Pub Date : 2024-06-28DOI: 10.1109/TMLCN.2024.3420268
Pavlos Doanis;Thrasyvoulos Spyropoulos
Data-driven network slicing has been recently explored as a major driver for beyond 5G networks. Nevertheless, we are still a long way before such solutions are practically applicable in real problems. Most solutions addressing the problem of dynamically placing virtual network function chains (“slices”) on top of a physical topology still face one or more of the following hurdles: (i) they focus on simple slicing setups (e.g. single domain, single slice, simple VNF chains and performance metrics); (ii) solutions based on modern reinforcement learning theory have to deal with astronomically high action spaces, when considering multi-VNF, multi-domain, multi-slice problems; (iii) the training of the algorithms is not particularly data-efficient, which can hinder their practical application given the scarce(r) availability of cellular network related data (as opposed to standard machine learning problems). To this end, we attempt to tackle all the above shortcomings in one common framework. For (i), we propose a generic, queuing network based model that captures the inter-slice orchestration setting, supporting complex VNF chain topologies and end-to-end performance metrics. For (ii), we explore multi-agent DQN algorithms that can reduce action space complexity by orders of magnitude compared to standard DQN. For (iii), we investigate two mechanisms to store to and select from the experience replay buffer, in order to speed up the training of DQN agents. The proposed scheme was validated to outperform both vanilla DQN (by orders of magnitude faster convergence) and static heuristics ( $3times $ cost improvement).
{"title":"Sample-Efficient Multi-Agent DQNs for Scalable Multi-Domain 5G+ Inter-Slice Orchestration","authors":"Pavlos Doanis;Thrasyvoulos Spyropoulos","doi":"10.1109/TMLCN.2024.3420268","DOIUrl":"https://doi.org/10.1109/TMLCN.2024.3420268","url":null,"abstract":"Data-driven network slicing has been recently explored as a major driver for beyond 5G networks. Nevertheless, we are still a long way before such solutions are practically applicable in real problems. Most solutions addressing the problem of dynamically placing virtual network function chains (“slices”) on top of a physical topology still face one or more of the following hurdles: (i) they focus on simple slicing setups (e.g. single domain, single slice, simple VNF chains and performance metrics); (ii) solutions based on modern reinforcement learning theory have to deal with astronomically high action spaces, when considering multi-VNF, multi-domain, multi-slice problems; (iii) the training of the algorithms is not particularly data-efficient, which can hinder their practical application given the scarce(r) availability of cellular network related data (as opposed to standard machine learning problems). To this end, we attempt to tackle all the above shortcomings in one common framework. For (i), we propose a generic, queuing network based model that captures the inter-slice orchestration setting, supporting complex VNF chain topologies and end-to-end performance metrics. For (ii), we explore multi-agent DQN algorithms that can reduce action space complexity by orders of magnitude compared to standard DQN. For (iii), we investigate two mechanisms to store to and select from the experience replay buffer, in order to speed up the training of DQN agents. The proposed scheme was validated to outperform both vanilla DQN (by orders of magnitude faster convergence) and static heuristics (\u0000<inline-formula> <tex-math>$3times $ </tex-math></inline-formula>\u0000 cost improvement).","PeriodicalId":100641,"journal":{"name":"IEEE Transactions on Machine Learning in Communications and Networking","volume":"2 ","pages":"956-977"},"PeriodicalIF":0.0,"publicationDate":"2024-06-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10577096","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141618103","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-06-28DOI: 10.1109/TMLCN.2024.3420269
Ruiming Wang;Chenyang Yang;Shengqian Han;Jiajun Wu;Shuangfeng Han;Xiaoyun Wang
Hybrid precoding is an efficient technique for achieving high rates at a low cost in millimeter wave (mmWave) multi-antenna systems. Many research efforts have explored the use of deep learning to optimize hybrid precoding, particularly in static channel scenarios. However, in mobile communication systems, the performance of mmWave communication severely degrades due to the channel aging effect. Furthermore, the learned precoding policy should be adaptable to dynamic environments, such as variations in the number of active users, to avoid the need for re-training. In this paper, resorting to the proactive optimization approach, we propose an end-to-end learning method to learn the downlink multi-user analog and digital hybrid precoders directly from the received uplink sounding reference signals, without explicit channel estimation and prediction. We take into account the frame structure used in practical cellular systems and design a parallel proactive optimization network (P-PONet) to concurrently learn hybrid precoding for multiple downlink subframes. The P-PONet consists of several graph neural networks, which enable the generalizability across different system scales. Simulation results show that the proposed P-PONet outperforms existing methods in terms of sum-rate performance and sounding overhead, and is generalizable to various system configurations.
