Evaluating deep learning variants for cyber-attacks detection and multi-class classification in IoT networks.

Area of Science:

• Cybersecurity
• Machine Learning
• Internet of Things (IoT)

Background:

• The Internet of Things (IoT) is rapidly expanding, increasing its vulnerability to sophisticated cyberattacks.
• Cyberattacks pose significant threats to businesses and organizations, impacting operations and data security.
• Machine learning (ML) and deep learning (DL) offer promising solutions for enhancing cybersecurity defenses.

Purpose of the Study:

• To investigate the effectiveness of deep learning models for detecting cyberattacks in IoT environments.
• To evaluate various deep learning architectures for identifying network traffic anomalies.
• To propose an efficient method for early network data segregation and cyberattack mitigation.

Main Methods:

• Utilized the CICDIoT2023 dataset for evaluating deep learning models.
• Implemented data preprocessing, including robust scalar and label encoding for categorical variables.
• Applied deep learning models such as Deep Neural Networks (DNNs), Convolutional Neural Networks (CNNs), and Recurrent Neural Networks (RNNs) for cyberattack detection.

Main Results:

• The Recurrent Neural Network (RNN) model demonstrated the highest accuracy at 96.56%.
• The proposed approach, utilizing deep learning, proved efficient in identifying cyberattacks within a realistic IoT setting.
• Experimental results confirmed the efficacy of the implemented deep learning models in network traffic analysis.

Conclusions:

• Deep learning models, particularly RNNs, are highly effective for detecting cyberattacks in IoT networks.
• The proposed method offers an efficient solution for mitigating cyber threats in the evolving landscape of IoT.
• Early detection and data segregation using deep learning are crucial for robust IoT security.