Tencent Meeting forensics based on memory reverse analysis

Shilong Yu, Binglong Li, Lin Zhu, Heyu Zhang, Sen Yang, Zhangxiao Li, Wenzheng FengPeerJ Comput. Sci 2025 · DOI

TL;DR

This paper introduces a novel memory forensics method for Tencent Meeting, a popular instant messaging and video conferencing tool. By analyzing memory dumps, the authors developed algorithms to extract crucial meeting metadata and communication data, addressing the challenge that this information isn't stored on disk. This work provides a vital tool for digital crime investigations involving Tencent Meeting.

What Problem Does This Paper Address?

In today's digital world, instant meeting software like Tencent Meeting, DingDing, and Zoom are indispensable for communication and collaboration. However, their widespread use also makes them targets for malicious activities, such as spreading malware, phishing, and disrupting meetings. Traditional disk forensics methods are insufficient for these applications because real-time data is often not saved to disk, and is instead held in volatile memory. This creates a significant challenge for digital forensics, as crucial evidence for legal proceedings and commercial disputes might be lost once a meeting ends or the system is shut down. While memory forensics has been applied to other communication tools, Tencent Meeting, a widely used platform in China, lacked a dedicated forensic approach. This paper aims to fill that gap by developing a method to extract evidence directly from Tencent Meeting's memory.

Methods

The core of this research lies in a memory reverse analysis approach for Tencent Meeting. The proposed method involves two main stages:

  1. Tencent Meeting Metadata Reverse Extraction Algorithm: This algorithm focuses on extracting key meeting identifiers. It begins with identifying the Tencent Meeting process in the memory dump. Then, it performs memory extraction of this process and finally extracts metadata such as the meeting theme and meeting number. The process involves using tools like Volatility to analyze memory dumps, identifying the wemeetapp.exe process, and then dumping specific memory files. The extracted .dat files are then scanned and converted to strings to identify and extract the metadata.
  2. Tencent Meeting Data Stream Engraving Algorithm: This algorithm is designed to extract the actual communication data transmitted during meetings. It involves three steps:
  3. Reverse Reconstruction of the Tencent Meeting Data Structure: This step aims to build an understanding of how Tencent Meeting stores data in memory. The researchers manually analyzed memory dumps using a hexadecimal editor (WinHex) to reverse-engineer the structures of text and file messages, identifying key-value pairs and array structures.
  4. Meeting Data Flow Matching: To efficiently locate relevant data structures within the memory image, the Knuth-Morris-Pratt (KMP) algorithm is adapted. This optimized approach first identifies the core message_elem_array structure and then uses validation data to filter and match specific text and file message elements.
  5. Meeting Data Stream Extraction: This final stage uses the identified data structures and matching positions to extract specific data blocks, such as timestamps, participant nicknames, unique IDs, and message content. The extracted information is then organized chronologically to reconstruct the meeting flow, providing a comprehensive view of the communication.

Datasets

The paper does not explicitly mention the use of pre-existing datasets for training or evaluation. Instead, the authors created their own experimental environment.

  • Custom Experimental Environment: A Windows 10 virtual machine was set up using VMware. A Tencent Meeting was initiated with the theme "Memory Forensic" and a specific meeting number (788-357-045) was recorded. Memory image files (.vmem) of this virtual machine were captured during the meeting for analysis. Text messages ("how are you") and file messages ("story for us.mp3") were exchanged within the meeting to generate data for extraction.

Key Insights and Ideas

  • Memory is Key for Instant Messaging Forensics: The paper strongly emphasizes that for applications like Tencent Meeting, where data isn't persistently stored on disk, memory analysis is the only viable forensic approach.
  • Reverse Engineering Data Structures is Crucial: Understanding the internal data structures of Tencent Meeting in memory is fundamental to extracting meaningful information. The authors demonstrated how to reverse-engineer these structures from raw hexadecimal data.
  • KMP Algorithm Adaptation for Efficiency: The adaptation of the KMP algorithm for data flow matching significantly improves the efficiency and accuracy of locating relevant forensic data within large memory dumps.
  • Comprehensive Data Extraction: The proposed method successfully extracts not only metadata (meeting theme, number) but also detailed communication content (text messages, file transfers), providing a much richer forensic picture than traditional methods.
  • Tencent Meeting vs. VooV Meeting: The paper clarifies that VooV Meeting is the international version of Tencent Meeting, and the study focuses on the Chinese version due to author limitations.

Key Results

The experimental verification demonstrated the effectiveness of the proposed method:

  • Metadata Extraction: The method successfully extracted meeting metadata, including the meeting theme ("Memory Forensic"), meeting number (788-357-045), and timestamps.
  • Communication Data Extraction: Text messages, such as "how are you" and "helllo," were accurately recovered.
  • File Transfer Recovery: Transmitted files, including "story for us.mp3," "test.txt," and "analysis515.docx," were successfully extracted along with their relevant details (name, type, size).
  • Reconstruction of Meeting Flow: The extracted data was organized chronologically to reconstruct the meeting process, providing a coherent narrative of the communication.
  • Comparison with Google Meeting: The paper highlights that their method extracted more comprehensive information (including file transfers) compared to existing methods for Google Meeting, which primarily focused on browser-based memory data.

Limitations and Future Work

  • Real-time Voice Data: The authors acknowledge that extracting real-time voice data remains a significant challenge due to audio data encoding complexities. This is identified as a key area for future research, requiring specialized knowledge in audio processing.
  • Generalizability: While the authors suggest their method can be analogically extended to other meeting software, further research would be needed to validate its applicability across different platforms.
  • Tool Development: The paper mentions a lack of tools specifically developed for Tencent Meeting forensics, implying a need for more specialized tools to streamline the process.

Summary updated

BibTeX
@article{yu2025tencent,
  author = {Shilong Yu and Binglong Li and Lin Zhu and Heyu Zhang and Sen Yang and Zhangxiao Li and Wenzheng Feng},
  title = {{Tencent Meeting forensics based on memory reverse analysis}},
  journal = {PeerJ Comput. Sci.},
  year = {2025},
  doi = {10.7717/peerj-cs.2963},
}
Figure