{"title":"Learning End-to-End Hybrid Precoding for Multi-User mmWave Mobile System With GNNs","authors":"Ruiming Wang;Chenyang Yang;Shengqian Han;Jiajun Wu;Shuangfeng Han;Xiaoyun Wang","doi":"10.1109/TMLCN.2024.3420269","DOIUrl":"https://doi.org/10.1109/TMLCN.2024.3420269","url":null,"abstract":"Hybrid precoding is an efficient technique for achieving high rates at a low cost in millimeter wave (mmWave) multi-antenna systems. Many research efforts have explored the use of deep learning to optimize hybrid precoding, particularly in static channel scenarios. However, in mobile communication systems, the performance of mmWave communication severely degrades due to the channel aging effect. Furthermore, the learned precoding policy should be adaptable to dynamic environments, such as variations in the number of active users, to avoid the need for re-training. In this paper, resorting to the proactive optimization approach, we propose an end-to-end learning method to learn the downlink multi-user analog and digital hybrid precoders directly from the received uplink sounding reference signals, without explicit channel estimation and prediction. We take into account the frame structure used in practical cellular systems and design a parallel proactive optimization network (P-PONet) to concurrently learn hybrid precoding for multiple downlink subframes. The P-PONet consists of several graph neural networks, which enable the generalizability across different system scales. Simulation results show that the proposed P-PONet outperforms existing methods in terms of sum-rate performance and sounding overhead, and is generalizable to various system configurations.","PeriodicalId":100641,"journal":{"name":"IEEE Transactions on Machine Learning in Communications and Networking","volume":"2 ","pages":"978-993"},"PeriodicalIF":0.0,"publicationDate":"2024-06-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10577095","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141618063","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Hybrid beamforming (HBF) and antenna selection are promising techniques for improving the energy efficiency (EE) of massive multiple-input multiple-output (mMIMO) systems. However, the transmitter architecture may contain several parameters that need to be optimized, such as the power allocated to the antennas and the connections between the antennas and the radio frequency chains. Therefore, finding the optimal transmitter architecture requires solving a non-convex mixed integer problem in a large search space. In this paper, we consider the problem of maximizing the EE of fully digital precoder (FDP) and HBF transmitters. First, we propose an energy model for different beamforming structures. Then, based on the proposed energy model, we develop a self-supervised learning (SSL) method to maximize the EE by designing the transmitter configuration for FDP and HBF. The proposed deep neural networks can provide different trade-offs between spectral efficiency and energy consumption while adapting to different numbers of active users. Finally, towards obtaining a system that can be trained using in-the-field measurements, we investigate the ability of the model to be trained exclusively using imperfect channel state information (CSI), both for the input to the deep learning model and for the calculation of the loss function. Simulation results show that the proposed solutions can outperform conventional methods in terms of EE while being trained with imperfect CSI. Furthermore, we show that the proposed solutions are less complex and more robust to noise than conventional methods.
