Using Machine Learning Multiclass Classification Technique to Detect IoT Attacks in Real Time.

Area of Science:

• Cybersecurity
• Network Security
• Machine Learning

Background:

• The proliferation of Internet of Things (IoT) devices has led to an increase in sophisticated cyberattacks.
• Existing intrusion detection systems (IDS) often struggle with the volume and velocity of IoT network traffic.
• There is a need for efficient and accurate real-time detection of IoT-specific threats.

Purpose of the Study:

• To develop and evaluate a real-time intrusion detection system (IDS) for identifying Internet of Things (IoT) attacks.
• To enhance detection accuracy and minimize prediction time using multiclass classification models.
• To leverage the PySpark architecture for scalable big data processing in cybersecurity.

Main Methods:

• Utilized the IoT-23 dataset comprising network traffic from smart home IoT devices.
• Applied data preprocessing techniques including cleaning, transformation, scaling, and Synthetic Minority Oversampling Technique (SMOTE).
• Employed multiclass classification algorithms with the OneVsRest (OVR) technique and feature selection methods within the PySpark framework.

Main Results:

• Extreme Gradient Boosting achieved a high accuracy of 98.89% in detecting IoT attacks.
• Random Forest exhibited the fastest prediction time at 0.0311 seconds.
• The proposed IDS demonstrated superior performance compared to existing approaches in real-time detection.

Conclusions:

• The developed real-time IDS effectively detects IoT attacks with high accuracy and efficiency.
• The PySpark architecture facilitates scalable and rapid analysis of network traffic for intrusion detection.
• This research contributes a robust solution for securing the expanding landscape of Internet of Things devices.