Last updated: 2024-11-03 04:01 UTC
All documents
Number of pages: 129
Author(s) | Title | Year | Publication | Keywords | ||
---|---|---|---|---|---|---|
Ru Huo, Xiangfeng Cheng, Chuang Sun, Tao Huang | A Cluster-Based Data Transmission Strategy for Blockchain Network in the Industrial Internet of Things | 2024 | Early Access | Blockchains Industrial Internet of Things Edge computing Data communication Computer architecture Topology Cloud computing Industrial Internet of Things (IIoT) blockchain edge computing clustering data transmission strategy | The proliferation of devices and data in the Industrial Internet of Things (IIoT) has rendered the traditional centralized cloud model unable to meet the stringent requirements of wide-scale and low latency in these IIoT scenarios. As emerging technologies, edge computing enables real-time processing and analysis on devices situated closer to the data source while reducing bandwidth requirements. Blockchain, being decentralized, could enhance data security. Therefore, edge computing and blockchain are integrated in IIoT to reduce latency and improve security. However, the inefficient data transmission of blockchain leads to increased transmission latency in the IIoT. To address this issue, we propose a cluster-based data transmission strategy (CDTS) for blockchain network. Initially, an improved weighted label propagation algorithm (WLPA) is proposed for clustering blockchain nodes. Subsequently, a spanning tree topology construction (STTC) is designed to simplify the blockchain network topology, based on the above node clustering results. Additionally, leveraging clustered nodes and tree topology, we propose a data transmission strategy to speed up data transmission. Simulation experiments show that CDTS effectively reduces data transmission time and better supports large-scale IIoT scenarios. | 10.1109/TNSM.2024.3387120 |
Pratyush Dikshit, Mike Kosek, Nils Faulhaber, Jayasree Sengupta, Vaibhav Bajpai | Evaluating DNS Resiliency and Responsiveness With Truncation, Fragmentation & DoTCP Fallback | 2024 | Early Access | Domain Name System Resilience Probes Internet Time factors Servers IP networks DNS DNS-over-TCP DNS-over-UDP Response Time Failure Rate EDNS(0) | Since its introduction in 1987, the DNS has become one of the core components of the Internet. While it was designed to work with both TCP and UDP, DNS-over-UDP (DoUDP) has become the default option due to its low overhead. As new Resource Records were introduced, the sizes of DNS responses increased considerably. This expansion of the message body has led to truncation and IP fragmentation more often in recent years where large UDP responses make DNS an easy vector for amplifying denial-of-service attacks which can reduce the resiliency of DNS services. This paper investigates the resiliency, responsiveness, and usage of DoTCP and DoUDP over IPv4 and IPv6 for 10 widely used public DNS resolvers. The paper specifically measures the resiliency of the DNS infrastructure in the age of increasing DNS response sizes that lead to truncation and fragmentation. Our results offer key insights into the management of robust and reliable DNS network services. While DNS Flag Day 2020 recommends 1232 bytes of buffer sizes, we find out that 3/10 resolvers mainly announce very large EDNS(0) buffer sizes both from the edge as well as from the core, which potentially causes fragmentation. In reaction to large response sizes from authoritative name servers, we find that resolvers do not fall back to the usage of DoTCP in many cases, bearing the risk of fragmented responses. As the message sizes in the DNS are expected to grow further, this problem will become more urgent in the future. This paper demonstrates the key results (particularly as a consequence of the DNS Flag Day 2020) which may support network service providers make informed choices to better manage their critical DNS services. | 10.1109/TNSM.2024.3365303 |
Bing Shi, Zhifeng Chen, Zhuohan Xu | A Deep Reinforcement Learning Based Approach for Optimizing Trajectory and Frequency in Energy Constrained Multi-UAV Assisted MEC System | 2024 | Early Access | Autonomous aerial vehicles Task analysis Trajectory Optimization Servers Computer architecture Computational modeling Mobile Edge Computing Unmanned Aerial Vehicle Multi-Agent Deep Reinforcement Learning | Mobile Edge Computing (MEC) is a technology that shows great promise in enhancing the computational power of smart devices (SDs) in the Internet of Things (IoT). However, the fixed location and limited coverage of MEC servers constrain their performance. To overcome this issue, this paper explores a multiple unmanned aerial vehicle (UAV) assisted MEC system. The proposed system considers a scenario where multiple UAVs work together to provide computing services while dynamically adjusting their frequency based on the task size, under the constraint of limited energy. This paper aims to maximize computation bits, SDs’ fairness, and UAVs’ load balancing in multi-UAV assisted MEC system by jointly optimizing the trajectory and frequency. To address this challenge, we model it as a Partially Observable Markov Decision Process and propose a joint optimization strategy based on multi-agent deep reinforcement learning. The effectiveness of the proposed strategy is evaluated on both synthetic and realistic datasets. The results demonstrate that our strategy outperforms other benchmark strategies. | 10.1109/TNSM.2024.3362949 |
