With the proliferation of various mobile smart devices, edge computing and computational offloading technologies have emerged as pivotal support mechanisms, enhancing the service quality of these devices. To facilitate the learning of task offloading strategies across diverse and complex scenarios, this study introduces a federated learning strategy algorithm based on multi-agent deep reinforcement learning. This algorithm aims to aggregate the training strategies of multiple edge computing devices, thereby enabling the synthesis of superior task offloading strategies tailored to a wide array of environments while also providing safeguards against malicious nodes. This paper specifically examines the reward inversion attack, illustrating the algorithm’s capability to identify and counteract such malicious threats effectively. Experimental results validate that the proposed algorithm not only robustly defends against these attacks but also adeptly learns the task offloading strategies pertinent to each node.
Due to the open characteristics of the electromagnetic channel, there are malicious nodes that jam the normal data flow, prevent the legitimate receiver from obtaining information, and then intercept and tamper with the data, which makes the research of communication anti-jamming becomes more and more important. Traditional anti-jamming methods employ a single anti-jamming approach and cannot adaptively adjust their anti-jamming strategy based on the environment, often can’t achieve a good anti-jamming effect in the complex communication environment. In order to solve these challenges, this paper studies the anti-jamming communication model based on deep reinforcement learning (DRL), and builds a simulation system to realize intelligent anti-jamming decision-making by using DRL algorithm. The simulation results show that the proposed intelligent anti-jamming decision can choose the best anti-jamming scheme according to the complex environment and effectively improve the communication quality.
HF diversity communication network adopts multiple frequencies to support user communication to improve the reliability of receiving.In the case of multi-user concurrent access,genetic algorithm can be used to plan the frequency of users,so as to save frequency resources under the condition of meeting the needs of given users.However,the planning process takes a long time and the optimization effect is not good.To address these problems,a new improved genetic algorithm is proposed to plan frequency resources of HF diversity communication network.The simulation results indicate that the improved algorithm has shorter planning time and better optimization effect than the original genetic algorithm,which provides a basis for the practical application of HF diversity communication network.
Due to its high flexibility and other advantages,aerial drone base stations can expand the communication support range for ground users and solve the mobile coverage problem of ground mobile users in battlefield environments.Considering the impact of high-level terrain features on radio signals in the battlefield environment,the differentiated communication needs of users,and the energy consumption issues of UAV(Unmanned Aerial Vehicle)base stations,a reinforcement learning communication coverage algorithm based on heterogeneous mobile users(ABS-RL)is proposed to model the channel capacity of ground users,the energy consumption of UAV,and the dynamic deployment of UAV base stations,aiming to provide high-quality communication services for ground users.The simulation results demonstrate that the algorithm offers significant advantages in enhancing the channel capacity for ground users and reducing the total energy consumption of UAV base stations.
In order to achieve the long-term work of the wireless sensor network in the border area,a network combining UAV and NOMA(Non-Orthogonal Multiple Access)technology is used to collect data from sensors distributed along the narrow border,so as to effectively reduce the energy consumption of WSNs(Wireless Sensor Networks).The sensor user group uploads data to UAV in a multi-carrier NOMA mode to minimize the total energy consumption of all sensors by optimizing sensor user scheduling and grouping,transmit power and UAV trajectory under the constraints of maximum transmit power,user QoS(Quality of Service)and UAV mobility.The optimization problem is a mixed integer nonlinear optimization problem,which is decomposed into three subproblems to solve.By using exhaustive search,Lagrange duality decomposition and SCA(Successive Convex Approximation),the subproblems are solved successively,and the minimum energy consumption is achieved.Simulation results indicate that the proposed scheme can effectively reduce the total energy consumption of the WSNs.
A new secure vTPM virtual machine live migration protocol and a practical implementation method are proposed to address the lack of security considerations in the implementation of vTPM(virtual trusted platform Module)live migration in the open source component QEMU under the current KVM(Kernel-based Virtual Machine)platform.First,a vTPM virtual certificate chain extension method based on pTPM(physical Trusted Platform Module)is designed to avoid key regeneration after vTPM migration.Then,this key structure is used to build a secure vTPM live migration protocol for private cloud scenarios,which effectively reduces denial-of-service attacks during the vTPM migration phase.In addition,the bidirectional remote trusted proof method in the proposed protocol and the security protection during the vTPM data migration process are all implemented based on pTPM,which can provide hardware based trust root security strength for vTPM live migration.
