Alipay DeepLink + JSBridge Security Research

17 Verified Vulnerabilities | 3 Devices | 308 Server Log Entries

Full Report

Summary

This repository documents a comprehensive security research project that uncovered 17 security vulnerabilities in Alipay's DeepLink URI scheme (alipays://) and its Nebula WebView container.

Key Findings

Severity Count Examples
CRITICAL 3 GPS silent theft, Transfer pre-fill, Payment initiation
HIGH 6 Device fingerprinting, UI spoofing, Session leak
MEDIUM 8 Network info, Chain WebView, Scheme injection

Attack Chain

External SMS/QQ/WeChat Link
    → Browser opens alipays:// DeepLink
    → Alipay launches with attacker's URL in WebView
    → AlipayJSBridge APIs exposed to external page
    → Silent data collection (GPS, device info, session)
    → UI spoofing (title bar, toast notifications)
    → Sensitive page navigation (transaction history, transfer)

Cross-Platform Verification

  • Samsung Galaxy S25 Ultra (Android 15, New Zealand)
  • Redmi 12 (Android 14, Malaysia)
  • iPhone 16 Pro (iOS 18.3, China)

Live PoC (Read-Only Demo)

No data is collected or transmitted. All results display locally only.

Responsible Disclosure Timeline

Date Action
2026-02-25 Initial report sent to Ant Group SRC (TLS/SSL findings)
2026-03-07 Full report V3 sent with 17 vulnerabilities + 308 log entries
2026-03-10 Ant Group response: "These are normal features" (正常功能)
2026-03-11 Public disclosure after vendor declined to acknowledge

Repository Structure

├── index.html          # Full bilingual (CN/EN) research blog
├── poc/
│   ├── trigger.html    # Attack trigger simulation page
│   ├── verify.html     # JSBridge exploitation PoC
│   └── chain.html      # Chain WebView demonstration
├── review_kimi.md      # Kimi K2 cross-validation review
├── review_sonnet.md    # Sonnet review
├── review_summary.md   # Review summary
└── README.md           # This file

Evidence

  • 308 server exfiltration log entries (JSONL format, not included in public repo)
  • 42 real-device screenshots (not included in public repo)
  • Full evidence available upon request: feng@innora.ai

This research is conducted for educational and security improvement purposes only. All testing was performed on accounts owned by the researcher. No unauthorized access to third-party accounts or data occurred.

The PoC pages are read-only demonstrations with all data exfiltration endpoints disabled. They only display results locally in the browser.

Mirrors & Archives

To prevent single-point deletion, this research is archived at multiple locations:

If any mirror is taken down, please check the other locations.

Readers are encouraged to fork this repository as backup.

Contact


This research follows responsible disclosure practices. The vendor was given adequate time to respond before public disclosure.

Description
No description provided
Readme 3.7 MiB
Languages
HTML 100%