An Intrusion Detection System (IDS) is essential to identify cyber-attacks and implement appropriate measures for each risk. The efficiency of the Machine Learning (ML) techniques is compromised in the presence of irrelevant features and class imbalance. In this research, an efficient data pre-processing strategy was proposed to enhance the model’s generalizability. The class dissimilarity is addressed using k-Means SMOTE. After this, we furnish a hybrid feature selection method that combines filters and wrappers. Further, a hyperparameter-tuned Light Gradient Boosting Machine (LGBM) is analyzed by varying the optimal feature subsets. The experiments used the datasets – UNSW-NB15 and CICIDS-2017, yielding an accuracy of 90.71% and 99.98%, respectively. As the transparency and generalizability of the model depend significantly on understanding each component of the prediction, we employed the eXplainable Artificial Intelligence (XAI) method, SHapley Additive exPlanation (SHAP), to improve the comprehension of forecasted results.
{"title":"Enhancing Intrusion Detection with Explainable AI: A Transparent Approach to Network Security","authors":"Seshu Bhavani Mallampati, Hari Seetha","doi":"10.2478/cait-2024-0006","DOIUrl":"https://doi.org/10.2478/cait-2024-0006","url":null,"abstract":"\u0000 An Intrusion Detection System (IDS) is essential to identify cyber-attacks and implement appropriate measures for each risk. The efficiency of the Machine Learning (ML) techniques is compromised in the presence of irrelevant features and class imbalance. In this research, an efficient data pre-processing strategy was proposed to enhance the model’s generalizability. The class dissimilarity is addressed using k-Means SMOTE. After this, we furnish a hybrid feature selection method that combines filters and wrappers. Further, a hyperparameter-tuned Light Gradient Boosting Machine (LGBM) is analyzed by varying the optimal feature subsets. The experiments used the datasets – UNSW-NB15 and CICIDS-2017, yielding an accuracy of 90.71% and 99.98%, respectively. As the transparency and generalizability of the model depend significantly on understanding each component of the prediction, we employed the eXplainable Artificial Intelligence (XAI) method, SHapley Additive exPlanation (SHAP), to improve the comprehension of forecasted results.","PeriodicalId":45562,"journal":{"name":"Cybernetics and Information Technologies","volume":null,"pages":null},"PeriodicalIF":1.2,"publicationDate":"2024-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140404468","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Abstract The research explores applying hierarchical clustering methods, namely single linkage and complete linkage, in IoT Sensor Networks (ISNs). ISNs are distributed systems comprising numerous sensor nodes that collect data from the environment and communicate with each other to transmit the data to a base station. Hierarchical clustering is a technique that groups nodes into clusters based on proximity and similarity. This paper implements and compares the performance of single linkage and complete linkage methods in terms of cluster size, network lifetime, and cluster quality. The study’s findings provide guidance for ISN researchers and designers in selecting the appropriate clustering method that meets their specific requirements.
{"title":"Exploring the Performance and Characteristics of Single Linkage and Complete Linkage Hierarchical Clustering Methods for IoT Sensor Networks","authors":"Fuad Bajaber","doi":"10.2478/cait-2023-0041","DOIUrl":"https://doi.org/10.2478/cait-2023-0041","url":null,"abstract":"Abstract The research explores applying hierarchical clustering methods, namely single linkage and complete linkage, in IoT Sensor Networks (ISNs). ISNs are distributed systems comprising numerous sensor nodes that collect data from the environment and communicate with each other to transmit the data to a base station. Hierarchical clustering is a technique that groups nodes into clusters based on proximity and similarity. This paper implements and compares the performance of single linkage and complete linkage methods in terms of cluster size, network lifetime, and cluster quality. The study’s findings provide guidance for ISN researchers and designers in selecting the appropriate clustering method that meets their specific requirements.","PeriodicalId":45562,"journal":{"name":"Cybernetics and Information Technologies","volume":null,"pages":null},"PeriodicalIF":1.2,"publicationDate":"2023-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139298587","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Muhammad Arham Tariq, Allah Bux Sargano, Muhammad Aksam Iftikhar, Z. Habib
Abstract Predicting students’ academic performance is a critical research area, yet imbalanced educational datasets, characterized by unequal academic-level representation, present challenges for classifiers. While prior research has addressed the imbalance in binary-class datasets, this study focuses on multi-class datasets. A comparison of ten resampling methods (SMOTE, Adasyn, Distance SMOTE, BorderLineSMOTE, KmeansSMOTE, SVMSMOTE, LN SMOTE, MWSMOTE, Safe Level SMOTE, and SMOTETomek) is conducted alongside nine classification models: K-Nearest Neighbors (KNN), Linear Discriminant Analysis (LDA), Quadratic Discriminant Analysis (QDA), Support Vector Machine (SVM), Logistic Regression (LR), Extra Tree (ET), Random Forest (RT), Extreme Gradient Boosting (XGB), and Ada Boost (AdaB). Following a rigorous evaluation, including hyperparameter tuning and 10 fold cross-validations, KNN with SmoteTomek attains the highest accuracy of 83.7%, as demonstrated through an ablation study. These results emphasize SMOTETomek’s effectiveness in mitigating class imbalance in educational datasets and highlight KNN’s potential as an educational data mining classifier.
