Scholarlink

Researchers trained models on social networking data to successfully predict behaviors and profile users. However, it seems that there is no testing of these models against a malicious network because of adversarial actions. This paper presents a systematic evaluation of the influence of evasion and poisoning attacks on model robustness in social data mining. We used the myPersonality dataset to check the performance of three classifiers (Logistic Regression, Random Forest, Support Vector Machine) under various attacks. If something is changed even a little, the performance is really bad. A systematic response is started to different types of attacks and their quantification of predictive robustness. The relative error does go up to 90%. Use data analytics to boost cybersecurity resilience.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Adversarial Vulnerability Assessment of Social Data Models for Defensive Cyber Operations

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Fahad Amin

摘要

Researchers trained models on social networking data to successfully predict behaviors and profile users. However, it seems that there is no testing of these models against a malicious network because of adversarial actions. This paper presents a systematic evaluation of the influence of evasion and poisoning attacks on model robustness in social data mining. We used the myPersonality dataset to check the performance of three classifiers (Logistic Regression, Random Forest, Support Vector Machine) under various attacks. If something is changed even a little, the performance is really bad. A systematic response is started to different types of attacks and their quantification of predictive robustness. The relative error does go up to 90%. Use data analytics to boost cybersecurity resilience.