#introduction I'm Jiqiang Feng, independent security researcher at Innora AI. I found 17 vulnerabilities (CVSS up to 9.3) in Alipay, a payment app used by 1B+ people. 18 CVEs filed with MITRE. Peer-reviewed paper published by IACR.
My Twitter/X account was permanently suspended during this disclosure. 8 research articles were also deleted from WeChat by the vendor's lawyers.
innora.ai | github.com/sgInnora
THREAD: Alipay SecurityGuard SDK — What we found and what happened next.
1/7 We reverse-engineered Alipay's SecurityGuard SDK (v10.8.30.8000, 89K Java source files). Found 17 vulnerabilities including a whitelist bypass (CVSS 9.3) that makes all 17 remotely exploitable via a single crafted URL.
18 CVEs filed across 4 MITRE tickets. Vendor says: 'normal functionality.'
2/7 Key findings:
- 976 proxy classes intercepting 208 system API categories (GPS, camera, clipboard, crypto)
- 97.1% of internal APIs (396/408) have ZERO access control
- PatchProxy: every security method remotely replaceable without app update
- SM4 encryption remotely disableable by server config
Full analysis: github.com/sgInnora/alipay-securityguard-analysis
3/7 The cryptographic infrastructure is broken:
- APK signing cert uses MD5+RSA-1024 (collision in 9 seconds)
- 27 server RSA private keys recovered via batch GCD
- Hardcoded DES keys
11 verified PoCs: github.com/sgInnora/hash-collision-lab
IACR paper: eprint.iacr.org/2026/526
4/7 Then came the censorship.
March 15: 4 research articles deleted from WeChat after Ant Group's law firm filed takedown requests.
WeChat initially REJECTED the complaint. It was resubmitted under China's Cybersecurity Law — articles removed without specific provision cited.
March 20: 4 MORE articles deleted. 8/8 = 100% censored.
5/7 Cross-platform suppression:
WeChat: 8 articles deleted (March 15-20)
Twitter/X: Account permanently suspended (March 16-17)
Meanwhile, the research was independently validated by IACR, MITRE (18 CVEs), Packet Storm (#217089), and acknowledged by 12+ regulatory agencies worldwide.
Full timeline: innora.ai/zfb/article_censorship.html
6/7 Regulatory responses (12+ jurisdictions):
- CSSF Luxembourg: CSSFWB-2026-080
- CNPD Luxembourg: GDPR investigation
- HKMA Hong Kong: CE20260313175412
- PDPC Singapore: #00629724
- BSP Philippines, PCPD HK, BNM Malaysia
- Google Play, CISA/CERT
- MITRE: 18 CVEs across 4 tickets
7/7 All evidence permanently preserved on IPFS:
gateway.pinata.cloud/ipfs/QmWUnbmgHsb3BMLufJWhzVaaZqd8j7XMjN2YVUmAGRGJ4C
Please fork github.com/sgInnora/alipay-securityguard-analysis as backup against further takedowns.
If you've experienced similar vendor retaliation for security research, I'd like to hear from you.
When the signature verifier is itself remotely replaceable...
PatchProxy controls 146,173 methods in Alipay, including verifyApk() — the trust anchor recursively under attacker control. Turtles all the way down.
Batch-3 filed (10 new, 28 total):
- PatchProxy RCE: CVSS 9.8
- Lua VM RCE: CVSS 9.8
- Payment auth bypass: CVSS 9.1
Details: innora.ai/zfb/
Jiqiang Feng | Innora AI Security