{"title":"Comparing Different Oversampling Methods in Predicting Multi-Class Educational Datasets Using Machine Learning Techniques","authors":"Muhammad Arham Tariq, Allah Bux Sargano, Muhammad Aksam Iftikhar, Z. Habib","doi":"10.2478/cait-2023-0044","DOIUrl":"https://doi.org/10.2478/cait-2023-0044","url":null,"abstract":"Abstract Predicting students’ academic performance is a critical research area, yet imbalanced educational datasets, characterized by unequal academic-level representation, present challenges for classifiers. While prior research has addressed the imbalance in binary-class datasets, this study focuses on multi-class datasets. A comparison of ten resampling methods (SMOTE, Adasyn, Distance SMOTE, BorderLineSMOTE, KmeansSMOTE, SVMSMOTE, LN SMOTE, MWSMOTE, Safe Level SMOTE, and SMOTETomek) is conducted alongside nine classification models: K-Nearest Neighbors (KNN), Linear Discriminant Analysis (LDA), Quadratic Discriminant Analysis (QDA), Support Vector Machine (SVM), Logistic Regression (LR), Extra Tree (ET), Random Forest (RT), Extreme Gradient Boosting (XGB), and Ada Boost (AdaB). Following a rigorous evaluation, including hyperparameter tuning and 10 fold cross-validations, KNN with SmoteTomek attains the highest accuracy of 83.7%, as demonstrated through an ablation study. These results emphasize SMOTETomek’s effectiveness in mitigating class imbalance in educational datasets and highlight KNN’s potential as an educational data mining classifier.","PeriodicalId":45562,"journal":{"name":"Cybernetics and Information Technologies","volume":null,"pages":null},"PeriodicalIF":1.2,"publicationDate":"2023-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139294540","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
J. Zraqou, Adnan H. Al-Helali, Waleed Maqableh, H. Fakhouri, Wesam Alkhadour
Abstract Effective spam filtering plays a crucial role in enhancing user experience by sparing them from unwanted messages. This imperative underscores the importance of safeguarding email systems, prompting scholars across diverse fields to delve deeper into this subject. The primary objective of this research is to mitigate the disruptive effects of spam on email usage by introducing improved security measures compared to existing methods. This goal can be accomplished through the development of a novel spam filtering technique designed to prevent spam from infiltrating users’ inboxes. Consequently, a hybrid filtering approach that combines an information gain philter and a Wrapper Grey Wolf Optimizer feature selection algorithm with a Naive Bayes Classifier, is proposed, denoted as GWO-NBC. This research is rigorously tested using the WEKA software and the SPAMBASE dataset. Thorough performance evaluations demonstrated that the proposed approach surpasses existing solutions in terms of both security and accuracy.
{"title":"Robust Email Spam Filtering Using a Hybrid of Grey Wolf Optimiser and Naive Bayes Classifier","authors":"J. Zraqou, Adnan H. Al-Helali, Waleed Maqableh, H. Fakhouri, Wesam Alkhadour","doi":"10.2478/cait-2023-0037","DOIUrl":"https://doi.org/10.2478/cait-2023-0037","url":null,"abstract":"Abstract Effective spam filtering plays a crucial role in enhancing user experience by sparing them from unwanted messages. This imperative underscores the importance of safeguarding email systems, prompting scholars across diverse fields to delve deeper into this subject. The primary objective of this research is to mitigate the disruptive effects of spam on email usage by introducing improved security measures compared to existing methods. This goal can be accomplished through the development of a novel spam filtering technique designed to prevent spam from infiltrating users’ inboxes. Consequently, a hybrid filtering approach that combines an information gain philter and a Wrapper Grey Wolf Optimizer feature selection algorithm with a Naive Bayes Classifier, is proposed, denoted as GWO-NBC. This research is rigorously tested using the WEKA software and the SPAMBASE dataset. Thorough performance evaluations demonstrated that the proposed approach surpasses existing solutions in terms of both security and accuracy.","PeriodicalId":45562,"journal":{"name":"Cybernetics and Information Technologies","volume":null,"pages":null},"PeriodicalIF":1.2,"publicationDate":"2023-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139305594","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Abstract The occurrence of large-scale crises is a great challenge for people. In such cases, many levels of public life are affected and recovery takes time and considerable resources. Therefore, approaches and tools for predicting and preventing crises, as well as models and methods for crisis management and crisis overcoming, are necessary. In this review, we present approaches, models, and methods that support decision-making in relation to the prevention and resolution of large-scale crises. We divide crises into three types: natural disasters, pandemics, and economic crises. For each type of crisis situation, the types of applied tasks that are solved and the corresponding models and methods that are used to support decision-makers in overcoming the crises are discussed. Conclusions are drawn on the state of the art in this area and some directions for future work are outlined.
