A secure edge computing model using machine learning and IDS to detect and isolate intruders

Area of Science:

• Computer Science
• Cybersecurity
• Machine Learning

Background:

• The rapid expansion of Internet of Things (IoT) and edge computing introduces significant security challenges.
• Existing security models struggle to effectively detect and mitigate intrusions in dynamic edge environments.

Purpose of the Study:

• To propose a secure edge computing model for robust intrusion detection and isolation.
• To enhance the security of edge networks, particularly for Internet of Things (IoT) applications.

Main Methods:

• A hybrid Intrusion Detection System (IDS) model combining Linear Discriminant Analysis (LDA) and Logistic Regression (LR).
• Utilizing machine learning for real-time analysis and identification of intrusive activities at edge nodes.
• Implementing a device and data isolation mechanism without alerting neighboring nodes to prevent lateral movement of intruders.

Main Results:

• The proposed IDS achieved high performance with 96.56% accuracy and 95.78% precision.
• Demonstrated a significantly fast training time of 0.04 seconds.
• Effectively detected and isolated various types of cyber-attacks in simulated edge environments.

Conclusions:

• The hybrid LDA-LR model offers a secure and efficient solution for intrusion detection in edge computing.
• The proposed system enhances the overall security posture of IoT-enabled edge networks.
• The isolation mechanism prevents intruders from compromising the entire edge network.