Hao Xu, Harry Chang, Kun Qiu, Yang Hong, Wenjun Zhu, Xiang Wang, Baoqian Li, Jin Zhao | Accelerating Deep Packet Inspection With SIMD-Based Multi-Literal Matching Engine | 2024 | Early Access | Engines Software algorithms Inspection Telecommunication traffic Software Payloads Network security network security DPI SIMD parallel computing | Deep Packet Inspection (DPI) has been one of the most significant network security techniques. It is widely used to identify and classify network traffic in various applications such as web application firewall and intrusion detection. Different from traditional packet filtering that only examines packet headers, DPI detects payloads as well by comparing them with an existing signature database. The literal matching engine, which plays a key role in DPI, is the primary determinant of the system performance. FDR, an engine that utilizes 3 SIMD operations to match 1 character with multiple literals, has been developed and is currently one of the fastest literal matching engines. However, FDR has significant performance drop-off when faced with small-scale literal rule sets, whose proportion is more than 90% in modern databases. In this paper, we designed Teddy, an engine that is highly optimized for small-scale literal rule sets. Compared with FDR, Teddy significantly improves the matching efficiency by a novel shift-or matching algorithm that can simultaneously match up to 64 characters with only 15 SIMD operations. We evaluate Teddy with real-world traffic and rule sets. Experimental results show that its performance is up to 43.07x that of Aho-corasick (AC) and 2.17x that of FDR. Teddy has been successfully integrated into Hyperscan, together with which it is widely deployed in modern popular DPI applications such as Snort and Suricata. | 10.1109/TNSM.2024.3354985 |
Patrick Luiz de Araújo, Murillo G. Carneiro, Luis M. Contreras, Rafael Pasquini | MTP-NT: A Mobile Traffic Predictor Enhanced by Neighboring and Transportation Data | 2024 | Early Access | Telecommunication traffic Urban areas Predictive models Data models Autoregressive processes Deep learning Analytical models 5G mobile communication Long short term memory Recurrent neural networks Mobile Networks 5G Time Series Deep Learning Network Traffic Forecasting Network Function Virtualization (NFV) Network Traffic Monitoring and Analysis (NTMA) | The development of techniques able to forecast the mobile network traffic in a city can feed data driven applications, as Virtual Network Functions (VNF) orchestrators, optimizing the resource allocation and increasing the capacity of mobile networks. Despite the fact that several studies have addressed this problem, many did not consider neither the traffic relationship among city regions nor the information retrieved from public transport stations, which may provide useful information to better anticipate the network traffic. In this paper, we propose a new deep learning based architecture to forecast the network traffic using representation learning and recurrent neural networks. The framework, named Mobile Traffic Predictor Enhanced by Neighboring and Transportation Data (MTP-NT), has two major components: the first one is responsible of learning from the time series of the region to be predicted, with the second one learning from the time series of both neighboring regions and public transportation stations. Several experiments were conducted over a dataset from the city of Milan, as well as comparisons against widely adopted and state-of-the-art techniques. The results shown in this paper demonstrate that the usage of public transport information contributes to improve the forecasts in central areas of the city, as well as in regions with aperiodic demands, such as tourist regions. | 10.1109/TNSM.2024.3488568 |