{"title":"Optimization Models and Strategy Approaches Dealing with Economic Crises, Natural Disasters, and Pandemics – An Overview","authors":"V. Guliashki, L. Kirilov, Alsa B. Nuzi","doi":"10.2478/cait-2023-0033","DOIUrl":"https://doi.org/10.2478/cait-2023-0033","url":null,"abstract":"Abstract The occurrence of large-scale crises is a great challenge for people. In such cases, many levels of public life are affected and recovery takes time and considerable resources. Therefore, approaches and tools for predicting and preventing crises, as well as models and methods for crisis management and crisis overcoming, are necessary. In this review, we present approaches, models, and methods that support decision-making in relation to the prevention and resolution of large-scale crises. We divide crises into three types: natural disasters, pandemics, and economic crises. For each type of crisis situation, the types of applied tasks that are solved and the corresponding models and methods that are used to support decision-makers in overcoming the crises are discussed. Conclusions are drawn on the state of the art in this area and some directions for future work are outlined.","PeriodicalId":45562,"journal":{"name":"Cybernetics and Information Technologies","volume":null,"pages":null},"PeriodicalIF":1.2,"publicationDate":"2023-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139291759","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Abstract One of the methods most frequently used to recommend films is collaborative filtering. We examine the potential of collaborative filtering in our paper’s discussion of product suggestions. In addition to utilizing collaborative filtering in a new application, the proposed system will present a better technique that focuses especially on resolving the cold start issue. The suggested system will compute similarity using the Pearson Correlation Coefficient (PCC). Collaborative filtering that uses PCC suffers from the cold start problem or a lack of information on new users to generate useful recommendations. The proposed system solves the issue of cold start by gauging each new user by certain arbitrary parameters and recommending based on the choices of other users in that demographic. The proposed system also solves the issue of users’ reluctance to provide ratings by implementing a keyword-based perception system that will aid users in finding the right product for them.
{"title":"An Improved Product Recommender System Using Collaborative Filtering and a Comparative Study of ML Algorithms","authors":"S. Amutha, R. Vikram Surya","doi":"10.2478/cait-2023-0035","DOIUrl":"https://doi.org/10.2478/cait-2023-0035","url":null,"abstract":"Abstract One of the methods most frequently used to recommend films is collaborative filtering. We examine the potential of collaborative filtering in our paper’s discussion of product suggestions. In addition to utilizing collaborative filtering in a new application, the proposed system will present a better technique that focuses especially on resolving the cold start issue. The suggested system will compute similarity using the Pearson Correlation Coefficient (PCC). Collaborative filtering that uses PCC suffers from the cold start problem or a lack of information on new users to generate useful recommendations. The proposed system solves the issue of cold start by gauging each new user by certain arbitrary parameters and recommending based on the choices of other users in that demographic. The proposed system also solves the issue of users’ reluctance to provide ratings by implementing a keyword-based perception system that will aid users in finding the right product for them.","PeriodicalId":45562,"journal":{"name":"Cybernetics and Information Technologies","volume":null,"pages":null},"PeriodicalIF":1.2,"publicationDate":"2023-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139305308","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Abstract A startup is a recently established business venture led by entrepreneurs, to create and offer new products or services. The discovery of promising startups is a challenging task for creditors, policymakers, and investors. Therefore, the startup survival rate prediction is required to be developed for the success/failure of startup companies. In this paper, the feature selection using the Convex Least Angle Regression Least Absolute Shrinkage and Selection Operator (CLAR-LASSO) is proposed to improve the classification of startup survival rate prediction. The Swish Activation Function based Long Short-Term Memory (SAFLSTM) is developed for classifying the survival rate of startups. Further, the Local Interpretable Model-agnostic Explanations (LIME) model interprets the predicted classification to the user. Existing research such as Hyper Parameter Tuning (HPT)-Logistic regression, HPT-Support Vector Machine (SVM), HPT-XGBoost, and SAFLSTM are used to compare the CLAR-LASSO. The accuracy of the CLAR-LASSO is 95.67% which is high when compared to the HPT-Logistic regression, HPT-SVM, HPT-XGBoost, and SAFLSTM.
