Pub Date : 2023-08-12DOI: 10.13052/jcsm2245-1439.1253
G. Du
With the wide application of 5G networks, many digital images must rely on networks for transmission. Traditional image encryption algorithms can no longer meet modern security requirements, and it is important to protect digital images in network transmission more securely. To address the shortcomings of traditional chaotic algorithms in image encryption, such as the strong randomness of image pixel replacement and time-consuming computations of image pixel iteration, we use a fractional-order Fourier transform to replace the image pixel matrix, a one-dimensional logistic chaos algorithm to reduce the problem of strong randomness of image pixels, and a sine chaos-based idea to optimize the diffusion algorithm to reduce the computational complexity. After encrypting a digital image in simulation experiments, we achieved better results through statistical analysis, adjacent pixel correlation, and resistance to differential attack performance analysis index tests, and verified the protection effect of this algorithm in digital images during network attacks.
{"title":"Security of Encrypted Images in Network Transmission Based on an Improved Chaos Algorithm","authors":"G. Du","doi":"10.13052/jcsm2245-1439.1253","DOIUrl":"https://doi.org/10.13052/jcsm2245-1439.1253","url":null,"abstract":"With the wide application of 5G networks, many digital images must rely on networks for transmission. Traditional image encryption algorithms can no longer meet modern security requirements, and it is important to protect digital images in network transmission more securely. To address the shortcomings of traditional chaotic algorithms in image encryption, such as the strong randomness of image pixel replacement and time-consuming computations of image pixel iteration, we use a fractional-order Fourier transform to replace the image pixel matrix, a one-dimensional logistic chaos algorithm to reduce the problem of strong randomness of image pixels, and a sine chaos-based idea to optimize the diffusion algorithm to reduce the computational complexity. After encrypting a digital image in simulation experiments, we achieved better results through statistical analysis, adjacent pixel correlation, and resistance to differential attack performance analysis index tests, and verified the protection effect of this algorithm in digital images during network attacks.","PeriodicalId":37820,"journal":{"name":"Journal of Cyber Security and Mobility","volume":"45 1","pages":"675-696"},"PeriodicalIF":0.0,"publicationDate":"2023-08-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"83150854","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}
Pub Date : 2023-08-12DOI: 10.13052/jcsm2245-1439.1256
Min Chen
At present, most cameras use internal networks and use methods such as Traceroute for security protection, which cannot meet the requirements of camera network mapping. Therefore, a camera mapping scheme of network information and physical distribution is proposed. Firstly, the network topology problem of video content information collection was analyzed. This paper uses the mapping relationship between network space and physical space to propose the subnet division conjecture method and complete the preliminary mapping of the network through video data screening. Considering the insufficient coverage of topology mapping, a judgment and inference method based on Bayesian classification technology and network information is proposed, and the results are corrected and evaluated through the test of Jackard coefficient. In the preliminary network topology performance test, two state-of-the-art schemes are selected for experimental comparison. When the number of nodes in the proposed scheme is 5, 25, and 50, the mapping can be completed in the shortest time, and the accuracy reaches 80%. However, the surveying and mapping accuracy of the proposed scheme in the preliminary test is low, and the network information method is used for data screening. In the final surveying and mapping performance test, when the number of nodes is 40, the accuracy of the proposed scheme is 96%, which is better than previously proposed schemes, while the testing delay time is shorter. The technology proposed in the study has the best overall performance. It can effectively solve the problem of intranet surveying and mapping and has important reference value for the security protection of the camera network.