Bo Yang, Liquan Chen, Jiaorui Shen, Huaqun Wang, Yang Ma | FHE-Based Publicly Verifiable Sealed-bid Auction Protocol atop Cross-blockchain | 2024 | Early Access | Protocols Internet Advertising Blockchains Privacy Monopoly Homomorphic encryption Relays Costs Servers Fully homomorphic encryption Approximate comparison Cross-blockchain Public verifiability Privacy | Online auctions, which are widely used on Internet advertising platforms, reduce the participation costs for buyers and sellers, and promote the flow of tens of billions of dollars in the global economy. However, internet advertising platforms tend to be monopolistic and adopt a sealed bidding model. Therefore, when price is the sole determinant of the winner, how to publicly verify the correctness of auction results without disclosing bidding information has become a challenge. To address these issues, we propose a fully homomorphic encryption (FHE)-based sealed-bid auction protocol with public verifiability atop cross-blockchain. Through an approximate comparison algorithm, the proof of the winner consists of m-1 (or 1) homomorphic ciphertexts, significantly reducing communication costs, where m represents the number of bidders. Thus, anyone can check the winner’s proof and complete the public verification of correctness. Moreover, this paper designs a cross-blockchain auction system model, breaking the monopoly of platforms, and proposes a distributed private key sharing method, which realizes the auditing function of the relay chain. Finally, we formalize the security model, and verify the correctness, public verifiability and privacy of our scheme. The off-chain time overhead and on-chain gas consumption demonstrate the strong practicability of our protocol in large-scale auctions. | 10.1109/TNSM.2024.3488090 |
Ashutosh Dutta, Ehab Al-Shaer, Ehsan Aghaei, Qi Duan, Hasan Yasar | Security Control Grid for Optimized Cyber Defense Planning | 2024 | Early Access | Security Computer security Planning Decision making Risk mitigation Resilience Process control Optimization NIST Costs security controls risk management security management and optimization vulnerability management language modeling | Cybersecurity controls are essential for ensuring information confidentiality, integrity, and availability. However, selecting the most effective controls to maximize return on investment (RoI) in cyber defense is a complex task involving numerous factors such as vulnerabilities, threat prioritization, and budget constraints. This paper introduces an innovative model and optimization techniques to select cybersecurity controls (CSC) for optimal risk mitigation, balancing residual risk, budget, and resiliency requirements. Our approach features the Security Control Grid (SCG) model, which automatically determines the necessary controls based on their security functions (Identify, Protect, Detect, Respond, and Recover), strategic placement within the cyber environment, and effectiveness at different stages of the attack kill chain. We formulate cybersecurity control decision-making as a multidimensional optimization problem, solving it using Satisfiability Modulo Theories (SMT). Additionally, we integrate a domain-specific language model that links CSCs with Common Vulnerabilities and Exposures (CVEs). This approach is implemented in the SCG solver tool, which generates scalable and robust CSC deployment plans that optimize cybersecurity RoI and maintain acceptable residual risk for large-scale enterprises. | 10.1109/TNSM.2024.3488011 |
Tingnan Bao, Aisha Syed, William Sean Kennedy, Melike Erol-Kantarci | Sustainable Task Offloading in Secure UAV-Assisted Smart Farm Networks: A Multi-Agent DRL With Action Mask Approach | 2024 | Early Access | Internet of Things Autonomous aerial vehicles Smart agriculture Servers Resource management Energy consumption Delays Sustainable development Real-time systems Productivity Task Offloading Unmanned Aerial Vehicles (UAVs) Physical Layer Security Energy Consumption Deep Reinforcement Learning (DRL) | The integration of unmanned aerial vehicles (UAVs) with mobile edge computing (MEC) and Internet of Things (IoT) technology is crucial for efficient resource management and sustainable agricultural productivity in smart frams. This paper addresses the critical need for optimizing task offloading in secure UAV-assisted smart farm networks, aiming to reduce total delay and energy consumption while maintaining robust security in data communications. We propose a multi-agent deep reinforcement learning (DRL)-based approach using a deep double Q-network (DDQN) with an action mask (AM), designed to manage task offloading dynamically and efficiently. Simulation results demonstrate the superior performance of our method in managing task offloading, highlighting significant improvements in operational efficiency, such as reduced delay and energy consumption. This aligns with the goal of developing sustainable and energy-efficient solutions for next-generation network infrastructures, making our approach an advanced solution for performance and sustainability in smart farming applications. | 10.1109/TNSM.2024.3486288 |