混合波束成形（HBF）和天线选择是提高大规模多输入多输出（mMIMO）系统能效（EE）的有效技术。然而，发射机架构可能包含多个需要优化的参数，例如分配给天线的功率以及天线与射频链之间的连接。因此，要找到最佳的发射机架构，需要在一个很大的搜索空间内解决一个非凸混合整数问题。在本文中，我们考虑了最大化全数字前置编码器（FDP）和 HBF 发射机 EE 的问题。首先，我们提出了不同波束成形结构的能量模型。然后，基于所提出的能量模型，我们开发了一种自监督学习（SSL）方法，通过设计 FDP 和 HBF 的发射机配置来最大化 EE。所提出的深度神经网络可以在频谱效率和能耗之间进行不同的权衡，同时适应不同数量的活跃用户。最后，为了获得一个可利用现场测量进行训练的系统，我们研究了完全利用不完善的信道状态信息（CSI）来训练模型的能力，包括深度学习模型的输入和损失函数的计算。仿真结果表明，在使用不完美 CSI 进行训练时，所提出的解决方案在 EE 方面优于传统方法。此外，我们还表明，与传统方法相比，所提出的解决方案复杂度更低，对噪声的鲁棒性更高。
{"title":"Learning Energy-Efficient Transmitter Configurations for Massive MIMO Beamforming","authors":"Hamed Hojatian;Zoubeir Mlika;Jérémy Nadal;Jean-François Frigon;François Leduc-Primeau","doi":"10.1109/TMLCN.2024.3419728","DOIUrl":"https://doi.org/10.1109/TMLCN.2024.3419728","url":null,"abstract":"Hybrid beamforming (HBF) and antenna selection are promising techniques for improving the energy efficiency (EE) of massive multiple-input multiple-output (mMIMO) systems. However, the transmitter architecture may contain several parameters that need to be optimized, such as the power allocated to the antennas and the connections between the antennas and the radio frequency chains. Therefore, finding the optimal transmitter architecture requires solving a non-convex mixed integer problem in a large search space. In this paper, we consider the problem of maximizing the EE of fully digital precoder (FDP) and HBF transmitters. First, we propose an energy model for different beamforming structures. Then, based on the proposed energy model, we develop a self-supervised learning (SSL) method to maximize the EE by designing the transmitter configuration for FDP and HBF. The proposed deep neural networks can provide different trade-offs between spectral efficiency and energy consumption while adapting to different numbers of active users. Finally, towards obtaining a system that can be trained using in-the-field measurements, we investigate the ability of the model to be trained exclusively using imperfect channel state information (CSI), both for the input to the deep learning model and for the calculation of the loss function. Simulation results show that the proposed solutions can outperform conventional methods in terms of EE while being trained with imperfect CSI. Furthermore, we show that the proposed solutions are less complex and more robust to noise than conventional methods.","PeriodicalId":100641,"journal":{"name":"IEEE Transactions on Machine Learning in Communications and Networking","volume":"2 ","pages":"939-955"},"PeriodicalIF":0.0,"publicationDate":"2024-06-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10574840","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141618104","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-06-25DOI: 10.1109/TMLCN.2024.3418933
Almoatssimbillah Saifaldawla;Flor Ortiz;Eva Lagunas;Abuzar B. M. Adam;Symeon Chatzinotas
Recent advancements in satellite communications have highlighted the challenge of interference detection, especially with the new generation of non-geostationary orbit satellites (NGSOs) that share the same frequency bands as legacy geostationary orbit satellites (GSOs). Despite existing radio regulations during the filing stage, this heightened congestion in the spectrum is likely to lead to instances of interference during real-time operations. This paper addresses the NGSO-to-GSO interference problem by proposing advanced artificial intelligence (AI) models to detect interference events. In particular, we focus on the downlink interference case, where signals from low-Earth orbit satellites (LEOs) potentially impact the signals received at the GSO ground stations (GGSs). In addition to the widely used autoencoder-based models (AEs), we design, develop, and train two generative AI-based models (GenAI), which are a variational autoencoder (VAE) and a transformer-based interference detector (TrID). These models generate samples of the expected GSO signal, whose error with respect to the input signal is used to flag interference. Actual satellite positions, trajectories, and realistic system parameters are used to emulate the interference scenarios and validate the proposed models. Numerical evaluation reveals that the models exhibit higher accuracy for detecting interference in the time-domain signal representations compared to the frequency-domain representations. Furthermore, the results demonstrate that TrID significantly outperforms the other models as well as the traditional energy detector (ED) approach, showing an increase of up to 31.23% in interference detection accuracy, offering an innovative and efficient solution to a pressing challenge in satellite communications.
