Addressing smart city security: machine and deep learning methodology combining feature selection and two-tier cooperative framework tuned by metaheuristics

The creation of safe and effective smart cities depends on the integration of internet of things (IoT) systems with cutting-edge technology. Conventional security techniques have not been able to keep up with the increasingly complex attacks targeting IoT networks. Adaptive artificial intelligence (...

Full description

Saved in:
Bibliographic Details
Published in:Cluster computing Vol. 28; no. 11; p. 705
Main Authors: Zivkovic, Miodrag, Jovanovic, Luka, Jocovic, Vladimir, Nikolic, Bosko, Zeljkovic, Vico, Abdel-Salam, Mahmoud, Mravik, Milos, Muthusamy, Suresh, Bacanin, Nebojsa
Format: Journal Article
Language:English
Published: New York Springer US 01.10.2025
Springer Nature B.V
Subjects:
Accuracy
Adaptive algorithms
Algorithms
Artificial intelligence
Artificial neural networks
Blockchain
Cities
Computer Communication Networks
Computer Science
Cybersecurity
Datasets
Deep learning
Economic growth
Evolutionary algorithms
Explainable artificial intelligence
Feature selection
Heuristic methods
Infrastructure
Internet of Things
Machine learning
Methodology
Operating Systems
Optimization
Processor Architectures
Public services
Security systems
Smart cities
Waste management
AdaBoost
Feature selection
Smart cities
LightGBM
Firefly algorithm
Security
Optimization
CNNs
ISSN:1386-7857, 1573-7543
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:The creation of safe and effective smart cities depends on the integration of internet of things (IoT) systems with cutting-edge technology. Conventional security techniques have not been able to keep up with the increasingly complex attacks targeting IoT networks. Adaptive artificial intelligence (AI)-driven solutions have surfaced to address these issues, improving the security and functionality of the IoT framework while ensuring dependable settings for people and systems. Convolutional neural networks (CNNs) and cutting-edge boosting machine learning classifiers, specifically adaptive boosting (AdaBoost), light gradient boosting (LigthGBM), and eXtreme gradient boosting (XGBoost), are combined together for addressing smart city security issue. Moreover, to produce robust and adaptable solutions, metaheuristics approaches are employed for machine learning hyperparameters’ tweaking and performance optimization. Therefore, this study presents two approaches to tackle this significant issue. First, metaheuristics are leveraged to simultaneously perform feature selection and hyperparameter tuning for AdaBoost classifier, using a composite objective function. Second, the proposed methodology introduces a two-tier cooperative framework that integrates CNNs with XGBoost and LigthGBM, with further performance optimization through metaheuristic techniques. To identify particular kinds of harmful actions, a realistic dataset was used for multi-class classification. The evolutionary adaptive firefly algorithm (EAFA), that is also introduced for the purpose of this study, achieved a noteworthy accuracy of 0.995565 in the best performing simulation, along with macro average precision of 0.958948, sensitivity of 0.842059 and f1-score equal to 0.883886. Explainable AI (XAI) approaches were used to further increase model interpretability, highlighting important feature implications towards focused data gathering and system enhancements. By striking a balance between high accuracy and computational economy, the suggested design outperformed simple CNNs. Proposed methodology may be used practically in integrated IoT systems to manage data flows, prevent network-wide assaults, and improve device security. Incorporating this design into the Metaverse may also improve device security, boost user confidence, and connect virtual and physical worlds, opening the door to more intelligent and secure IoT infrastructures in smart cities.
Bibliography:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:1386-7857
1573-7543
DOI:10.1007/s10586-025-05378-x