Zhaowei Zhang, Chunfeng Liu, Wenyu Qu, Zhao Zhao, Weisi Guo | PRobust: A Percolation-Based Robustness Optimization Model for Underwater Acoustic Sensor Networks | 2024 | Early Access | Robustness Optimization Network topology Topology Underwater acoustics Routing Data communication Computational modeling Wireless sensor networks Heuristic algorithms Underwater Acoustic Sensor Networks Robustness Optimization Percolation Current Movement Data Transmission Network Traffic | In Underwater Acoustic Sensor Networks (UASNs), the robustness of network is greatly affected by complex marine environments when implementing multi-hop data transmission. Factors such as the underwater acoustic channel and dynamic topological changes induced by multi-layered oceanic vortices exacerbate this influence. However, there is currently a research gap in the specific area of robustness optimization for UASNs. Existing studies on robustness optimization are unsuitable for UASNs as they neglect the considerations of the marine environment and node characteristics (e.g., residual energy). In this work, we propose PRobust, a percolation-based robustness optimization model for UASNs. PRobust consists of two distinct phases: percolation modeling and bottleneck optimization. In the percolation modeling phase, we incorporate both node and edge features, considering the physical and topological properties, and introduce a novel approach for calculating link quality. In the bottleneck optimization phase, we devise a graph theory-based method to identify bottlenecks, leveraging the flow information recorded by nodes to improve the accuracy of bottleneck discovery. Moreover, we integrated time slots and a current movement model into the proposed model, allowing its applicability to dynamically changing UASNs. Extensive simulation results indicate that, compared to existing methods, PRobust significantly enhances network robustness and performance with the same overhead after bottleneck optimization. | 10.1109/TNSM.2024.3487956 |
Jia Xu, Xiao Liu, Jiong Jin, Wuzhen Pan, Xuejun Li, Yun Yang | Holistic Service Provisioning in a UAV-UGV Integrated Network for Last-Mile Delivery | 2024 | Early Access | Autonomous aerial vehicles Energy consumption Logistics Collaboration Servers Network architecture Vehicle dynamics Real-time systems Quality of service Optimization Service Provisioning Drone-as-a-Service Multi-access Edge Computing Last-Mile Delivery | Effective last-mile delivery is pivotal in smart logistics system. While existing delivery network architectures, such as Drone-as-a-Service (DaaS), are capable of enhancing order delivery effectiveness, they often fall short in provisioning diverse delivery services. Furthermore, DaaS-based last-mile delivery systems face challenges from limited payload capacity and range. In this paper, we propose a UAV-UGV integrated network architecture based on Multi-access Edge Computing (MEC), denoted as DaaS+, encompassing the diverse delivery services from both Unmanned Aerial Vehicle (UAV) and Unmanned Ground Vehicle (UGV) for last-mile delivery. To optimize the effectiveness in the delivery process, the intricate overhead and constraints of heterogeneous delivery services are taken into the full consideration. Specifically, we present an energy-aware service model for the UAV-UGV integrated network that considers the deadline constraints of services. Additionally, we address issues of service unavailability during service provisioning. To identify optimal service provisioning plans, we design a novel Energy-Aware Holistic Service Provisioning Mechanism based on Particle Swarm Optimization (ES-PSO), which minimizes delivery energy consumption while adhering to service deadline constraints. Experimental results substantiate the effectiveness of our proposed solution, demonstrating its ability to generate superior service provisioning plans and significantly reduce total delivery energy consumption. | 10.1109/TNSM.2024.3487357 |
