An intrusion detection system in the Internet of Things with deep learning and an improved arithmetic optimization...

Area of Science:

• Cybersecurity
• Network Security
• Artificial Intelligence

Background:

• Internet of Things (IoT) devices face significant cybersecurity risks due to inherent security limitations and resource constraints.
• Traditional Intrusion Detection Systems (IDS) struggle with imbalanced datasets, high-dimensional network traffic, and detecting zero-day cyber-attacks.

Purpose of the Study:

• To propose an advanced IDS framework for the Internet of Things (IoT) to enhance the detection of cyber-attacks.
• To address challenges of imbalanced datasets, high-dimensional traffic, and the detection of novel threats in IoT environments.

Main Methods:

• Utilized game-theory-based Generative Adversarial Networks (GAN) for dataset balancing.
• Employed a hybrid Arithmetic Optimization Algorithm (AOA) and Sine Cosine Algorithm (SCA) for optimal feature selection.
• Developed a Parallel Convolutional Neural Network (PCNN) integrated with a Long Short-Term Memory (LSTM) layer for accurate attack classification.

Main Results:

• The proposed ASPCNNLSTM model achieved 99.86% precision on the NSL-KDD dataset.
• Demonstrated an attack detection accuracy of 98.67% on the UNSW-NB15 dataset.
• Significantly outperformed traditional CNN, LSTM, and existing feature selection techniques.

Conclusions:

• The advanced IDS framework effectively balances datasets and selects optimal features for improved network traffic analysis.
• The ASPCNNLSTM model enhances spatial and temporal feature learning, providing robust detection against complex and unknown cyber threats in IoT ecosystems.