{"title":"GenAI-Based Models for NGSO Satellites Interference Detection","authors":"Almoatssimbillah Saifaldawla;Flor Ortiz;Eva Lagunas;Abuzar B. M. Adam;Symeon Chatzinotas","doi":"10.1109/TMLCN.2024.3418933","DOIUrl":"https://doi.org/10.1109/TMLCN.2024.3418933","url":null,"abstract":"Recent advancements in satellite communications have highlighted the challenge of interference detection, especially with the new generation of non-geostationary orbit satellites (NGSOs) that share the same frequency bands as legacy geostationary orbit satellites (GSOs). Despite existing radio regulations during the filing stage, this heightened congestion in the spectrum is likely to lead to instances of interference during real-time operations. This paper addresses the NGSO-to-GSO interference problem by proposing advanced artificial intelligence (AI) models to detect interference events. In particular, we focus on the downlink interference case, where signals from low-Earth orbit satellites (LEOs) potentially impact the signals received at the GSO ground stations (GGSs). In addition to the widely used autoencoder-based models (AEs), we design, develop, and train two generative AI-based models (GenAI), which are a variational autoencoder (VAE) and a transformer-based interference detector (TrID). These models generate samples of the expected GSO signal, whose error with respect to the input signal is used to flag interference. Actual satellite positions, trajectories, and realistic system parameters are used to emulate the interference scenarios and validate the proposed models. Numerical evaluation reveals that the models exhibit higher accuracy for detecting interference in the time-domain signal representations compared to the frequency-domain representations. Furthermore, the results demonstrate that TrID significantly outperforms the other models as well as the traditional energy detector (ED) approach, showing an increase of up to 31.23% in interference detection accuracy, offering an innovative and efficient solution to a pressing challenge in satellite communications.","PeriodicalId":100641,"journal":{"name":"IEEE Transactions on Machine Learning in Communications and Networking","volume":"2 ","pages":"904-924"},"PeriodicalIF":0.0,"publicationDate":"2024-06-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10570488","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141500330","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-06-24DOI: 10.1109/TMLCN.2024.3418756
Ulya Sabeel;Shahram Shah Heydari;Khalil El-Khatib;Khalid Elgazzar
AI-based Network Intrusion Detection Systems (NIDS) provide effective mechanisms for cybersecurity analysts to gain insights and thwart several network attacks. Although current IDS can identify known/typical attacks with high accuracy, current research shows that such systems perform poorly when facing atypical and dynamically changing (polymorphic) attacks. In this paper, we focus on improving detection capability of the IDS for atypical and polymorphic network attacks. Our system generates adversarial polymorphic attacks against the IDS to examine its performance and incrementally retrains it to strengthen its detection of new attacks, specifically for minority attack samples in the input data. The employed attack quality analysis ensures that the adversarial atypical/polymorphic attacks generated through our system resemble original network attacks. We showcase the high performance of the IDS that we have proposed by training it using the CICIDS2017 and CICIoT2023 benchmark datasets and evaluating its performance against several atypical/polymorphic attack flows. The results indicate that the proposed technique, through adaptive training, learns the pattern of dynamically changing atypical/polymorphic attacks, identifies such attacks with approximately 90% balanced accuracy for most of the cases, and surpasses various state-of-the-art detection and class balancing techniques.
