Submit

Login Login

Register Register

Home / Journals / CMC / Online First / doi:10.32604/cmc.2026.083690
Journal Menu
Special Issues
Table of Content

All issues

Online First

2026

2025

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

2004

Open Access

ARTICLE

Optimized Hybrid Deep Learning Frameworks for IoT Cybersecurity against IoT Attacks in Smart Cities

Muhammad Usman Ghani1, Muhammad Javed1, Zeeshan Ali Haider2, Mohd Faizal Bin Yusof3, Jamil Abedalrahim Jamil Alsayaydeh4,*, Inam Ullah5,*, Fida Muhammad Khan2
1 Gomal Research Institute of Computing (GRIC), Gomal University, Dera Ismail Khan, Pakistan
2 Department of Computer Science, Qurtuba University of Science & Information Technology, Peshawar, Pakistan
3 General Education and Foundation Program, Faculty of Resilience, Rabdan Academy, 65 Al Inshirah Street, Abu Dhabi, United Arab Emirates
4 Department of Engineering Technology, Fakulti Teknologi dan Kejuruteraan Elektronik dan Komputer (FTKEK), Universiti Teknikal Malaysia Melaka (UTeM), Melaka, Malaysia
5 Department of Computer Engineering, Gachon University, Seongnam, Republic of Korea
* Corresponding Author: Jamil Abedalrahim Jamil Alsayaydeh. Email: email; Inam Ullah. Email: email
(This article belongs to the Special Issue: Advances in Machine Learning and Artificial Intelligence for Intrusion Detection Systems, 2nd Edition)

Computers, Materials & Continua https://doi.org/10.32604/cmc.2026.083690

Received 08 April 2026; Accepted 05 June 2026; Published online 18 August 2026

Abstract

The Internet of Things (IoT) networks in smart cities experience high-dimensional, time-dependent traffic types, and the detection of attacks is difficult in a timely fashion, particularly in the case of imbalanced classes of attacks. Two hybrid deep learning-based intrusion detection frameworks, TimeSpaceNet and ContextFusionNet, are proposed for IoT intrusion detection: TimeSpaceNet, a CNN-LSTM model enhanced with spatial-temporal normalization, and ContextFusionNet, a CNN-BiLSTM model strengthened with contextual fusion attention. For both models, class imbalance is addressed with SMOTE, and training convergence is assisted by the ADOPT optimizer. All the models are tested on the IoT Bot dataset, which contains benign traffic and several categories of IoT attack, including DDoS, DoS, key logging, and data theft. The experimental results show that ContextFusionNet achieves the best overall performance, with accuracy of 97.2%, precision of 96.3%, recall of 94.6%, and F1-score of 95.4%. TimeSpaceNet also has high accuracy (95.4%), precision (94.2%), recall (91.0%), and F1-score (92.5%). These results indicate that a combination of spatial feature extraction, bidirectional temporal modeling, and attention-based fusion is beneficial for improved threat detection in IoT systems, across both attack classes and the accuracy for underrepresented attacks. The proposed framework provides a promising basis for smart-city IoT cybersecurity, while future work should evaluate deployment performance on larger real-time and edge-based environments.

Keywords

IoT security; deep learning; TimeSpaceNet; ContextFusionNet; spatial-temporal normalization; contextual fusion attention; cyberattack detection; smart cities

  • 196

    View

  • 43

    Download

  • 0

    Like

Related articles
Copyright© 2026 Tech Science Press
© 1997-2026 TSP (Henderson, USA) unless otherwise stated

Share Link