Muhammad Asghar Khan, Neeraj Kumar, Saeed Hamood Alsamhi, Gordana Barb, Justyna Zywiołek, Insaf Ullah, Fazal Noor, Jawad Ali Shah, Abdullah M. Almuhaideb | Security and Privacy Issues and Solutions for UAVs in B5G Networks: A Review | 2024 | Early Access | Security Autonomous aerial vehicles Privacy Reviews 6G mobile communication Data models Cryptography 5G mobile communication Drones Mathematical models AI/ML B5G networks blockchain federated learning security physical layer security privacy post-quantum cryptography | Unmanned aerial vehicles (UAVs) in beyond 5G (B5G) are crucial for revolutionizing various industries, including surveillance, agriculture, and logistics, by enabling high-speed data transfer, ultra-low latency communication, and ultra-reliable connectivity. However, integrating UAVs into B5G networks poses various security and privacy concerns. These risks encompass the possibility of unauthorized access, breaches of data, and cyber-physical attacks which jeopardize the integrity, confidentiality, and availability of UAV operations. Moreover, UAVs in B5G networks are also at high risk from the application of machine learning (ML)-based attacks by exploiting vulnerabilities in ML models, leading to adversarial manipulation, data poisoning and model evasion techniques, which can compromise the integrity of UAV operations, lead to navigation errors, and expose sensitive data collected by UAVs. Considering the aforementioned security and privacy concerns, this review article presents emerging security and privacy solutions for UAVs in B5G networks. Firstly, We introduce the essential background of integrating UAVs into B5G networks and discuss the advantages and security challenges which the emerging integrated network architecture have. Then, we proceed to analyze and examine the security and privacy landscape by including threats and requirements of UAVs in B5G networks. Based on these threats and requirements, solutions from physical layer security (PLS), blockchain (BC), federated learning (FL) and post-quantum cryptography (PQC) are discussed and explored in details. Moreover, potential future research directions are discussed in details as open research issues. | 10.1109/TNSM.2024.3487265 |
Shimin Sun, Xiangyun Liu, Meiyu Wang, Ze Wang, Li Han | Predictive Control Plane Balancing in SD-IoT Networks Based on Elitism Genetic Algorithm and Non-Cooperative Game Theory | 2024 | Early Access | Load management Switches Internet of Things Genetic algorithms Prediction algorithms Heuristic algorithms Load modeling Artificial intelligence Quality of service Games Software Defined Internet of Thing (SD-IoTs) load balancing switch migration elitism genetic algorithm non-cooperative game theory | In the evolving landscape of Software Defined Internet of Thing (SD-IoT), the proliferation of IoT devices and applications has led to a drastic expansion of network traffic. Because of the overwhelming influx of control messages, the SDN controller may not have sufficient capacity to adequately address them. The main challenge is to properly deploy multiple controllers to enhance resource utilization and boost network performance, taking into account different factors for varying network scenarios. This paper presents a Predictive and Elitism genetic algorithm with Non-cooperative Game (PENG) strategy, tailored to address the control plane load imbalance. PENG incorporates a Gated Recurrent Units (GRU) based traffic prediction model, an improved elitism genetic algorithm, and the non-cooperative game theory to synergistically optimize the load balancing strategy. The study formulates a multi-objective optimization model that takes into account the degree of load balancing, control plane latency, and migration expenses as utility functions. This paper is structured around pivotal modules such as traffic prediction, controller overload identification, switch migration, and controller-switch mapping matrix reconfiguration. Moreover, an improved elitism genetic reallocation algorithm (IEGR) is designed, featuring a novel similarity factor to expedite convergence and improve the accuracy of identifying the optimal solution. Further, the detailed algorithm of PENG is present, outlining proactive and predictive switch migration to preemptively address potential load imbalance. The proposed methodology is simulated and the experimental results demonstrate that the proposal outperforms the comparisons in optimizing the load balancing degree, reducing average latency and migration cost. | 10.1109/TNSM.2024.3486379 |
He Bai, Hui Li, Jianming Que, Abla Smahi, Minglong Zhang, Peter Han Joo Chong, Shuo-Yen Robert Li, Xiyu Wang, Ping Lu | QSCCP: A QoS-Aware Congestion Control Protocol for Information-Centric Networking | 2024 | Early Access | Quality of service Protocols Resource management Packet loss Receivers Scalability Information-centric networking IP networks Diffserv networks Collaboration Information-Centric Networks Congestion Control Transport Protocol Quality of Service (QoS) | Information-Centric Networking (ICN) is a promising future network architecture that shifts the host-based network paradigm to a content-oriented one. Over the past decade, numerous ICN congestion control (CC) schemes have been proposed, tailored to address congestion issues based on ICN’s transmission characteristics. However, several key challenges still need to be addressed. One critical issue is that most existing CC studies for ICN do not