增强社交网络安全社交机器人账号检测与防御方法及装置
By optimizing the social robot detection model through adversarial training and reweighting mechanisms, the vulnerability of existing models to cyberattacks is addressed, the robustness and defense capabilities of the detection model are improved, and the security of social platforms is ensured.
CN117614687BActive Publication Date: 2026-07-17TIANJIN UNIV
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- TIANJIN UNIV
- Filing Date
- 2023-11-24
- Publication Date
- 2026-07-17
AI Technical Summary
Technical Problem
Existing social bot detection models are vulnerable to cyberattacks and prone to failure, leading to a proliferation of bot accounts and spam accounts on social media platforms and a loss of control over network management.
Method used
An adversarial attack strategy based on graph node injection is adopted. The social robot detection model is optimized through adversarial training and reweighting mechanism, and a composite loss function is constructed to enhance the robustness and defense capability of the model.
Benefits of technology
This improves the social robot detection model's defense capabilities under various network attack scenarios, enhances the model's security and robustness, and ensures reliable detection in complex network environments.
✦ Generated by Eureka AI based on patent content.
Smart Images
Figure CN117614687B_ABST
Abstract
本发明公开了一种增强社交网络安全社交机器人账号检测与防御方法及装置，方法包括：部署一基于图节点注入的对抗攻击策略，用以模拟新社交机器人节点的注入；将筛选后的有效扰动节点添加至社交网络图中，创建含有对抗性扰动的社交网络图；在获取所有新增节点后，将其与原始训练样本集合并，构建新的对抗训练样本集；启动社交机器人检测模型的对抗训练过程，并引入重加权机制；构建复合损失函数，融合原始样本的训练损失与新增节点的对抗损失；在完成对抗训练并获得增强鲁棒性的社交机器人检测模型后，用于识别和检测潜在的社交机器人账户。装置包括：处理器和存储器。本发明加强对网络平台的监管和检测，对网络攻击加强清洁和抵御能力。
Need to check novelty before this filing date? Find Prior Art