Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

TrendSim: Simulating Trending Topics in Social Media Under Poisoning Attacks with LLM-based Multi-agent System

Zeyu Zhang, Jianxun Lian, Chen Ma, Yaning Qu, Ye Luo, Lei Wang, Rui Li, Xu Chen, Yankai Lin, Le Wu, Xing Xie, Ji-Rong Wen

TL;DR

TrendSim introduces an LLM-based multi-agent framework to simulate trending topics on social media under poisoning attacks. It integrates a time-aware interaction mechanism, centralized message dissemination, and memory-driven user agents along with prototype-based attackers to model realistic dynamics and psychological states. Through multi-aspect evaluations, TrendSim demonstrates attacker efficacy, defense potentials such as content censorship, and the framework’s scalability, while acknowledging limitations and ethical considerations. The work provides a foundation for interpretable, extensible simulations of social phenomena in real-time trending topics with implications for defense policy and platform design.

Abstract

Trending topics have become a significant part of modern social media, attracting users to participate in discussions of breaking events. However, they also bring in a new channel for poisoning attacks, resulting in negative impacts on society. Therefore, it is urgent to study this critical problem and develop effective strategies for defense. In this paper, we propose TrendSim, an LLM-based multi-agent system to simulate trending topics in social media under poisoning attacks. Specifically, we create a simulation environment for trending topics that incorporates a time-aware interaction mechanism, centralized message dissemination, and an interactive system. Moreover, we develop LLM-based human-like agents to simulate users in social media, and propose prototype-based attackers to replicate poisoning attacks. Besides, we evaluate TrendSim from multiple aspects to validate its effectiveness. Based on TrendSim, we conduct simulation experiments to study four critical problems about poisoning attacks on trending topics for social benefit.

TrendSim: Simulating Trending Topics in Social Media Under Poisoning Attacks with LLM-based Multi-agent System

TL;DR

TrendSim introduces an LLM-based multi-agent framework to simulate trending topics on social media under poisoning attacks. It integrates a time-aware interaction mechanism, centralized message dissemination, and memory-driven user agents along with prototype-based attackers to model realistic dynamics and psychological states. Through multi-aspect evaluations, TrendSim demonstrates attacker efficacy, defense potentials such as content censorship, and the framework’s scalability, while acknowledging limitations and ethical considerations. The work provides a foundation for interpretable, extensible simulations of social phenomena in real-time trending topics with implications for defense policy and platform design.

Abstract

Trending topics have become a significant part of modern social media, attracting users to participate in discussions of breaking events. However, they also bring in a new channel for poisoning attacks, resulting in negative impacts on society. Therefore, it is urgent to study this critical problem and develop effective strategies for defense. In this paper, we propose TrendSim, an LLM-based multi-agent system to simulate trending topics in social media under poisoning attacks. Specifically, we create a simulation environment for trending topics that incorporates a time-aware interaction mechanism, centralized message dissemination, and an interactive system. Moreover, we develop LLM-based human-like agents to simulate users in social media, and propose prototype-based attackers to replicate poisoning attacks. Besides, we evaluate TrendSim from multiple aspects to validate its effectiveness. Based on TrendSim, we conduct simulation experiments to study four critical problems about poisoning attacks on trending topics for social benefit.

Paper Structure

This paper contains 46 sections, 2 theorems, 7 equations, 8 figures, 9 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. Poisoning Attack in Social Media
  4. LLM-based Multi-agent Social Simulation
  5. Methods
  6. Overview of TrendSim
  7. Multi-agent Simulation Environment
  8. Time-aware Interaction Mechanism
  9. Centralized Message Dissemination
  10. User-Environment Interactive System
  11. LLM-based User Agent
  12. Perception Module
  13. Memory Module
  14. Action Module
  15. Prototype-based Attacker Agent
  16. ...and 31 more sections

Key Result

Theorem 1

The function $P(t)$ is $G_0$-smooth.

Figures (8)

  • Figure 1: The framework of TrendSim for simulating trending topics in social media.
  • Figure 2: The framework of LLM-based user agents in TrendSim.
  • Figure 3: The user's psychological conditions in social media platforms over time. The curves represent mean values, and the shading represents standard deviations.
  • Figure 4: The user psychological conditions in different groups of simulation experiments.
  • Figure 5: Compare $f(t)$ with real-world populations.
  • ...and 3 more figures

Theorems & Definitions (4)

  • Theorem 1
  • proof
  • Theorem 2
  • proof