consider the diverse Quality of Service (QoS) requirements of modern network applications. This limitation hinders their applicability across various applications with different network performance preferences. Another ongoing challenge lies in improving transmission performance, particularly considering how to appropriately coordinate congestion control participants to enhance content retrieval efficiency and ensure reasonable resource allocation, especially in multipath scenarios. To tackle these challenges, we propose QSCCP, a QoS-aware congestion control protocol built upon NDN (Named Data Networking), a well-known ICN architecture. In QSCCP, diverse QoS preferences of various traffic are supported within a collaborative congestion control framework. A novel multi-level, class-based scheduling and forwarding mechanism is designed to ensure varied and fine-grained QoS guarantees. A distributed congestion notification and precise feedback mechanism is also provided, which efficiently collaborates with an adaptive multipath forwarding strategy and consumer rate adjustment to rationally allocate network resources and improve transmission efficiency, particularly in multipath scenarios. Extensive experimental results demonstrate that QSCCP satisfies diverse QoS requirements while achieving outstanding transmission performance. It outperforms existing schemes in throughput, fairness, delay, and packet loss, with a rapid convergence rate and excellent stability. | 10.1109/TNSM.2024.3486052 |
Ting Yu, Zijian Gao, Kele Xu, Xu Wang, Peichang Shi, Bo Ding, Dawei Feng | Dual Temporal Masked Modeling for KPI Anomaly Detection via Similarity Aggregation | 2024 | Early Access | Time series analysis Anomaly detection Transformers Monitoring Frequency-domain analysis Data models Contrastive learning Reviews Interference Feature extraction Anomaly detection Key performance indicators Self-supervised learning Monitoring | With the expanding scale of current industries, monitoring systems centered around Key Performance Indicators (KPIs) play an increasingly crucial role. KPI anomaly detection can monitor the potential risks according to KPI data and has garnered widespread attention due to its rapid responsiveness and adaptability to dynamic changes. Considering the absence of labels and the high cost of manual annotation of KPI data, the self-supervised approaches are proposed. Among them, mask modeling methods draw great attention and can learn the intrinsic distribution of data without relying on prior assumptions. However, conventional mask modeling often overlooks the examination of relationships between unsynchronized variables, treating them with equal importance, and inducing inaccurate detection results. To address this, this paper proposes a Dual Masked modeling Approach combined with Similarity Aggregation, named DMASA. Starting from a self-supervised approach based on mask modeling, DMASA incorporates spectral residual techniques to explore inter-variable dependencies and aggregates information from similar data to eliminate interference from irrelevant variables in anomaly detection. Extensive experiments on eight datasets and state-of-the-art results demonstrate the effectiveness of our approach. | 10.1109/TNSM.2024.3486167 |
Reo Uneyama, Takehiro Sato, Eiji Oki | Flow Update Model Based on Probability Distribution of Migration Time in Software-defined Networks | 2024 | Early Access | Probability distribution Routing Packet loss Optimization Transient analysis Control systems Computational modeling Wide area networks Processor scheduling Maintenance Software-defined network network update problem capacity consistency probability distribution two-phase commit | In a software-defined network (SDN), routes of packet flows need to be updated in situations such as maintenance and router replacement. Each flow is migrated from its old path to new path. The SDN update has an asynchronous nature; the time when the switches process commands by the controller varies depending on flows. Therefore, it is difficult to control an order of flow migrations, and packets can be lost by congestion. Existing models divide the time axis into rounds and assign migrations to these rounds. However, congestion caused by multiple migrations in the same round is uncontrollable. Based on the probability distribution of time required for each migration, congestion can occur. This paper proposes a flow update model which minimizes the expected amount of excessive traffic by shifting the probability distributions. The time axis is divided into time slots which are fine-grained than rounds, so that each probability distribution is shifted. The proposed model assigns the time when the controller injects a command of flow migration to time slots. The proposed model is formulated as an optimization problem to determine the command times to minimize the expected amount. This paper introduces two methods to compute the expected amount. This paper also introduces a two-stage scheduling scheme (2SS) that divides the optimization problem into two stages. 2SS suppresses the computation time from O(|T||F|-1) to O(|T||F|-1/2) at the cost of including at most 0.12% error. 2SS suppresses the amount of excessive traffic than an existing model by at most 71.2%. | 10.1109/TNSM.2024.3485753 |