{"title":"Convex Least Angle Regression Based LASSO Feature Selection and Swish Activation Function Model for Startup Survival Rate","authors":"Ramakrishna Allu, V. N. R. Padmanabhuni","doi":"10.2478/cait-2023-0039","DOIUrl":"https://doi.org/10.2478/cait-2023-0039","url":null,"abstract":"Abstract A startup is a recently established business venture led by entrepreneurs, to create and offer new products or services. The discovery of promising startups is a challenging task for creditors, policymakers, and investors. Therefore, the startup survival rate prediction is required to be developed for the success/failure of startup companies. In this paper, the feature selection using the Convex Least Angle Regression Least Absolute Shrinkage and Selection Operator (CLAR-LASSO) is proposed to improve the classification of startup survival rate prediction. The Swish Activation Function based Long Short-Term Memory (SAFLSTM) is developed for classifying the survival rate of startups. Further, the Local Interpretable Model-agnostic Explanations (LIME) model interprets the predicted classification to the user. Existing research such as Hyper Parameter Tuning (HPT)-Logistic regression, HPT-Support Vector Machine (SVM), HPT-XGBoost, and SAFLSTM are used to compare the CLAR-LASSO. The accuracy of the CLAR-LASSO is 95.67% which is high when compared to the HPT-Logistic regression, HPT-SVM, HPT-XGBoost, and SAFLSTM.","PeriodicalId":45562,"journal":{"name":"Cybernetics and Information Technologies","volume":null,"pages":null},"PeriodicalIF":1.2,"publicationDate":"2023-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139295096","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Abstract Cognitive Radio Networks (CRNs) present a compelling possibility to enable secondary users to take advantage of unused frequency bands in constrained spectrum resources. However, the network is vulnerable to a wide range of jamming attacks, which adversely affect its performance. Several countermeasures proposed in the literature require prior knowledge of the communication network and jamming strategy that are computationally intensive. These solutions may not be suitable for many real-world critical applications of the Internet of Things (IoT). Therefore, a novel self-exploration approach based on deep reinforcement learning is proposed to learn an optimal policy against dynamic attacks in CRN-based IoT applications. This method reduces computational complexity, without prior knowledge of the communication network or jamming strategy. A simulation of the proposed scheme eliminates interference effectively, consumes less power, and has a better Signal-to-Noise Ratio (SNR) than other algorithms. A platform-agnostic and efficient anti-jamming solution is provided to improve CRN’s performance when jamming occurs.
{"title":"A Novel Self-Exploration Scheme for Learning Optimal Policies against Dynamic Jamming Attacks in Cognitive Radio Networks","authors":"Y. Sudha, V. Sarasvathi","doi":"10.2478/cait-2023-0040","DOIUrl":"https://doi.org/10.2478/cait-2023-0040","url":null,"abstract":"Abstract Cognitive Radio Networks (CRNs) present a compelling possibility to enable secondary users to take advantage of unused frequency bands in constrained spectrum resources. However, the network is vulnerable to a wide range of jamming attacks, which adversely affect its performance. Several countermeasures proposed in the literature require prior knowledge of the communication network and jamming strategy that are computationally intensive. These solutions may not be suitable for many real-world critical applications of the Internet of Things (IoT). Therefore, a novel self-exploration approach based on deep reinforcement learning is proposed to learn an optimal policy against dynamic attacks in CRN-based IoT applications. This method reduces computational complexity, without prior knowledge of the communication network or jamming strategy. A simulation of the proposed scheme eliminates interference effectively, consumes less power, and has a better Signal-to-Noise Ratio (SNR) than other algorithms. A platform-agnostic and efficient anti-jamming solution is provided to improve CRN’s performance when jamming occurs.","PeriodicalId":45562,"journal":{"name":"Cybernetics and Information Technologies","volume":null,"pages":null},"PeriodicalIF":1.2,"publicationDate":"2023-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139295169","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Abstract A novel proposed Binary Volleyball Premier League algorithm (BVPL) has shown some promising results in a Parkinson’s Disease (PD) dataset related to fitness and accuracy [1]. This paper evaluates and provides an overview of the efficiency of BVPL in feature selection compared to various metaheuristic optimization algorithms and PD datasets. Moreover, an improved variant of BVPL is proposed that integrates the opposite-based solution to enlarge search domains and increase the possibility of getting rid of the local optima. The performance of BVPL is validated using the accuracy of the k-Nearest Neighbor Algorithm. The superiority of BVPL over the competing algorithms for each dataset is measured using statistical tests. The conclusive results indicate that the BVPL exhibits significant competitiveness compared to most metaheuristic algorithms, thereby establishing its potential for accurate prediction of PD. Overall, BVPL shows high potential to be employed in feature selection.