{"title":"Incremental Adversarial Learning for Polymorphic Attack Detection","authors":"Ulya Sabeel;Shahram Shah Heydari;Khalil El-Khatib;Khalid Elgazzar","doi":"10.1109/TMLCN.2024.3418756","DOIUrl":"https://doi.org/10.1109/TMLCN.2024.3418756","url":null,"abstract":"AI-based Network Intrusion Detection Systems (NIDS) provide effective mechanisms for cybersecurity analysts to gain insights and thwart several network attacks. Although current IDS can identify known/typical attacks with high accuracy, current research shows that such systems perform poorly when facing atypical and dynamically changing (polymorphic) attacks. In this paper, we focus on improving detection capability of the IDS for atypical and polymorphic network attacks. Our system generates adversarial polymorphic attacks against the IDS to examine its performance and incrementally retrains it to strengthen its detection of new attacks, specifically for minority attack samples in the input data. The employed attack quality analysis ensures that the adversarial atypical/polymorphic attacks generated through our system resemble original network attacks. We showcase the high performance of the IDS that we have proposed by training it using the CICIDS2017 and CICIoT2023 benchmark datasets and evaluating its performance against several atypical/polymorphic attack flows. The results indicate that the proposed technique, through adaptive training, learns the pattern of dynamically changing atypical/polymorphic attacks, identifies such attacks with approximately 90% balanced accuracy for most of the cases, and surpasses various state-of-the-art detection and class balancing techniques.","PeriodicalId":100641,"journal":{"name":"IEEE Transactions on Machine Learning in Communications and Networking","volume":"2 ","pages":"869-887"},"PeriodicalIF":0.0,"publicationDate":"2024-06-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10570491","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141494827","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-06-24DOI: 10.1109/TMLCN.2024.3418748
Guo Hao Thng;Said Mikki
In recent years, the research involving the use of machine learning in the field of communication networks have shown promising results, in particular, improving receiver sensitivity against noise and link impairment. The proposal of analog radio-over-fiber fronthaul solutions simplifies the overall base station configuration by generating wireless signals at the desired transmission frequency, directly after photodiode heterodyne detection, without requiring additional frequency upconversion components. However, analog radio-over-fiber signals is more susceptible to nonlinear distortions originating from the optical transmission system. This paper explores the use of machine learning in an analog radio-over-fiber link, improving receiver sensitivity in the presence of phase noise. The machine learning algorithm is implemented at the receiver. To evaluate the feasibility of the proposed machine learning based phase noise correction approach, software simulations were conducted to collect data needed for machine leanring algorithm training. Initial findings suggests that the proposed machine-learning-based receiver’s can perform close to conventional heterodyned-based receivers in terms of detection accuracy, exhibiting great tolerance against phase-induced noise, with a symbol error rate improvement from $10^{-2}$ to $10^{-5}$ , using a relatively simple machine learning algorithm with only 3 hidden layers consisting of fully connected feedforward neural networks.
{"title":"A Machine Learning Aided Reference-Tone-Based Phase Noise Correction Framework for Fiber-Wireless Systems","authors":"Guo Hao Thng;Said Mikki","doi":"10.1109/TMLCN.2024.3418748","DOIUrl":"https://doi.org/10.1109/TMLCN.2024.3418748","url":null,"abstract":"In recent years, the research involving the use of machine learning in the field of communication networks have shown promising results, in particular, improving receiver sensitivity against noise and link impairment. The proposal of analog radio-over-fiber fronthaul solutions simplifies the overall base station configuration by generating wireless signals at the desired transmission frequency, directly after photodiode heterodyne detection, without requiring additional frequency upconversion components. However, analog radio-over-fiber signals is more susceptible to nonlinear distortions originating from the optical transmission system. This paper explores the use of machine learning in an analog radio-over-fiber link, improving receiver sensitivity in the presence of phase noise. The machine learning algorithm is implemented at the receiver. To evaluate the feasibility of the proposed machine learning based phase noise correction approach, software simulations were conducted to collect data needed for machine leanring algorithm training. Initial findings suggests that the proposed machine-learning-based receiver’s can perform close to conventional heterodyned-based receivers in terms of detection accuracy, exhibiting great tolerance against phase-induced noise, with a symbol error rate improvement from \u0000<inline-formula> <tex-math>$10^{-2}$ </tex-math></inline-formula>\u0000 to \u0000<inline-formula> <tex-math>$10^{-5}$ </tex-math></inline-formula>\u0000, using a relatively simple machine learning algorithm with only 3 hidden layers consisting of fully connected feedforward neural networks.","PeriodicalId":100641,"journal":{"name":"IEEE Transactions on Machine Learning in Communications and Networking","volume":"2 ","pages":"888-903"},"PeriodicalIF":0.0,"publicationDate":"2024-06-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10570478","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141494826","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-06-21DOI: 10.1109/TMLCN.2024.3417806
Da Huang;Akram Al-Hourani
The proliferation of the Internet of Things (IoT) has created significant opportunities for future telecommunications. A popular category of IoT devices is oriented toward low-cost and low-power applications. However, certain aspects of such category, including the authentication process, remain inadequately investigated against cyber vulnerabilities. This is caused by the inherent trade-off between device complexity and security rigor. In this work, we propose an authentication method based on radio frequency fingerprinting (RFF) using deep learning. This method can be implemented on the base station side without increasing the complexity of the IoT devices. Specifically, we propose four representation modalities based on continuous wavelet transform (CWT) to exploit tempo-spectral radio fingerprints. Accordingly, we utilize the generative adversarial network (GAN) and convolutional neural network (CNN) for spoof detection and authentication. For empirical validation, we consider the widely popular LoRa system with a focus on the preamble of the radio frame. The presented experimental test involves 20 off-the-shelf LoRa modules to demonstrate the feasibility of the proposed approach, showing reliable detection results of spoofing devices and high-level accuracy in authentication of 92.4%.