Redha A. Alliche, Ramon Aparicio-Pardo, Lucile Sassatelli | O-DQR: A Multi-Agent Deep Reinforcement Learning for Multihop Routing in Overlay Networks | 2024 | Early Access | Routing Training Topology Overlay networks Network topology Convergence Stability analysis Servers Optimization Delays Multi-Agent Deep Reinforcement Learning Distributed Packet Routing Overlay Networks Autonomous Network Control | This paper addresses the problem of dynamic packet routing in overlay networks using a fully decentralized MA-DRL. Overlay networks are built by having a virtual topology on top of an ISP underlay network, where those nodes are running a fixed, single path routing policy decided by the ISP. In such a scenario, the underlay topology and the traffic are unknown by the overlay network. In this setting, we propose O-DQR, which is an MA-DRL framework working under DTDE, where the agents are allowed to communicate only with their immediate overlay neighbors during both training and inference. We address three fundamental aspects for deploying such a solution: (i) performance (delay, loss rate), where the framework can achieve near-optimal performance, (ii) control overhead, which is reduced by enabling the agents to send control packets only when needed dynamically; and (iii) training convergence stability, which is improved by proposing a guided reward mechanism for dynamically learning the penalty applied when a packet is lost. Finally, we evaluate our solution through extensive experimentation in a realistic network simulation in both offline training and continual learning settings. | 10.1109/TNSM.2024.3485196 |
Hongping Gan, Hejie Zheng, Zhangfa Wu, Chunyan Ma, Jie Liu | TFD-Net: Transformer Deviation Network for Weakly Supervised Anomaly Detection | 2024 | Early Access | Anomaly detection Transformers Feature extraction Accuracy Training Data models Noise Knowledge engineering Time series analysis Computer architecture Weakly supervised anomaly detection Transformer imbalanced samples TFD-Loss | Deep Learning (DL)-based weakly supervised anomaly detection methods enhance the security and performance of communication and networks by promptly identifying and addressing anomalies within imbalanced samples, thus ensuring reliable communication and smooth network operations. However, existing DL-based methods often overly emphasize the local feature representations of samples, thereby neglecting the long-range dependencies and the prior knowledge of the samples, which imposes potential limitations on anomaly detection with a limited number of abnormal samples. To mitigate these challenges, we propose a Transformer deviation network for weakly supervised anomaly detection, called TFD-Net, which can effectively leverage the interdependencies and data priors of samples, yielding enhanced anomaly detection performance. Specifically, we first use a Transformer-based feature extraction module that proficiently captures the dependencies of global features in the samples. Subsequently, TFD-Net employs an anomaly score generation module to obtain corresponding anomaly scores. Finally, we introduce an innovative loss function for TFD-Net, named Transformer Deviation Loss Function (TFD-Loss), which can adequately incorporate prior knowledge of samples into the network training process, addressing the issue of imbalanced samples, and thereby enhancing the detection efficiency. Experimental results on public benchmark datasets demonstrate that TFD-Net substantially outperforms other DL-based methods in weakly supervised anomaly detection task. | 10.1109/TNSM.2024.3485545 |