To address the end-to-end secure transmission problem in fiber optic cable networks,this paper proposes a scheme of anti-interception transmission based on QNRC(Quantum Noise Randomized Cipher).First,this paper introduces the encryption principle and implementation method of QNRC,then it takes the method of optical domain decryption and coherent detection to experimentally realize a PSK-QNRC secure transmission system with 10 Gbps rate at 50 km transmission distance.Experimental results indicate that the proposed scheme is highly feasible and can ensure the high-speed transmission with high security between the legitimate sender and receiver.
Due to its characteristics of determinacy,low-latency,and high-reliability,time-sensitive networks will be used as a new type of network infrastructure for industrial internet and other application scenarios,and its security is crucial.Currently,MACsec encryption is the main protection measure,but there is the problem that hop-by-hop encryption introduces a large cumulative delay as well as the leakage of sensitive data information brought about by intermediate node decryption realization.Based on the concept of software definition,this paper constructs a time-sensitive network security framework to minimize the introduced encryption latency through centralized security configuration management and integrated network and encryption design.It analyzes a security application scenario for a typical multi-service shared network transmission and some of the advantages it brings.The relevant design ideas and methods can provide a reference for building secure time-sensitive networks.
At present,typical homomorphic cryptography algorithms such as BGV scheme,BFV scheme,CKKS scheme etc.have large resource consumption and slow performance in polynomial ring multiplication,which is the difficulty in efficient hardware implementation of homomorphic cryptography algorithms.The fast NTT(Number Theoretic Transforms)algorithm is usually used to accelerate the design.By analyzing the key issues in the hardware design of resource intensive NTT algorithms,as well as optimizing the typical NTT algorithm process,this paper proposes two kinds of hardware design schemes:multi-cycle parallelization and single butterfly unit pipelining.It focuses on analyzing the core points of pipelining design and gives the overall structure of the module and simulation verification results.Combined with performance and resource consumption evaluation,this provides a reference for practical hardware design of different parameters in resource intensive NTT algorithms.Research shows that in a trade-off between resources and performance,the use of pipelining design has a higher resource performance ratio and is the preferred choice.
The power supply and distribution system of airborne mission equipment must maintain high reliability and high power quality under any working conditions.However,with the widespread use of high-power nonlinear electrical equipment,generator voltage and current waveforms can become distorted,resulting in power quality degradation.In addition,problems often occur in areas such as compatibility of the airborne power supply system with electrical equipment.In view of these problems,this paper briefly describes the development and classification of the output categories of airborne power supply systems,summarizes the scheme design of power distribution architecture and rectification technology for electrical equipment,elucidates the contents and interrelationships of power supply adaptability characteristic standards,testing standards,and testing method standards,clarifies the adaptability design principles of the electrical equipment,and analyzes the advantages and disadvantages of pulse load power matching solutions for airborne generators,which provides some important guiding roles for the design and testing of the new generation airborne power supply systems and electrical equipments.
In recent years,the design of linear precoding algorithms for MU-MIMO(Multi-User Multiple-Input Multiple-Output)downlink systems attracts increasing interest from researchers.However,currently there is no research on linear precoding algorithms under the constraint of PAPCS(Per-Antenna Power Constraints)with known channel error probability distribution at the base station.To address this issue,linear precoding algorithms based on different optimization criteria are proposed,considering the objectives of traversal and maximum expected sum rate.Some of these algorithms are designed based on CSSCA(Constrained Stochastic Successive Convex Approximation),second-order duality,ADMM(Alternating Direction Method of Multipliers),and Gaussian randomization techniques.These algorithms are more suitable for practical applications and perform well through experimental simulations.
