An IoT-Focused Intrusion Detection System Approach Based on Preprocessing Characterization for Cybersecurity...

Area of Science:

• Cybersecurity
• Machine Learning
• Network Traffic Analysis

Background:

• Internet of Things (IoT) networks handle vast data, making them vulnerable to increasingly sophisticated cybersecurity attacks.
• Existing intrusion detection systems require enhanced accuracy to effectively mitigate these threats.
• Machine learning algorithms show promise for improving intrusion detection system (IDS) performance.

Purpose of the Study:

• To evaluate various data preprocessing techniques for machine learning-based intrusion detection in IoT networks.
• To investigate the impact of network traffic categorization on the accuracy of intrusion detection models.
• To identify optimal preprocessing strategies for enhancing the classification of cyberattacks.

Main Methods:

• Utilized two benchmark datasets (UGR16, UNSW-NB15) and KDD99 for evaluation.
• Applied preprocessing techniques including scalar and normalization functions.
• Categorized network traffic features into four groups: basic connection, content, statistical, and traffic/connection direction-based characteristics.

Main Results:

• The proposed traffic categorization and preprocessing methods enhanced model accuracy by up to 45%.
• Preprocessing specific feature groups proved more effective in improving classification accuracy.
• The machine learning algorithm demonstrated improved ability to correctly classify attack-related parameters.

Conclusions:

• Network traffic categorization combined with targeted data preprocessing significantly boosts intrusion detection accuracy in IoT environments.
• This approach offers a viable strategy for developing more robust and effective cybersecurity solutions for IoT.
• Further research into feature selection within categorized traffic can lead to even more precise attack detection.