaponizing Telegram Bots for Credential Exfiltration](https://securityboulevard.com/2026/03/weaponizing-telegram-bots-how-threat-actors-exfiltrate-credentials/)
- [MDN: getUserMedia API Documentation — Browser Permission Model](https://developer.mozilla.org/en-US/docs/Web/API/MediaDevices/getUserMedia)
- [CYFIRMA: Raven Stealer — Telegram-Based Data Exfiltration Analysis](https://www.cyfirma.com/research/raven-stealer-unmasked-telegram-based-data-exfiltration/)
- [Cyble: Deepfake-as-a-Service Threat Landscape 2025-2026](https://cyble.com/knowledge-hub/deepfake-as-a-service-exploded-in-2025/)
- [Microsoft Security Blog: AI as Tradecraft — How Threat Actors Operationalize AI](https://www.microsoft.com/en-us/security/blog/2026/03/06/ai-as-tradecraft-how-threat-actors-operationalize-ai/)

## Full data

Detection queries (Splunk SPL / Microsoft KQL / Sigma) and IOC values require the Threadlinqs MCP server (Purple tier): https://intel.threadlinqs.com/mcp

Canonical: https://intel.threadlinqs.com/threat/TL-2026-0233