Scholarlink

Insider threats are a growing concern for the security and stability of critical infrastructure systems. By exploiting authorized access, insiders can bypass traditional security measures, leading to significant disruptions. This study explores how machine learning, specifically Long Short-Term Memory (LSTM) networks and Support Vector Machines (SVM), can help tackle this challenge. Using real-world data, we manually tested 100 individual cases to assess each model’s performance. LSTM stood out with an accuracy of 86%, thanks to its ability to understand sequences and detect subtle behavioral patterns, while SVM achieved 76%, showing its strength with simpler cases. Through this hands-on testing, we gained a clear picture of where these models shine and where they fall short. Our findings suggest that blending deep learning with traditional methods could strike a balance between accuracy and efficiency. This research highlights practical ways to use AI for real-time detection and stronger defense against insider threats, ensuring critical systems remain secure.

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

MShield LSTM Model for Detecting Insider Threat Attacks in Critical Infrastructure Using Machine Learning

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Venkatesh Mannuru,
  • Ali Al-Sinayyid,
  • Sasidhar kadiyala,
  • Rohith Reddy Battula,
  • Vaishnavi kottala

摘要

Insider threats are a growing concern for the security and stability of critical infrastructure systems. By exploiting authorized access, insiders can bypass traditional security measures, leading to significant disruptions. This study explores how machine learning, specifically Long Short-Term Memory (LSTM) networks and Support Vector Machines (SVM), can help tackle this challenge. Using real-world data, we manually tested 100 individual cases to assess each model’s performance. LSTM stood out with an accuracy of 86%, thanks to its ability to understand sequences and detect subtle behavioral patterns, while SVM achieved 76%, showing its strength with simpler cases. Through this hands-on testing, we gained a clear picture of where these models shine and where they fall short. Our findings suggest that blending deep learning with traditional methods could strike a balance between accuracy and efficiency. This research highlights practical ways to use AI for real-time detection and stronger defense against insider threats, ensuring critical systems remain secure.