{"title":"Camera Network Topology Mapping Based on the Integration of Network Information and Physical Distribution Under the Background of Communication Security","authors":"Min Chen","doi":"10.13052/jcsm2245-1439.1256","DOIUrl":"https://doi.org/10.13052/jcsm2245-1439.1256","url":null,"abstract":"At present, most cameras use internal networks and use methods such as Traceroute for security protection, which cannot meet the requirements of camera network mapping. Therefore, a camera mapping scheme of network information and physical distribution is proposed. Firstly, the network topology problem of video content information collection was analyzed. This paper uses the mapping relationship between network space and physical space to propose the subnet division conjecture method and complete the preliminary mapping of the network through video data screening. Considering the insufficient coverage of topology mapping, a judgment and inference method based on Bayesian classification technology and network information is proposed, and the results are corrected and evaluated through the test of Jackard coefficient. In the preliminary network topology performance test, two state-of-the-art schemes are selected for experimental comparison. When the number of nodes in the proposed scheme is 5, 25, and 50, the mapping can be completed in the shortest time, and the accuracy reaches 80%. However, the surveying and mapping accuracy of the proposed scheme in the preliminary test is low, and the network information method is used for data screening. In the final surveying and mapping performance test, when the number of nodes is 40, the accuracy of the proposed scheme is 96%, which is better than previously proposed schemes, while the testing delay time is shorter. The technology proposed in the study has the best overall performance. It can effectively solve the problem of intranet surveying and mapping and has important reference value for the security protection of the camera network.","PeriodicalId":37820,"journal":{"name":"Journal of Cyber Security and Mobility","volume":"79 1","pages":"733-756"},"PeriodicalIF":0.0,"publicationDate":"2023-08-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"85155661","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}
Pub Date : 2023-08-12DOI: 10.13052/jcsm2245-1439.1258
Fadhil Mohammed Salman, Ahssan Ahmed Mohammed, Fanar Ali Joda
Wireless mesh networks have recently presented a promising environment for many researchers to develop large-scale wireless communication. Traffic in WMNs often suffers from congestion due to heavy traffic load’s saturation of certain routes. Therefore, this article proposes an efficient approach for congestion awareness and load balancing in WMNs, based on the Ant Colony Optimization (ACO) approach. The proposed approach aims to raise the performance of the WMN by distributing the traffic load between optimal routes and avoiding severe traffic congestion. The proposed approach relies on three basic mechanisms: detection of severe congestion within the ideal paths used for data transmission, creation of ideal secondary paths with updated pheromone values, and distribution of the traffic load (data packet flow) between the primary and secondary ideal paths. According to the results of the NS2 simulator, the suggested approach increased the WMN throughput by 14.8% when compared to the CACO approach and by 37% when employing the WCETT approach. The results also showed that the proposed approach achieved an average end-to-end delay closing of 0.0562, while WCETT and CACO approaches achieved an average end-to-end delay close to 0.1021 and 0.0976, respectively. The results indicated that the proposed approach achieved a lower percentage of dropped packets by 6.97% and 0.99% compared to the WCETT and CACO approaches. Thus, the proposed approach is effective in improving the performance of WMNs.
{"title":"Adaptation of the Ant Colony Algorithm to Avoid Congestion in Wireless Mesh Networks","authors":"Fadhil Mohammed Salman, Ahssan Ahmed Mohammed, Fanar Ali Joda","doi":"10.13052/jcsm2245-1439.1258","DOIUrl":"https://doi.org/10.13052/jcsm2245-1439.1258","url":null,"abstract":"Wireless mesh networks have recently presented a promising environment for many researchers to develop large-scale wireless communication. Traffic in WMNs often suffers from congestion due to heavy traffic load’s saturation of certain routes. Therefore, this article proposes an efficient approach for congestion awareness and load balancing in WMNs, based on the Ant Colony Optimization (ACO) approach. The proposed approach aims to raise the performance of the WMN by distributing the traffic load between optimal routes and avoiding severe traffic congestion. The proposed approach relies on three basic mechanisms: detection of severe congestion within the ideal paths used for data transmission, creation of ideal secondary paths with updated pheromone values, and distribution of the traffic load (data packet flow) between the primary and secondary ideal paths. According to the results of the NS2 simulator, the suggested approach increased the WMN