An analysis of WeChat's network protocol MMTLS finds that it is a modified version of TLS 1.3 and WeChat developers' changes to its cryptography add weaknesses
Key contributions — We performed the first public analysis of the security and privacy properties of MMTLS … X: @citizenlab , @thegrugq , @jsrailton , and @citizenlab X: @citizenlab : WeChat messages are encrypted twice, yet, double encryption doesn't make it more secure. In #WeChat, messages are first encrypted with an old and vulnerable custom protocol called “Business-layer Encryption”, then encrypted again with MMTLS, which contains minor weaknesses. [image] Thaddeus E. Grugq / @thegrugq : It's weird but... I believe the only people who will benefit from cryptologic attacks on WeChat are NSA. - China has direct access. - China uses it for managing intelligence assets. If NSA can monitor it via cryptologic attack; then —> good for USA bad for China. ¯\_(ツ)_/¯ John Scott-Railton / @jsrailton : New work by my @citizenlab colleagues digging into Chinese chat app @Weixin_WeChat's encryption. The big takeaway: they rolled their own crypto. Which you should never do. Especially not if you have a billion users. By Mona Wang, Pellaeon Lin & Jeffrey Knockel👇 [image] @citizenlab : 💬NEW REPORT: The Citizen Lab takes a deep dive into the network encryption protocol used by #WeChat, an app with over one billion users. The app uses a custom #encryption protocol called “MMTLS” that introduces cryptographic weaknesses. Read the report: https://citizenlab.ca/...
Context & Ripple Effects
This extends a run of research describing privacy and security exposure around WeChat’s surrounding systems: reverse engineering previously found behavioral tracking in Mini Programs, while a separate Citizen Lab analysis identified encryption flaws in Tencent’s Sogou Input Method. The MMTLS findings move scrutiny to the transport layer used by the main app.
Earlier reporting also documented automated image censorship in private chats, underscoring that confidentiality, content controls, and telemetry are distinct properties. Encryption alone does not resolve the broader privacy questions raised by WeChat’s architecture.
First-order effects
- WeChat and its developers face a concrete protocol-hardening task: assess the reported weaknesses in their TLS 1.3 modifications and the older business-layer encryption, then update affected clients and services where warranted.
- Users and organizations relying on WeChat’s encryption claims gain a more qualified security picture: two encryption layers do not automatically provide stronger protection when one is legacy custom cryptography and the other departs from a standard protocol.
Second-order effects
- The findings give security reviewers and enterprise users a reason to scrutinize WeChat’s implementation choices rather than treating the use of TLS-derived encryption as sufficient evidence of robust protection.
- They also reinforce pressure on messaging-platform developers to minimize custom cryptographic changes, because deviations add audit and maintenance burden without necessarily improving security.
Third-order effects
- If repeated across major platforms, this pattern favors widely reviewed protocols and independently assessable implementations over proprietary security layers whose properties are difficult for users to verify.
- The broader consequence is a more granular standard for communication privacy: transport encryption, application-layer encryption, tracking, and content governance will increasingly be evaluated separately rather than collapsed into a single claim of “secure messaging.”
The trend: Messaging security is shifting from judging whether an app encrypts traffic to auditing how its full stack implements encryption, data collection, and platform controls.