Hybrid evolutionary machine learning model for advanced intrusion detection architecture for cyber threat...

Area of Science:

• Computer Science
• Cybersecurity
• Machine Learning

Background:

• The increasing complexity and evolution of cyber threats necessitate advanced network security solutions.
• Traditional intrusion detection systems (IDS) struggle with adaptability to novel attacks and scalability across diverse network environments.

Purpose of the Study:

• To propose and validate an Evolutionary Machine Learning Algorithm for robust intrusion detection.
• To address the challenges of adaptability to new threats and scalability in network security.
• To develop a preemptively adaptive IDS capable of handling emerging cyber threats.

Main Methods:

• Utilized a hybrid approach combining Genetic Algorithm (GA) based feature selection with Decision Tree-Support Vector Machine (DT-SVM) classification.
• Employed GA for evolutionary feature selection to identify and optimize relevant network data characteristics.
• Validated the algorithm on the BoT-IoT and UNSW-NB15 datasets, encompassing IoT-specific and general network intrusion scenarios.

Main Results:

• The proposed IDS demonstrated superior performance compared to traditional methodologies.
• Achieved high accuracy, recall, and low false positive rates in identifying both known and novel threats.
• The GA-driven feature selection effectively optimized the classification model's focus on critical network data components.

Conclusions:

• The Evolutionary Machine Learning Algorithm offers a scalable and adaptable solution for robust intrusion detection.
• The hybrid DT-SVM and GA-based feature selection approach effectively balances efficiency and accuracy.
• The developed IDS is capable of real-time data stream processing for prompt and precise intrusion detection.