throughput by 14.8% when compared to the CACO approach and by 37% when employing the WCETT approach. The results also showed that the proposed approach achieved an average end-to-end delay closing of 0.0562, while WCETT and CACO approaches achieved an average end-to-end delay close to 0.1021 and 0.0976, respectively. The results indicated that the proposed approach achieved a lower percentage of dropped packets by 6.97% and 0.99% compared to the WCETT and CACO approaches. Thus, the proposed approach is effective in improving the performance of WMNs.","PeriodicalId":37820,"journal":{"name":"Journal of Cyber Security and Mobility","volume":"148 1","pages":"785-812"},"PeriodicalIF":0.0,"publicationDate":"2023-08-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"86076916","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}
Pub Date : 2023-08-12DOI: 10.13052/jcsm2245-1439.1254
Ying Xue
The existence of vulnerabilities is a serious threat to the security of networks, which needs to be detected timely. In this paper, machine learning methods were mainly studied. Firstly, network security vulnerabilities were briefly introduced, and then a Convolutional Neural Network (CNN) + Long Short-Term Memory (LSTM) method was designed to extract and match vulnerability features by preprocessing vulnerability data based on National Vulnerability Database. It was found that the CNN-LSTM method had high training accuracy, and its recall rate, precision, F1, and Mathews correlation coefficient (MCC) values were better than those of support vector machine and other methods in detecting the test set; its F1 and MCC values reached 0.8807 and 0.9738, respectively; the F1 value was above 0.85 in detecting different categories of vulnerabilities. The results demonstrate the reliability of the CNN-LSTM method for vulnerability detection. The CNN-LSTM method can be applied to real networks.
{"title":"Machine Learning: Research on Detection of Network Security Vulnerabilities by Extracting and Matching Features","authors":"Ying Xue","doi":"10.13052/jcsm2245-1439.1254","DOIUrl":"https://doi.org/10.13052/jcsm2245-1439.1254","url":null,"abstract":"The existence of vulnerabilities is a serious threat to the security of networks, which needs to be detected timely. In this paper, machine learning methods were mainly studied. Firstly, network security vulnerabilities were briefly introduced, and then a Convolutional Neural Network (CNN) + Long Short-Term Memory (LSTM) method was designed to extract and match vulnerability features by preprocessing vulnerability data based on National Vulnerability Database. It was found that the CNN-LSTM method had high training accuracy, and its recall rate, precision, F1, and Mathews correlation coefficient (MCC) values were better than those of support vector machine and other methods in detecting the test set; its F1 and MCC values reached 0.8807 and 0.9738, respectively; the F1 value was above 0.85 in detecting different categories of vulnerabilities. The results demonstrate the reliability of the CNN-LSTM method for vulnerability detection. The CNN-LSTM method can be applied to real networks.","PeriodicalId":37820,"journal":{"name":"Journal of Cyber Security and Mobility","volume":"32 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-08-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134977357","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}
Pub Date : 2023-08-12DOI: 10.13052/jcsm2245-1439.1255
Xingfeng Li
With more and more control systems accessing computer networks, the increase in their associated vulnerabilities has led to a decreasing security evaluation of the networks. It is essential to secure computer networks from attacks. To this end, the study constructs a network of computer network security evaluation model based on an optimized circular algorithm. To avoid detecting the model’s parameters falling into the local optimum, the model is first optimized based on the Corsi grey wolf optimization (CGWO) algorithm for the hyperparameters of the Gaussian process (GP). To solve the problem of unbalanced data and the GP not having memory capability, the study proposes an optimized Gaussian Mixture Model-Recurrent neural networks (GMM-RNN) algorithm. Experimental results of attack type recognition accuracy showed that the research CGWO-GP algorithm can jump out of the local optimum, and its average value of accuracy reached 98.99%. The average value of the leakage rate was 0.42%, and the average value of the false alarm rate was 0.11%. The average detection accuracy of the GMM-RNN model for eight attack types was 95.899%. The optimal detection accuracy of this model performance was 96.3948%. The training time of the GMM-RNN model was 67.96 s, and the detection time of the test set was 6.45 s, which greatly optimized the real-time performance. The GMM-RNN model was more effective in predicting the security posture of computer networks, and the prediction value can reach 97.65%. The research model was significantly better than other algorithmic models in the performance and evaluation of computer network security and had certain research values.