{"title":"Physical Layer Spoof Detection and Authentication for IoT Devices Using Deep Learning Methods","authors":"Da Huang;Akram Al-Hourani","doi":"10.1109/TMLCN.2024.3417806","DOIUrl":"https://doi.org/10.1109/TMLCN.2024.3417806","url":null,"abstract":"The proliferation of the Internet of Things (IoT) has created significant opportunities for future telecommunications. A popular category of IoT devices is oriented toward low-cost and low-power applications. However, certain aspects of such category, including the authentication process, remain inadequately investigated against cyber vulnerabilities. This is caused by the inherent trade-off between device complexity and security rigor. In this work, we propose an authentication method based on radio frequency fingerprinting (RFF) using deep learning. This method can be implemented on the base station side without increasing the complexity of the IoT devices. Specifically, we propose four representation modalities based on continuous wavelet transform (CWT) to exploit tempo-spectral radio fingerprints. Accordingly, we utilize the generative adversarial network (GAN) and convolutional neural network (CNN) for spoof detection and authentication. For empirical validation, we consider the widely popular LoRa system with a focus on the preamble of the radio frame. The presented experimental test involves 20 off-the-shelf LoRa modules to demonstrate the feasibility of the proposed approach, showing reliable detection results of spoofing devices and high-level accuracy in authentication of 92.4%.","PeriodicalId":100641,"journal":{"name":"IEEE Transactions on Machine Learning in Communications and Networking","volume":"2 ","pages":"841-854"},"PeriodicalIF":0.0,"publicationDate":"2024-06-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10568158","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141474855","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-06-21DOI: 10.1109/TMLCN.2024.3417889
Nageswara S. V. Rao;Chris Y. T. Ma;Fei He
Data transfer infrastructures composed of Data Transfer Nodes (DTN) are critical to meeting distributed computing and storage demands of clouds, data repositories, and complexes of supercomputers and instruments. The infrastructure’s throughput profile, estimated as a function of the connection round trip time using Machine Learning (ML) methods, is an indicator of its operational state, and has been utilized for monitoring, diagnosis and optimization purposes. We show that the inherent statistical variations and precision of throughput profiles estimated by ML methods can be exploited for unauthorized use of DTNs’ computing and network capacity. We present a game theoretic formulation that captures the cost-benefit trade-offs between an attacker that attempts to hide under the profile’s statistical variations and a provider that attempts to balance compromise detection with the cost of throughput measurements. The Nash equilibrium conditions adapted to this game provide qualitative insights and bounds for the success probabilities of the attacker and provider, by utilizing the generalization equation of ML-estimate. We present experimental results that illustrate this game wherein a significant portion of DTN computing capacity is compromised without being detected by an attacker that exploits the ML estimate properties.
