Hybrid IoT-CAD system: optimized feature selection based gated recurrent residual deep learning for cyber attack detection in IoT networks
Abstract The growth in intelligent services with fewer resources and advanced communication technologies has positioned the Internet of Things (IoT) as the leading framework for less power lossy networks. However, IoT systems face significant risks from cyberattacks due to drawbacks in storage and c...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Springer
2025-08-01
|
| Series: | Discover Internet of Things |
| Subjects: | |
| Online Access: | https://doi.org/10.1007/s43926-025-00190-w |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Abstract The growth in intelligent services with fewer resources and advanced communication technologies has positioned the Internet of Things (IoT) as the leading framework for less power lossy networks. However, IoT systems face significant risks from cyberattacks due to drawbacks in storage and computation. In this work, a novel approach is introduced that combines deep learning (DL) and machine learning (ML) to improve cyber-attack detection in IoT networks, which solves cybersecurity issues. The proposed IoT system includes four steps: pre-processing, augmentation, feature selection, and classification. In the initial phase, data are collected from the ToN-IoT dataset and the BoT-IoT dataset. Then, pre-processing is employed using Z-score normalization and missing value imputation to improve the quality of data. Next, to solve the data imbalance problem, Conditional Generative Adversarial Networks (CGAN) are employed. After that, the Binary Genetic Dung Beetle Optimization (B-GDBO) method is utilized for optimal feature selection and also reduces the risk of overfitting. Ultimately, a one-dimensional Gated Recurrent Unit with attention based deep residual network combined with Support Vector Machine (1D GRU-AtDRN-SVM) is proposed for accurate attack detection and classification, and improves accuracy in classification. This hybrid model supports improved detection accuracy and resilience for protecting IoT networks from cyber threats. The experimental results attain an accuracy of 99.5%, a precision of 98.72%, and an F1-score of 98.26% for the ToN-IoT dataset. Also, the BoT-IoT dataset achieves an accuracy of 99.3%, a precision of 99.01%, and an F1-score of 98.89%. |
|---|---|
| ISSN: | 2730-7239 |