RETRACTED: Enhancing IoT cybersecurity through lean-based hybrid feature selection and ensemble learning: A visual...

Area of Science:

• Cybersecurity
• Intrusion Detection Systems
• Internet of Things (IoT) Security

Background:

• The increasing sophistication of cyber threats in IoT environments necessitates advanced, scalable intrusion detection systems.
• Existing systems often struggle with high computational overhead and suboptimal detection efficiency.
• The need for real-time, accurate threat detection is critical for protecting IoT infrastructure.

Purpose of the Study:

• To propose a novel Lean-based hybrid Intrusion Detection framework for the IoT setting.
• To enhance the accuracy and efficiency of cyber threat detection in IoT systems.
• To develop a scalable and robust solution for real-time intrusion detection.

Main Methods:

• A hybrid framework combining Particle Swarm Optimization and Genetic Algorithm (PSO-GA) for feature selection.
• Utilizing Extreme Learning Machine and Bootstrap Aggregation (ELM-BA) for feature classification.
• Employing Lean principles for minimal computational overhead and optimal efficiency.

Main Results:

• Achieved high detection rates on the CICIDS-2017 dataset.
• Demonstrated 100% accuracy for critical attack categories including PortScan, SQL Injection, and Brute Force.
• Validated through statistical verification and visual evaluation metrics, proving model robustness.

Conclusions:

• The proposed Lean-based hybrid Intrusion Detection framework offers scalable and effective cyber threat detection for IoT.
• The framework minimizes false positives, reduces decision-making latency, and increases IoT infrastructure resilience.
• It is suitable for real-world deployment in smart cities and industrial IoT systems.