Tongxin xuebao (Nov 2024)

Intention embedding method based social bot detection

  • NIU Hongfeng,
  • LI Jiawei,
  • SONG Yunpeng,
  • CAI Zhongmin

Journal volume & issue
Vol. 45
pp. 194 – 205

Abstract

Read online

Artificial intelligence generated content technology has significantly enhanced the disguise capabilities of social bots, presenting new challenges to existing bot detection methods. By modeling the intentions of social media users through intention representation, a intention embedding method based social bot detection was proposed, thereby avoiding the difficulty of directly detecting bots with enhanced behavioral camouflage at the action level on social platforms. Experimental results show that the detection model using intention embedding improves the accuracy of social bot detection by 5.58 percentage points compared to models not utilizing intention embedding, and it enhances the recognition capability of specific types of social bots, verifying the effectiveness of intention embedding in improving the performance of human-bot detection tasks.

Keywords