{"title":"A Competitive Parkinson-Based Binary Volleyball Premier League Metaheuristic Algorithm for Feature Selection","authors":"Edjola K. Naka","doi":"10.2478/cait-2023-0038","DOIUrl":"https://doi.org/10.2478/cait-2023-0038","url":null,"abstract":"Abstract A novel proposed Binary Volleyball Premier League algorithm (BVPL) has shown some promising results in a Parkinson’s Disease (PD) dataset related to fitness and accuracy [1]. This paper evaluates and provides an overview of the efficiency of BVPL in feature selection compared to various metaheuristic optimization algorithms and PD datasets. Moreover, an improved variant of BVPL is proposed that integrates the opposite-based solution to enlarge search domains and increase the possibility of getting rid of the local optima. The performance of BVPL is validated using the accuracy of the k-Nearest Neighbor Algorithm. The superiority of BVPL over the competing algorithms for each dataset is measured using statistical tests. The conclusive results indicate that the BVPL exhibits significant competitiveness compared to most metaheuristic algorithms, thereby establishing its potential for accurate prediction of PD. Overall, BVPL shows high potential to be employed in feature selection.","PeriodicalId":45562,"journal":{"name":"Cybernetics and Information Technologies","volume":null,"pages":null},"PeriodicalIF":1.2,"publicationDate":"2023-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139298959","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Abstract The Denial of Service (DoS) attack threatens the availability of key components of Vehicular Ad-hoc Network (VANET). Various centralized and decentralized trust-based approaches have been proposed to secure the VANET from DoS attack. The centralized approach is less efficient because the attack on the central trust manager leads to the overall failure of services. In comparison, the cluster-based decentralized approach faces overhead because of frequent changes in cluster members due to the high speed of the vehicles. Therefore, we have proposed a cluster-based Denial-of-Service Resistant Trust model (DoSRT). It improves decentralized trust management using speed deviation-based clustering and detects DoS attack based on the frequency of messages sent. Through performance evaluation, we have found that DoSRT improves precision, recall, accuracy, and F-Score by around 19%, 16%, 20%, and 17% in the presence of 30% DoS attackers.
摘要 拒绝服务（DoS）攻击威胁着车载无线网络（VANET）关键组件的可用性。为了确保 VANET 免受 DoS 攻击，人们提出了各种基于信任的集中式和分散式方法。集中式方法的效率较低，因为对中央信任管理器的攻击会导致整体服务失效。相比之下，基于集群的分散式方法由于车辆的高速行驶而导致集群成员的频繁变化，因此面临着开销问题。因此，我们提出了一种基于集群的抗拒绝服务信任模型（DoSRT）。它利用基于速度偏差的聚类改进了分散式信任管理，并根据发送信息的频率检测 DoS 攻击。通过性能评估，我们发现在存在 30% DoS 攻击者的情况下，DoSRT 的精确度、召回率、准确度和 F-Score 分别提高了约 19%、16%、20% 和 17%。
{"title":"DoSRT: A Denial-of-Service Resistant Trust Model for VANET","authors":"Niharika Keshari, Dinesh Singh, Ashish Kumar Maurya","doi":"10.2478/cait-2023-0042","DOIUrl":"https://doi.org/10.2478/cait-2023-0042","url":null,"abstract":"Abstract The Denial of Service (DoS) attack threatens the availability of key components of Vehicular Ad-hoc Network (VANET). Various centralized and decentralized trust-based approaches have been proposed to secure the VANET from DoS attack. The centralized approach is less efficient because the attack on the central trust manager leads to the overall failure of services. In comparison, the cluster-based decentralized approach faces overhead because of frequent changes in cluster members due to the high speed of the vehicles. Therefore, we have proposed a cluster-based Denial-of-Service Resistant Trust model (DoSRT). It improves decentralized trust management using speed deviation-based clustering and detects DoS attack based on the frequency of messages sent. Through performance evaluation, we have found that DoSRT improves precision, recall, accuracy, and F-Score by around 19%, 16%, 20%, and 17% in the presence of 30% DoS attackers.","PeriodicalId":45562,"journal":{"name":"Cybernetics and Information Technologies","volume":null,"pages":null},"PeriodicalIF":1.2,"publicationDate":"2023-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139304633","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：