{"title":"An Evaluation Model for Network Security Based on an Optimized Circular Algorithm","authors":"Xingfeng Li","doi":"10.13052/jcsm2245-1439.1255","DOIUrl":"https://doi.org/10.13052/jcsm2245-1439.1255","url":null,"abstract":"With more and more control systems accessing computer networks, the increase in their associated vulnerabilities has led to a decreasing security evaluation of the networks. It is essential to secure computer networks from attacks. To this end, the study constructs a network of computer network security evaluation model based on an optimized circular algorithm. To avoid detecting the model’s parameters falling into the local optimum, the model is first optimized based on the Corsi grey wolf optimization (CGWO) algorithm for the hyperparameters of the Gaussian process (GP). To solve the problem of unbalanced data and the GP not having memory capability, the study proposes an optimized Gaussian Mixture Model-Recurrent neural networks (GMM-RNN) algorithm. Experimental results of attack type recognition accuracy showed that the research CGWO-GP algorithm can jump out of the local optimum, and its average value of accuracy reached 98.99%. The average value of the leakage rate was 0.42%, and the average value of the false alarm rate was 0.11%. The average detection accuracy of the GMM-RNN model for eight attack types was 95.899%. The optimal detection accuracy of this model performance was 96.3948%. The training time of the GMM-RNN model was 67.96 s, and the detection time of the test set was 6.45 s, which greatly optimized the real-time performance. The GMM-RNN model was more effective in predicting the security posture of computer networks, and the prediction value can reach 97.65%. The research model was significantly better than other algorithmic models in the performance and evaluation of computer network security and had certain research values.","PeriodicalId":37820,"journal":{"name":"Journal of Cyber Security and Mobility","volume":"79 1","pages":"711-732"},"PeriodicalIF":0.0,"publicationDate":"2023-08-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"79299595","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}
Pub Date : 2023-06-30DOI: 10.13052/jcsm2245-1439.1246
S. Alouneh
Domain Name System (DNS) is considered a vital service for the internet and networks operations, and practically this service is configured and accessible across networks’ firewall. Therefore, attackers take advantage of this open configuration to attack a network’s DNS server in order to use it as a reflector to achieve Denial of Service (DoS) attacks. Most of protection methods such as intrusion prevention and detection systems use blended tactics such as blocked-lists for suspicious sources, and thresholds for traffic volumes to detect and defend against DoS flooding attacks. However, these protection methods are not often successful. In this paper, we propose a new method to sense and protect DNS systems from DoS and Distributed DoS (DDoS) attacks. The main idea in our approach is to distribute the DNS request mapping into more than one DNS resolver such that an attack on one server should not affect the entire DNS services. Our approach uses the Multi-Protocol Label Switching (MPLS) along with multi-path routing to achieve this goal. Also, we use threshold secret sharing to code the distributed DNS requests. Our findings and results show that this approach performs better when compared with the traditional DNS structure.
