Securing SDON with hybrid evolutionary intrusion detection system: An ensemble algorithm for feature selection and classification

IF 2.6 3区计算机科学 Q2 ENGINEERING, ELECTRICAL & ELECTRONIC

Optical Fiber Technology Pub Date : 2025-03-20 DOI:10.1016/j.yofte.2025.104206

Benitha Christinal J. , Ameelia Roseline A.

{"title":"Securing SDON with hybrid evolutionary intrusion detection system: An ensemble algorithm for feature selection and classification","authors":"Benitha Christinal J. , Ameelia Roseline A.","doi":"10.1016/j.yofte.2025.104206","DOIUrl":null,"url":null,"abstract":"<div><div>Software-Defined Optical Network (SDON) is a modernized approach to manage optical networks by integrating the principles of Software-Defined Networking (SDN). This integration allows for enhanced programmability, flexibility, and automation in network operations, addressing the growing demand for efficient and adaptable network infrastructures. SDON is powered by a centralized controller, where the entirety of network intelligence is integrated into the control plane. However, this centralization introduces significant challenges related to data privacy and network security. To address these security issues, this paper proposes an adaptive IDS-SDON-EFSC (Intrusion Detection System for Software-Defined Optical Network with Ensemble Feature Selection and Classification) framework. In the proposed framework, during the initial phase, the gathered data is pre-processed to remove noise, and Adaptive Synthetic (ADASYN) sampling is implemented to balance imbalanced datasets. Overfitting is mitigated using Root Mean Square Deviation (RMSD) regularization, which reduces the loss between training and test data. In the second phase, the Deep Convolutional Neural Network Attention-based Bi-directional Long Short-Term Memory (DCNN-AttBiLSTM) system is employed for classification. The hybrid metaheuristic-Adam optimizer is used to optimize hyperparameter selection. Experiments demonstrate the efficacy of the IDS-SDON-EFSC model, which was trained and evaluated on the most current and practical datasets in the SDON domain: InSDN, SDN-IoT, and SDNFlow. The model achieved excellent performance, discerning various forms of intrusion with 100% accuracy on training data and 99.71% accuracy on test data. Additionally, it achieved a minimal latency rate of 1.6%, reduced controller overhead, and a minimal degradation rate of 2.1% leading to improved throughput performance when executed in real-time networks.</div></div>","PeriodicalId":19663,"journal":{"name":"Optical Fiber Technology","volume":"93 ","pages":"Article 104206"},"PeriodicalIF":2.6000,"publicationDate":"2025-03-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Optical Fiber Technology","FirstCategoryId":"94","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S1068520025000811","RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q2","JCRName":"ENGINEERING, ELECTRICAL & ELECTRONIC","Score":null,"Total":0}

引用次数: 0

Abstract

Software-Defined Optical Network (SDON) is a modernized approach to manage optical networks by integrating the principles of Software-Defined Networking (SDN). This integration allows for enhanced programmability, flexibility, and automation in network operations, addressing the growing demand for efficient and adaptable network infrastructures. SDON is powered by a centralized controller, where the entirety of network intelligence is integrated into the control plane. However, this centralization introduces significant challenges related to data privacy and network security. To address these security issues, this paper proposes an adaptive IDS-SDON-EFSC (Intrusion Detection System for Software-Defined Optical Network with Ensemble Feature Selection and Classification) framework. In the proposed framework, during the initial phase, the gathered data is pre-processed to remove noise, and Adaptive Synthetic (ADASYN) sampling is implemented to balance imbalanced datasets. Overfitting is mitigated using Root Mean Square Deviation (RMSD) regularization, which reduces the loss between training and test data. In the second phase, the Deep Convolutional Neural Network Attention-based Bi-directional Long Short-Term Memory (DCNN-AttBiLSTM) system is employed for classification. The hybrid metaheuristic-Adam optimizer is used to optimize hyperparameter selection. Experiments demonstrate the efficacy of the IDS-SDON-EFSC model, which was trained and evaluated on the most current and practical datasets in the SDON domain: InSDN, SDN-IoT, and SDNFlow. The model achieved excellent performance, discerning various forms of intrusion with 100% accuracy on training data and 99.71% accuracy on test data. Additionally, it achieved a minimal latency rate of 1.6%, reduced controller overhead, and a minimal degradation rate of 2.1% leading to improved throughput performance when executed in real-time networks.

查看原文本刊更多论文

求助全文

约1分钟内获得全文求助全文

来源期刊

Optical Fiber Technology 工程技术-电信学

CiteScore

4.80

自引率

11.10%

发文量

327

审稿时长

63 days

期刊介绍： Innovations in optical fiber technology are revolutionizing world communications. Newly developed fiber amplifiers allow for direct transmission of high-speed signals over transcontinental distances without the need for electronic regeneration. Optical fibers find new applications in data processing. The impact of fiber materials, devices, and systems on communications in the coming decades will create an abundance of primary literature and the need for up-to-date reviews. Optical Fiber Technology: Materials, Devices, and Systems is a new cutting-edge journal designed to fill a need in this rapidly evolving field for speedy publication of regular length papers. Both theoretical and experimental papers on fiber materials, devices, and system performance evaluation and measurements are eligible, with emphasis on practical applications.