Lightweight intrusion detection system using multiscale attention 1D CNN for large scale internet of things

Area of Science:

• Cybersecurity
• Machine Learning
• Network Security

Background:

• The Internet of Things (IoT) is rapidly expanding, increasing its vulnerability to sophisticated cyber attacks.
• Existing security measures are often insufficient against advanced, real-time threats targeting IoT networks.

Purpose of the Study:

• To develop an efficient, real-time intrusion detection system (IDS) for large-scale IoT networks.
• To enhance the detection of diverse cyber attack patterns and methods in IoT environments.

Main Methods:

• Proposed a novel intrusion detection system (IDS) leveraging multiscale attention 1D convolutional neural networks (MA-1D-CNN).
• Integrated multi-scale convolutional kernels with a dual attention mechanism for efficient feature extraction.
• Utilized spatial features to differentiate between normal and malicious IoT network traffic.

Main Results:

• The proposed MA-1D-CNN IDS achieved high accuracy, reaching 91.03% on the UM-NIDS dataset and 99.37% on the UNSW-NB15 dataset.
• Demonstrated superior performance compared to existing state-of-the-art intrusion detection models.
• Validated the model's inference efficiency and intrusion detection capabilities on benchmark datasets.

Conclusions:

• The MA-1D-CNN model is a lightweight, feature-efficient, and high-precision solution for real-time attack detection in IoT networks.
• The proposed IDS effectively addresses the need for robust security in rapidly growing IoT ecosystems.
• This approach offers a promising direction for securing large-scale IoT deployments against evolving cyber threats.