Chang-Lin Chen, Hanhan Zhou, Jiayu Chen, Mohammad Pedramfar, Tian Lan, Zheqing Zhu, Chi Zhou, Pol Mauri Ruiz, Neeraj Kumar, Hongbo Dong, Vaneet Aggarwal | Learning-Based Two-Tiered Online Optimization of Region-Wide Datacenter Resource Allocation | 2024 | Early Access | Servers Resource management Heuristic algorithms Hardware Dynamic scheduling Optimization Numerical models Decision trees Containers Business Cloud Computing Capacity Reservation Deep Reinforcement Learning Explainable Reinforcement Learning | Online optimization of resource management for large-scale data centers and infrastructures to meet dynamic capacity reservation demands and various practical constraints (e.g., feasibility and robustness) is a very challenging problem. Mixed Integer Programming (MIP) approaches suffer from recognized limitations in such a dynamic environment, while learning-based approaches may face with prohibitively large state/action spaces. To this end, this paper presents a novel two-tiered online optimization to enable a learning-based Resource Allowance System (RAS). To solve optimal server-to-reservation assignment in RAS in an online fashion, the proposed solution leverages a reinforcement learning (RL) agent to make high-level decisions, e.g., how much resource to select from the Main Switch Boards (MSBs), and then a low-level Mixed Integer Linear Programming (MILP) solver to generate the local server-to-reservation mapping, conditioned on the RL decisions. We take into account fault tolerance, server movement minimization, and network affinity requirements and apply the proposed solution to large-scale RAS problems. To provide interpretability, we further train a decision tree model to explain the learned policies and to prune unreasonable corner cases at the low-level MILP solver, resulting in further performance improvement. Extensive evaluations show that our two-tiered solution outperforms baselines such as pure MIP solver by over 15% while delivering 100× speedup in computation. | 10.1109/TNSM.2024.3484213 |
Bing Tang, Zhikang Wu, Wei Xu, Buqing Cao, Mingdong Tang, Qing Yang | TP-MDU: A Two-Phase Microservice Deployment Based on Minimal Deployment Unit in Edge Computing Environment | 2024 | Early Access | Microservice architectures Optimization Quality of service Dynamic scheduling Servers Reinforcement learning Resource management Cloud computing Time factors Load modeling mobile edge computing microservices minimal deployment unit two-phase deployment reinforcement learning | In mobile edge computing (MEC) environment, effective microservices deployment significantly reduces vendor costs and minimizes application latency. However, existing literatures overlook the impact of dynamic characteristics such as the frequency of user requests and geographical location, and lack in-depth consideration of the types of microservices and their interaction frequencies. To address these issues, we propose TP-MDU, a novel two-stage deployment framework for microservices. This framework is designed to learn users’ dynamic behaviors and introduces, for the first time, a minimal deployment unit. Initially, TP-MDU generates minimal deployment units online, tailored to the types of microservices and their interaction frequencies. In the initial deployment phase, aiming for load balancing, it employs a simulated annealing algorithm to achieve a superior deployment plan. During the optimization scheduling phase, it utilizes reinforcement learning algorithms and introduces dynamic information and new optimization objectives. Previous deployment plans serve as the initial state for policy learning, thus facilitating more optimal deployment decisions. This paper evaluates the performance of TP-MDU using a real dataset from Australia’s EUA and some related synthetic data. The experimental results indicate that TP-MDU outperforms other representative algorithms in performance. | 10.1109/TNSM.2024.3483634 |
Lifan Mei, Jinrui Gou, Jingrui Yang, Yujin Cai, Yong Liu | On Routing Optimization in Networks With Embedded Computational Services | 2024 | Early Access | Routing Computational modeling Delays Optimization Heuristic algorithms Servers Load modeling Resilience Performance evaluation Resource management Routing Edge Computing In-Network Computation Network Function Virtualization | Modern communication networks are increasingly equipped with in-network computational capabilities and services. Routing in such networks is significantly more complicated than the traditional routing. A legitimate route for a flow not only needs to have enough communication and computation resources, but also has to conform to various application-specific routing constraints. This paper presents a comprehensive study on routing optimization problems in networks with embedded computational services. We develop a set of routing optimization models and derive low-complexity heuristic routing algorithms for diverse computation scenarios. For dynamic demands, we also develop an online routing algorithm with performance guarantees. Through evaluations over emerging applications on real topologies, we demonstrate that our models can be flexibly customized to meet the diverse routing requirements of different computation applications. Our proposed heuristic algorithms significantly outperform baseline algorithms and can achieve close-to-optimal performance in various scenarios. | 10.1109/TNSM.2024.3483088 |