由数据传输节点（DTN）组成的数据传输基础设施对于满足云、数据存储库以及超级计算机和仪器群的分布式计算和存储需求至关重要。使用机器学习（ML）方法估算的基础设施吞吐量曲线是连接往返时间的函数，是其运行状态的指标，已被用于监控、诊断和优化目的。我们的研究表明，可以利用 ML 方法估算的吞吐量曲线的固有统计变化和精度，在未经授权的情况下使用 DTN 的计算和网络容量。我们提出了一个博弈论公式，它捕捉到了试图隐藏在吞吐量曲线统计变化下的攻击者与试图平衡破坏检测与吞吐量测量成本的提供者之间的成本效益权衡。通过利用 ML-estimate 的广义方程，适应该博弈的纳什均衡条件为攻击者和提供者的成功概率提供了定性的见解和界限。我们展示的实验结果说明了这种博弈，在这种博弈中，利用 ML 估计特性的攻击者可以在不被检测到的情况下破坏大部分 DTN 计算能力。
{"title":"Game Strategies for Data Transfer Infrastructures Against ML-Profile Exploits","authors":"Nageswara S. V. Rao;Chris Y. T. Ma;Fei He","doi":"10.1109/TMLCN.2024.3417889","DOIUrl":"https://doi.org/10.1109/TMLCN.2024.3417889","url":null,"abstract":"Data transfer infrastructures composed of Data Transfer Nodes (DTN) are critical to meeting distributed computing and storage demands of clouds, data repositories, and complexes of supercomputers and instruments. The infrastructure’s throughput profile, estimated as a function of the connection round trip time using Machine Learning (ML) methods, is an indicator of its operational state, and has been utilized for monitoring, diagnosis and optimization purposes. We show that the inherent statistical variations and precision of throughput profiles estimated by ML methods can be exploited for unauthorized use of DTNs’ computing and network capacity. We present a game theoretic formulation that captures the cost-benefit trade-offs between an attacker that attempts to hide under the profile’s statistical variations and a provider that attempts to balance compromise detection with the cost of throughput measurements. The Nash equilibrium conditions adapted to this game provide qualitative insights and bounds for the success probabilities of the attacker and provider, by utilizing the generalization equation of ML-estimate. We present experimental results that illustrate this game wherein a significant portion of DTN computing capacity is compromised without being detected by an attacker that exploits the ML estimate properties.","PeriodicalId":100641,"journal":{"name":"IEEE Transactions on Machine Learning in Communications and Networking","volume":"2 ","pages":"925-938"},"PeriodicalIF":0.0,"publicationDate":"2024-06-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10568232","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141539198","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-06-13DOI: 10.1109/TMLCN.2024.3414296
Zhikai Liu;Haifeng Luo;Tharmalingam Ratnarajah
Digital self-interference cancellation (DSIC) has become a pivotal strategy for implementing in-band full-duplex (IBFD) radios to overcome the hurdles posed by residual self-interference that persist after propagation and analog domain cancellation. This work proposes a novel reservoir computing-based DSIC (RC-DSIC) technique and compares it with traditional polynomial-based (PL-DSIC) and various existing neural network-based (NN-DSIC) approaches. We begin by delineating the structure of the RC and exploring its capability to address the DSIC task, highlighting its potential advantages over current methodologies. Subsequently, we examine the computational complexity of these approaches and undertake extensive simulations to compare the proposed RC-DSIC approach against PL-DSIC and existing NN-DSIC schemes. Our results reveal that the RC-DSIC scheme attains 99.84% of the performance offered by PL-based DSIC algorithms while requiring only 1.51% of the computational demand. Compared to many existing NN-DSIC schemes, the RC-DSIC method achieves at least 99.73% of its performance with no more than 36.61% of the computational demand. This performance justifies the viability of RC-DSIC as an effective and efficient solution for DSIC in IBFD, striking it is a better implementation method in terms of computational simplicity.