{"title":"A Multi-Path Approach to Protect DNS Against DDoS Attacks","authors":"S. Alouneh","doi":"10.13052/jcsm2245-1439.1246","DOIUrl":"https://doi.org/10.13052/jcsm2245-1439.1246","url":null,"abstract":"Domain Name System (DNS) is considered a vital service for the internet and networks operations, and practically this service is configured and accessible across networks’ firewall. Therefore, attackers take advantage of this open configuration to attack a network’s DNS server in order to use it as a reflector to achieve Denial of Service (DoS) attacks. Most of protection methods such as intrusion prevention and detection systems use blended tactics such as blocked-lists for suspicious sources, and thresholds for traffic volumes to detect and defend against DoS flooding attacks. However, these protection methods are not often successful. In this paper, we propose a new method to sense and protect DNS systems from DoS and Distributed DoS (DDoS) attacks. The main idea in our approach is to distribute the DNS request mapping into more than one DNS resolver such that an attack on one server should not affect the entire DNS services. Our approach uses the Multi-Protocol Label Switching (MPLS) along with multi-path routing to achieve this goal. Also, we use threshold secret sharing to code the distributed DNS requests. Our findings and results show that this approach performs better when compared with the traditional DNS structure.","PeriodicalId":37820,"journal":{"name":"Journal of Cyber Security and Mobility","volume":"9 1","pages":"569-588"},"PeriodicalIF":0.0,"publicationDate":"2023-06-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"83524505","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}
Pub Date : 2023-06-30DOI: 10.13052/jcsm2245-1439.1241
Ali Akram Abdul-Kareem, W. A. M. Al-Jawher
Image encryption is one of the most important techniques to maintain data confidentiality against illegal access and fraudulent usage. In this study, a new medical image encryption technique was developed by combining the discrete wavelet transform, the fast Fourier transform, the Arnold transform, and two multidimensional chaotic systems. The medical image is subjected to a discrete wavelet transform before the magic square shuffles the image sub-bands. Confusion operations are performed on each scrambled subdomain using the Uruk 4D chaotic system. To increase randomness and unpredictability, a second stage of confusion is implemented in the domain of the Fast Fourier transform using the Arnold transform. The final encrypted image is obtained utilizing secret keys derived from the WAM 3D chaotic system. In particular, the initial conditions for chaotic systems are derived from grayscale values, thereby increasing the algorithm’s sensitivity to the input image. The results of the tests and the security analysis indicate that the proposed algorithm is exceptionally reliable and secure.
{"title":"A Hybrid Domain Medical Image Encryption Scheme Using URUK and WAM Chaotic Maps with Wavelet - Fourier Transforms","authors":"Ali Akram Abdul-Kareem, W. A. M. Al-Jawher","doi":"10.13052/jcsm2245-1439.1241","DOIUrl":"https://doi.org/10.13052/jcsm2245-1439.1241","url":null,"abstract":"Image encryption is one of the most important techniques to maintain data confidentiality against illegal access and fraudulent usage. In this study, a new medical image encryption technique was developed by combining the discrete wavelet transform, the fast Fourier transform, the Arnold transform, and two multidimensional chaotic systems. The medical image is subjected to a discrete wavelet transform before the magic square shuffles the image sub-bands. Confusion operations are performed on each scrambled subdomain using the Uruk 4D chaotic system. To increase randomness and unpredictability, a second stage of confusion is implemented in the domain of the Fast Fourier transform using the Arnold transform. The final encrypted image is obtained utilizing secret keys derived from the WAM 3D chaotic system. In particular, the initial conditions for chaotic systems are derived from grayscale values, thereby increasing the algorithm’s sensitivity to the input image. The results of the tests and the security analysis indicate that the proposed algorithm is exceptionally reliable and secure.","PeriodicalId":37820,"journal":{"name":"Journal of Cyber Security and Mobility","volume":"34 1","pages":"435-464"},"PeriodicalIF":0.0,"publicationDate":"2023-06-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"87242718","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}
Pub Date : 2023-06-30DOI: 10.13052/jcsm2245-1439.1243
Julija Gavėnaitė-Sirvydienė, Algita Miečinskienė
Constantly evolving high technologies provide new approaches to business development and deliver unknown business risks. Online financial services and operations are integral to everyday life, making cyber risk one of the most relevant risks for the financial sector’s companies. As the survey conducted by the National Bank of Lithuania at the end of 2018 showed, the possibility of cyber threats and presumable effects on the financial system in Lithuania is one of the critical problems that should be prioritized. Therefore, it is essential to clarify what potential cyber threats in financial sector companies are considered the most significant and likely to occur. As well as identify how companies in the financial sector estimate their dispositions and preparedness for this cyber risk management and control. The findings of this research are significant for financial institutions as a tool to adopt their cyber risk management processes, increase preparedness and cyber security, and identify the possible threats to the organization.
