Adaptive weighted kernel support vector machine-based circle search approach for intrusion detection in IoT environments

Nowadays, the Internet of Things (IoT) is considered a globally implemented technology in automated network structures. However, the procedures utilized in IoT devices can have security exposures that may affect the entire system. IoT systems are vulnerable to attacks due to the lack of basic securi...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:Signal, image and video processing Ročník 18; číslo 5; s. 4479 - 4490
Hlavní autoři: Geetha, C., Johnson, Shiny Duela, Oliver, A. Sheryl, Lekha, D.
Médium: Journal Article
Jazyk:angličtina
Vydáno: London Springer London 01.07.2024
Springer Nature B.V
Témata:
Accuracy
Blockchain
Communications traffic
Computer Imaging
Computer Science
Cybersecurity
Data analysis
Datasets
Deep learning
Denial of service attacks
Design
Efficiency
False alarms
Feature extraction
Genetic algorithms
Image Processing and Computer Vision
Internet of Things
Intrusion detection systems
Kernel functions
Multimedia Information Systems
Neural networks
Original Paper
Pattern Recognition and Graphics
Preprocessing
Radio communications
Signal,Image and Speech Processing
Smart houses
Support vector machines
Vision
Circle search algorithm
Attack detection
Internet of Things
Adaptive weighted kernel support vector machine
ISSN:1863-1703, 1863-1711
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:Nowadays, the Internet of Things (IoT) is considered a globally implemented technology in automated network structures. However, the procedures utilized in IoT devices can have security exposures that may affect the entire system. IoT systems are vulnerable to attacks due to the lack of basic security protocols. An intrusion detection system (IDS) is a fundamental cybersecurity device used to identify any unusual host or network activity. The common approach to designing IDS methods is to use classifiers like detection units. However, classifiers often cause high misclassification and low detection rates due to the presence of inappropriate, noisy, and unnecessary features in the large feature space. To address these issues, this article proposes a novel IoT-based intrusion detection system. The proposed IoT-IDS system comprises data analysis, data preprocessing, feature selection, and attack prediction phases. During preprocessing, unprocessed data are transformed into a simpler and more efficient form to perform further processing. In the feature selection process, unnecessary and inappropriate features are extracted, and a feature subset is generated. To attain optimal features, this paper employs a Pareto-based ensemble technique. The normal and anomaly behaviors of network traffic are determined through a set of extraction rules. Finally, the proposed intrusion detection classifier, named adaptive weighted kernel support vector machine-based circle search (AWSVM-CS) approach, detects and discriminates normal and anomaly features accurately. After attack detection, the network generates a warning alert, which helps the IoT network administrator take further actions to solve the security issue. The efficiency of the proposed approach is investigated in terms of diverse evaluation measures, including accuracy, precision, sensitivity, F1-score, false alarm rate, average accuracy, attack detection rate, and attack accuracy. The experimental analysis reveals that the proposed approach achieves a high attack detection rate of about 93.3% for the UNSW-NB15 dataset and 93% for the KDD99 dataset. Overall, this AWSVM-CS approach provides a robust solution to the security challenges in IoT systems by effectively detecting and preventing potential security breaches.
Bibliografie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:1863-1703
1863-1711
DOI:10.1007/s11760-024-03088-2