An enhanced spotted hyena optimizer-based LSTM framework for accurate DDoS attack detection in internet of things networks
Abstract
The proliferation of IoT devices has made modern networks more vulnerable to Distributed Denial-of-Service (DDoS) attacks, which may overwhelm systems and cause them to crash. Due to the complex structure, enormous volume and ever-changing behavior of IoT-generated network data, conventional detection algorithms often produce too many false alarms and fail to maintain accuracy, making the detection of such attacks very challenging. This paper proposes an integrated strategy of optimization-deep learning for improving detection of distributed denial of service threats in Internet of Things systems. The model includes Long Short-Term Memory (LSTM) network and Enhanced Spotted Hyena Optimizer (E-SHO) for effective hyperparameter tweaking and feature selection. The revised optimizer is integrated with adaptive control strategies, the Levy flight exploration, the inertia weight adjustment and the opposition-based learning to improve the search process and prevent the premature convergence. While the optimization section adjusts model parameters, the LSTM network learns network data’s temporal patterns. It is tested using CICDDoS2019 and N-BaIoT benchmark datasets. The experimental results showed accuracy of 92.21%, precision of 92.22%, recall of 92.22%, F1-score of 92.22% for CICDDoS2019 Dataset and accuracy of 93.1%, precision of 92.8%, recall of 92.8%, F1-score of 92.9% for N-BaIoT Dataset. It also offers reduced false alarm rates and more stability than previous techniques.
Related
- Siamese-Enhanced Multi-View Representation Learning for Robust Attack Detection in the Internet of Vehicles
- ASC-YOLOv8n: Enhanced multi-scale feature fusion for accurate detection of four cigar appearance defects
- Machine learning–enhanced label-free Raman spectroscopy for accurate serum-based detection of early-stage breast cancer via tumor-associated antigens
- Securing intelligent transportation systems in the internet of vehicles through an AI-driven intrusion detection framework
- Hybrid deep feature fusion framework with maximum entropy based selection for accurate diabetic foot ulcer diagnosis from enhanced and segmented medical images
BibTeX
@article{nanthiya2026enhanced,
title = {{An enhanced spotted hyena optimizer-based LSTM framework for accurate DDoS attack detection in internet of things networks}},
author = {D. Nanthiya and R. Manjula Devi and S. B. Gopal},
journal = {Scientific Reports},
year = {2026},
month = oct,
doi = {10.1038/s41598-026-70449-x},
url = {https://www.nature.com/articles/s41598-026-70449-x},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - An enhanced spotted hyena optimizer-based LSTM framework for accurate DDoS attack detection in internet of things networks
AU - D. Nanthiya
AU - R. Manjula Devi
AU - S. B. Gopal
PY - 2026
DA - 2026/10/27
JO - Scientific Reports
DO - 10.1038/s41598-026-70449-x
UR - https://www.nature.com/articles/s41598-026-70449-x
AB - The proliferation of IoT devices has made modern networks more vulnerable to Distributed Denial-of-Service (DDoS) attacks, which may overwhelm systems and cause them to crash. Due to the complex structure, enormous volume and ever-changing behavior of IoT-generated network data, conventional detection algorithms often produce too many false alarms and fail to maintain accuracy, making the detection of such attacks very challenging. This paper proposes an integrated strategy of optimization-deep learning for improving detection of distributed denial of service threats in Internet of Things systems. The model includes Long Short-Term Memory (LSTM) network and Enhanced Spotted Hyena Optimizer (E-SHO) for effective hyperparameter tweaking and feature selection. The revised optimizer is integrated with adaptive control strategies, the Levy flight exploration, the inertia weight adjustment and the opposition-based learning to improve the search process and prevent the premature convergence. While the optimization section adjusts model parameters, the LSTM network learns network data’s temporal patterns. It is tested using CICDDoS2019 and N-BaIoT benchmark datasets. The experimental results showed accuracy of 92.21%, precision of 92.22%, recall of 92.22%, F1-score of 92.22% for CICDDoS2019 Dataset and accuracy of 93.1%, precision of 92.8%, recall of 92.8%, F1-score of 92.9% for N-BaIoT Dataset. It also offers reduced false alarm rates and more stability than previous techniques.
SN - 2045-2322
ER -