Towards secure IoT networks: A comprehensive study of metaheuristic algorithms in conjunction with CNN using a...

Area of Science:

• Computer Science
• Cybersecurity
• Network Engineering

Background:

• The proliferation of Internet of Things (IoT) devices has expanded the attack surface, increasing security vulnerabilities.
• Existing security datasets are often inadequate for the unique challenges of IoT environments.
• Developing robust security measures for IoT networks is critical due to widespread connectivity and automation.

Purpose of the Study:

• To address the lack of specialized datasets for IoT security by generating a realistic simulation.
• To evaluate the performance of metaheuristic algorithms combined with Convolutional Neural Networks (CNNs) for intrusion detection in IoT networks.
• To identify optimal hyperparameters for CNNs to maximize their effectiveness in detecting IoT-specific attacks.

Main Methods:

• Generation of a custom dataset using the Contiki-OS Cooja Simulator, mimicking four attack types: blackhole, sinkhole, flooding, and version number attacks.
• Evaluation of metaheuristic algorithms in conjunction with Convolutional Neural Networks (CNNs) for intrusion detection.
• Hyperparameter optimization for CNN models to enhance detection accuracy.

Main Results:

• The generated dataset effectively simulates real-world IoT attack scenarios.
• The combination of metaheuristic algorithms and CNNs demonstrates promising results in identifying network intrusions.
• Optimal CNN hyperparameters were identified for improved performance in detecting specific IoT attacks.

Conclusions:

• The study provides a valuable, customized dataset for IoT security research and development.
• The findings offer practical insights into enhancing the security of IoT networks through advanced machine learning techniques.
• This work contributes to a better understanding and implementation of effective security measures for the Internet of Things.