{"title":"The Assessment of Cyber Security's Significance in the Financial Sector of Lithuania","authors":"Julija Gavėnaitė-Sirvydienė, Algita Miečinskienė","doi":"10.13052/jcsm2245-1439.1243","DOIUrl":"https://doi.org/10.13052/jcsm2245-1439.1243","url":null,"abstract":"Constantly evolving high technologies provide new approaches to business development and deliver unknown business risks. Online financial services and operations are integral to everyday life, making cyber risk one of the most relevant risks for the financial sector’s companies. As the survey conducted by the National Bank of Lithuania at the end of 2018 showed, the possibility of cyber threats and presumable effects on the financial system in Lithuania is one of the critical problems that should be prioritized. Therefore, it is essential to clarify what potential cyber threats in financial sector companies are considered the most significant and likely to occur. As well as identify how companies in the financial sector estimate their dispositions and preparedness for this cyber risk management and control. The findings of this research are significant for financial institutions as a tool to adopt their cyber risk management processes, increase preparedness and cyber security, and identify the possible threats to the organization.","PeriodicalId":37820,"journal":{"name":"Journal of Cyber Security and Mobility","volume":"34 1","pages":"497-518"},"PeriodicalIF":0.0,"publicationDate":"2023-06-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"84715261","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}
Pub Date : 2023-06-30DOI: 10.13052/jcsm2245-1439.1242
Nisha P. Shetty, Balachandra Muniyal, Aman Priyanshu, Vedant Rishi Das
Cyberbullying has become one of the most pressing concerns for online platforms, putting individuals at risk and raising severe public concerns. Recent studies have shown a significant correlation between declining mental health and cyberbullying. Automated detection offers a great solution to this problem; however, the sensitivity of client-data becomes a concern during data collection, and as such, access may be restricted. This paper demonstrates FedBully, a federated approach for cyberbullying detection using sentence encoders for feature extraction. This paper introduces concepts of secure aggregation to ensure client privacy in a cross-device learning system. Optimal hyper-parameters were studied through comprehensive experiments, and a computationally and communicationally inexpensive network is proposed. Experiments reveal promising results with up to 93% classification AUC (Area Under the Curve) using only dense networks to fine-tune sentence embeddings on IID datasets and 91% AUC on non-IID datasets, where IID refers to Independent and Identically Distributed data. The analysis also shows that data independence profoundly impacts network performance, with AUC decreasing by a mean of 5.1% between Non-IID and IID. A rich and extensive study has also been performed on client network size and secure aggregation protocols, which prove the robustness and practicality of the proposed model. The novel approach presented offers an efficient and practical solution to training a cross-device cyberbullying detector while ensuring client-privacy.
