Zero-Click Vulnerability in WeChat Calls Spreads Among Friends

Tencent has fixed a zero-click vulnerability in WeChat’s calling feature. The attack succeeds even if the user does not answer, and account takeover is completed in seconds. The research organization Calif, which discovered it, named the flaw WeWorm. This article is based on The Register (All Rights Reserved). It relies on fair quotation under Article 32 of the Japanese Copyright Act.

According to reporting by Connor Jones of The Register, there was a memory corruption flaw in the VoIP processing of WeChat, which has more than 1.4 billion monthly users. The flaw affected both iOS and Android, and was structured to be exploitable via calls from registered friends. Tencent released a fix on August 21. Calif released a proof-of-concept video this week, but has kept the core technical details private.

How the Attack Hijacks Without an Answer

In Calif’s demonstration, when the caller sends crafted call data, processing on the receiving side breaks down. Even if the user does not answer the call, control is seized while the ringtone continues to ring. After the takeover, reading and sending messages, making calls, and acting on behalf of the user become possible. The original quote is as follows.

Exploitation takes only seconds, and gives us full control of the WeChat account We can read and send messages, make calls, and act on the victim’s behalf

It is said that rejecting the call avoided infection. In contrast, answering the call or letting it keep ringing did not prevent infection. The fact that an attacker can retry by calling back when the victim is absent also makes defense difficult. Because voice-call reception processing is always on standby, the attack surface tends to be broad. There are also many cases in the video encoding field where memory safety becomes an issue. As shown in Vulkan Video Encoding Returns on Intel Alchemist GPUs, media-processing implementations are becoming increasingly complex. Decoding call audio and jitter countermeasures likewise involve complex parsing.

Method of Spread Abusing Trusted Relationships

WeWorm spreads via friend relationships. An attacker needs to be registered as a friend of the victim, but chains from a hijacked account to the next friend. For this reason, after the initial intrusion, it can spread as a trusted caller. Even without any operation by the victim, the next call is automatically placed from the hijacked side.

This structure has properties similar to self-propagating email malware. The difference is that it starts with telephone incoming-call processing. Users can become victims without opening suspicious attachments. Trust in contacts alone cannot provide defense. In corporate environments, personal and business contacts tend to mix. A single compromise could spread to business partners and colleagues. Administrators are required to enforce thorough device management and app updates.

Discovery and Proof-of-Concept Process

Shortened by AI Use

Calif explains that it used AI to identify the flaw and achieve an initial proof of remote execution in about two days. The target was VoIP reception processing, apparently focusing on insufficient input validation. Tencent verified the content after the report. The detailed exploit chain has not been disclosed.

Faster work through AI also affects defenders. Methods that combine AI with fuzzing, reachability analysis, and generation of proof-of-concept code templates are spreading. As shown in Desktop Commander MCP Delegates Terminal Operations to AI, operations that delegate work to AI are expanding at development sites. Similarly in vulnerability research, automation of reproduction-environment setup and documentation is advancing. However, manual verification remains essential for final root-cause identification and impact assessment. This case showed that faster discovery directly leads to faster disclosure and patching.

Risk of Chained Compromise Affecting the

Entire Device

Calif says that WeWorm alone is limited to controlling the WeChat account. However, it explains that combined with other Android and iOS flaws, it could lead to control of the entire device. The specific chaining steps have not been disclosed. It also mentions a scenario in which privilege escalation like the OEMpocalypse technique is combined to seize WeChat from a flaw in another app.

The dangers of privilege escalation and inter-app cooperation have been confirmed in other cases. In Microsoft Defender Privilege Escalation Vulnerability ‘RoguePlanet’ Disclosed, a privilege issue in the defense mechanism itself was reported. Even when a single flaw appears to have a small impact, cases reaching administrator privileges through chaining are not uncommon. In this case as well, a pattern is assumed in which a flaw in call processing serves as the starting point and combines with flaws in the OS layer. Therefore, both OS updates and app updates are important. For users, enabling automatic updates and verifying the legitimacy of the source are basic countermeasures.

The fix was already released on August 21. Users need to update WeChat to the latest version and keep the OS up to date as well. Even with suspicious missed calls, it is hard to reject calls from friends. Although rejecting is said to have been effective, constant vigilance is not realistic. Organizations that use WeChat for business communications should track update status and consider using multi-factor authentication together. It is necessary to avoid obtaining it from sources other than official stores.

Editorial Opinion

In the short term, we expect inspections of large-scale apps with calling features to advance. Input validation for audio processing will be reviewed. Work to verify patch application will also increase.

In the long term, we assess that AI-assisted vulnerability research will become established. The time from discovery to proof-of-concept will shorten. Defenders will need to review operations and disclosure procedures.

The remaining question is how to design trust in friend relationships. Are countermeasures that rely solely on rejecting calls sufficient? How to ensure verification means for users is the question.

References

• “WeChat worm could pwn a friend before they even answered the call”, by Connor Jones — The Register, 2026-09-09T12:45:00.000Z (ARR)
• Source URL: https://www.theregister.com/security/2026/09/09/wechat-worm-could-pwn-a-friend-before-they-even-answered-the-call/5295234