Sequential Model Based Intrusion Detection System for IoT Servers Using Deep Learning Methods.

Area of Science:

• Cybersecurity
• Artificial Intelligence
• Internet of Things (IoT)

Background:

• IoT devices are integral to daily life, necessitating robust security measures against cyber threats.
• Traditional Intrusion Detection Systems (IDS) for IoT servers require enhanced accuracy and performance.
• Deep learning offers advanced capabilities for pattern recognition and data analysis, surpassing traditional machine learning.

Purpose of the Study:

• To propose novel deep learning-based methods for improving Intrusion Detection Systems (IDS) in IoT environments.
• To enhance the accuracy and performance of IoT security by leveraging sequential data modeling.
• To investigate the effectiveness of deep learning models in detecting cyber threats targeting IoT servers.

Main Methods:

• Utilized sequential models, specifically Text-CNN and Gated Recurrent Unit (GRU), treating network data as a language model.
• Collected features from both network layers (tcpdump packets) and application layers (system routines).
• Compared deep learning approaches against traditional methods for intrusion detection.

Main Results:

• Deep learning methods, particularly sequential models, demonstrated superior feature extraction capabilities.
• Experiments confirmed higher F1-scores for deep learning-based IDS compared to traditional approaches.
• The proposed methods showed significant improvements in detecting intrusions in IoT networks.

Conclusions:

• Sequential model-based IDS utilizing deep learning significantly contribute to the security of IoT servers.
• Deep learning methods offer a promising direction for developing more effective and efficient intrusion detection systems.
• The study highlights the potential of Text-CNN and GRU in securing the evolving landscape of IoT infrastructure.