{"title":"Reservoir Computing-Based Digital Self-Interference Cancellation for In-Band Full-Duplex Radios","authors":"Zhikai Liu;Haifeng Luo;Tharmalingam Ratnarajah","doi":"10.1109/TMLCN.2024.3414296","DOIUrl":"https://doi.org/10.1109/TMLCN.2024.3414296","url":null,"abstract":"Digital self-interference cancellation (DSIC) has become a pivotal strategy for implementing in-band full-duplex (IBFD) radios to overcome the hurdles posed by residual self-interference that persist after propagation and analog domain cancellation. This work proposes a novel reservoir computing-based DSIC (RC-DSIC) technique and compares it with traditional polynomial-based (PL-DSIC) and various existing neural network-based (NN-DSIC) approaches. We begin by delineating the structure of the RC and exploring its capability to address the DSIC task, highlighting its potential advantages over current methodologies. Subsequently, we examine the computational complexity of these approaches and undertake extensive simulations to compare the proposed RC-DSIC approach against PL-DSIC and existing NN-DSIC schemes. Our results reveal that the RC-DSIC scheme attains 99.84% of the performance offered by PL-based DSIC algorithms while requiring only 1.51% of the computational demand. Compared to many existing NN-DSIC schemes, the RC-DSIC method achieves at least 99.73% of its performance with no more than 36.61% of the computational demand. This performance justifies the viability of RC-DSIC as an effective and efficient solution for DSIC in IBFD, striking it is a better implementation method in terms of computational simplicity.","PeriodicalId":100641,"journal":{"name":"IEEE Transactions on Machine Learning in Communications and Networking","volume":"2 ","pages":"855-868"},"PeriodicalIF":0.0,"publicationDate":"2024-06-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10556632","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141474856","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-06-12DOI: 10.1109/TMLCN.2024.3413669
Shayan Zargari;Chintha Tellambura;Amine Maaref;Geoffrey Ye Li
In ambient backscatter communication (AmBC), battery-free devices (tags) harvest energy from ambient radio frequency (RF) signals and communicate with readers. Although reliable channel estimation (CE) is critical, classical pilot-based estimators tend to perform poorly. To address this challenge, we treat CE as a denoising problem using conditional generative adversarial networks (CGANs). A three-dimensional (3D) denoising block leverages spatial and temporal characteristics of pilot signals, considering both real and imaginary components of channel matrices. The proposed CGAN estimator is extensively evaluated against traditional estimators like minimum mean-squared error (MMSE), least squares (LS), convolutional neural network (CNN), CNN-based deep residual learning denoiser (CRLD), and blind estimation. Simulation results show 82% gain of the proposed estimator over CRLD and MMSE estimators at an SNR of 5 dB. Moreover, it has advanced learning capabilities and accurately replicates complex channel characteristics.
{"title":"Deep Conditional Generative Adversarial Networks for Efficient Channel Estimation in AmBC Systems","authors":"Shayan Zargari;Chintha Tellambura;Amine Maaref;Geoffrey Ye Li","doi":"10.1109/TMLCN.2024.3413669","DOIUrl":"https://doi.org/10.1109/TMLCN.2024.3413669","url":null,"abstract":"In ambient backscatter communication (AmBC), battery-free devices (tags) harvest energy from ambient radio frequency (RF) signals and communicate with readers. Although reliable channel estimation (CE) is critical, classical pilot-based estimators tend to perform poorly. To address this challenge, we treat CE as a denoising problem using conditional generative adversarial networks (CGANs). A three-dimensional (3D) denoising block leverages spatial and temporal characteristics of pilot signals, considering both real and imaginary components of channel matrices. The proposed CGAN estimator is extensively evaluated against traditional estimators like minimum mean-squared error (MMSE), least squares (LS), convolutional neural network (CNN), CNN-based deep residual learning denoiser (CRLD), and blind estimation. Simulation results show 82% gain of the proposed estimator over CRLD and MMSE estimators at an SNR of 5 dB. Moreover, it has advanced learning capabilities and accurately replicates complex channel characteristics.","PeriodicalId":100641,"journal":{"name":"IEEE Transactions on Machine Learning in Communications and Networking","volume":"2 ","pages":"805-822"},"PeriodicalIF":0.0,"publicationDate":"2024-06-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10555303","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141453439","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","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：