{"title":"FedBully: A Cross-Device Federated Approach for Privacy Enabled Cyber Bullying Detection using Sentence Encoders","authors":"Nisha P. Shetty, Balachandra Muniyal, Aman Priyanshu, Vedant Rishi Das","doi":"10.13052/jcsm2245-1439.1242","DOIUrl":"https://doi.org/10.13052/jcsm2245-1439.1242","url":null,"abstract":"Cyberbullying has become one of the most pressing concerns for online platforms, putting individuals at risk and raising severe public concerns. Recent studies have shown a significant correlation between declining mental health and cyberbullying. Automated detection offers a great solution to this problem; however, the sensitivity of client-data becomes a concern during data collection, and as such, access may be restricted. This paper demonstrates FedBully, a federated approach for cyberbullying detection using sentence encoders for feature extraction. This paper introduces concepts of secure aggregation to ensure client privacy in a cross-device learning system. Optimal hyper-parameters were studied through comprehensive experiments, and a computationally and communicationally inexpensive network is proposed. Experiments reveal promising results with up to 93% classification AUC (Area Under the Curve) using only dense networks to fine-tune sentence embeddings on IID datasets and 91% AUC on non-IID datasets, where IID refers to Independent and Identically Distributed data. The analysis also shows that data independence profoundly impacts network performance, with AUC decreasing by a mean of 5.1% between Non-IID and IID. A rich and extensive study has also been performed on client network size and secure aggregation protocols, which prove the robustness and practicality of the proposed model. The novel approach presented offers an efficient and practical solution to training a cross-device cyberbullying detector while ensuring client-privacy.","PeriodicalId":37820,"journal":{"name":"Journal of Cyber Security and Mobility","volume":"37 1","pages":"465-496"},"PeriodicalIF":0.0,"publicationDate":"2023-06-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"81264947","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}
Pub Date : 2023-06-30DOI: 10.13052/jcsm2245-1439.1245
Liu Zhang, Yanyu Liu
Technology development has promoted network construction, but malicious network attacks are still inevitable. To solve the problem that the current network security assessment is not practical and the assessment effect is poor, this study proposes a network security monitoring tool based on situation assessment and prediction to assist network security construction. The framework of the evaluation module is based on convolution neural network. The initial module is introduced to convert some large convolution cores into small convolution cores in series. This is to reduce the operating cost, because building multiple evaluators in series can maximize the retention of characteristic values. This module is the optimized form of Elman neural network. The delay operator is added to the model to respond to the time property of network attack. At the same time, particle swarm optimization algorithm is used to solve the initial weight dependence problem. The research adopts two methods of security situation assessment and situation prediction to carry out model application test. During the test, the commonly used KDD Cup99 is used as intrusion detection data. The experimental results of the network security situation evaluation module show that the optimization reduces the evaluation error by 3.34%, and the accuracy meets the evaluation requirements. The model is superior to the back propagation neural network and the standard Elman model. The model proposed in this study achieves better prediction of posture scores from 0.3 to 0.9, which is more stable than BP neural network. It proves that the model designed by the research can achieve more stable and higher prediction than similar models. It is more practical to obtain better results on the basis of a more stable model architecture and lower implementation costs, which is a meaningful attempt in the wide application of network security.
{"title":"Network Security Prediction and Situational Assessment Using Neural Network-based Method","authors":"Liu Zhang, Yanyu Liu","doi":"10.13052/jcsm2245-1439.1245","DOIUrl":"https://doi.org/10.13052/jcsm2245-1439.1245","url":null,"abstract":"Technology development has promoted network construction, but malicious network attacks are still inevitable. To solve the problem that the current network security assessment is not practical and the assessment effect is poor, this study proposes a network security monitoring tool based on situation assessment and prediction to assist network security construction. The framework of the evaluation module is based on convolution neural network. The initial module is introduced to convert some large convolution cores into small convolution cores in series. This is to reduce the operating cost, because building multiple evaluators in series can maximize the retention of characteristic values. This module is the optimized form of Elman neural network. The delay operator is added to the model to respond to the time property of network attack. At the same time, particle swarm optimization algorithm is used to solve the initial weight dependence problem. The research adopts two methods of security situation assessment and situation prediction to carry out model application test. During the test, the commonly used KDD Cup99 is used as intrusion detection data. The experimental results of the network security situation evaluation module show that the optimization reduces the evaluation error by 3.34%, and the accuracy meets the evaluation requirements. The model is superior to the back propagation neural network and the standard Elman model. The model proposed in this study achieves better prediction of posture scores from 0.3 to 0.9, which is more stable than BP neural network. It proves that the model designed by the research can achieve more stable and higher prediction than similar models. It is more practical to obtain better results on the basis of a more stable model architecture and lower implementation costs, which is a meaningful attempt in the wide application of network security.","PeriodicalId":37820,"journal":{"name":"Journal of Cyber Security and Mobility","volume":"24 1","pages":"547-568"},"PeriodicalIF":0.0,"publicationDate":"2023-06-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"77926762","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：