To address the common issue of imbalanced malicious traffic samples in network traffic detection,this paper proposes a classification method based on data augmentation and ensemble learning.The approach first employs K-means + SMOTE data augmentation to balance data samples of different categories,and then uses an ensemble learning model to enhance the generalization capability of the classification model.Experimental results on two publicly available datasets indicate that when using XGBoost for binary and multi-class classification,the accuracy reaches 99.1%and 97.19%,respectively.Compared to models such as CNN,Random Forest,and LightGBM,the approach proposed in this paper consistently shows significant performance advantages.
To address the problem of biased results caused by most deep learning algorithms using a single modality for classification,this paper proposes a hybrid neural network based on bimodal features.The method can use two different modalities to train the classification model and improve the accuracy of the classification model.First,the payload features of the transport layer traffic data packets are used as the packet-level modals and the length sequence features of the data packets are used as the flow-level modals.Then,a neural network is used on two paths to analyze the bimodal features,and the high-dimensional features extracted from the two paths are fused.Finally,the classification results of the model are output.Two public datasets are used to train and test the model.Experimental results indicate that the classification precision of the bimodal model reaches 96.46%and 93.01%,respectively,which is significantly improved compared with the current four excellent single-modal and multi-modal method.
Existing algorithms for determining cluster heads by establishing mathematical models are too complex and inflexible due to the high speed movement of vehicles and frequent changes in network topology in in-vehicle networks.This paper proposes a scheme for selecting the optimal cluster head based on fuzzy logic in VANET clustering,which uses the relative average speed among vehicles,adjacency,and RSU link quality as the metrics,and selects the most optimal cluster head based on fuzzy logic,which increases the flexibility of cluster head selection.The scheme selects adjacent vehicles as cluster heads,which can improve the speed and quality of intra-cluster communication and effectively reduce the latency.The performance simulation comparison with CROWN algorithm indicates that the proposed scheme has better service discovery latency and service consumption latency.
A device fingerprint generation method based on multidimensional feature dynamic weighting is proposed to address the issues of conventional device type recognition methods,such as incomplete data feature extraction,difficult to cover various platform device types,and some fingerprints may undergo slight changes with temperature,humidity,etc.,which makes it difficult to accurately recognize the device identity.First,multi-dimensional static and dynamic feature information of the device is extracted through a plug-in approach.Then,based on the expert library dynamic weighting technology,the feature value mapping is completed to make it robust.Finally,a secure and lightweight summary value algorithm is selected to encrypt the feature values to generate device fingerprints,so that they are secure,real-time,stable,and unforgeable,and can represent the identity of the platform device.
At present,FPGA(Field-Programmable Gate Array)module verification has problems such as high complexity,large scale,and high time-consuming,etc.In response to these problems,this paper proposes an efficient FPGA module verification model.The model adopts a flow line design that includes five phases:SVS,ALV,DLV,BLV,and DM,each of which elaborates the specific verification methods,test types,and rationalization steps.Among them,the efficiency of the model is discussed in detail.In addition,functional simulation and physical testing methods are employed to verify the optimization means.Finally,a large amount of sample data is used to summarize and analyze the high-efficiency model time-consuming reduction ratio α of 10 000 lines and defect rate β of 10 000 lines.It is concluded from the experimental data that the proposed verification model has complete structure,high reliability,strong pertinence,and high verification efficiency.
With the development of information countermeasure,both the enemy and we begin to use a large amount of radar and interference equipment,electromagnetic interference environment is becoming increasingly complex,and interference suppression becomes a problem that needs to be addressed.Interference suppression is usually achieved through the use of interference cancellation techniques and spatial adaptive interference suppression techniques.In response to the problem of needing to know the DOA(Direction of Arrival)of signal sources in spatial adaptive interference,this paper proposes the use of spatial spectrum estimation direction finding techniques for DOA estimation before adaptive interference suppression.This approach does not add additional channels and hardware resources to the system,and it can simultaneously estimate the direction of multiple signal sources with high direction finding accuracy and good stability.The system can effectively eliminate multiple interference components in the received signal and restore the original communication signal without requiring auxiliary channels and a priori information about other interference sources.
The increased scale and complexity of information systems poses challenges for the service,organization,and use of edge environments.In order to bring more services with guaranteed quality to various terminal devices,the edge cloud is constructed through intensive use of server resources in the edge environment,and the service capability of the"cloud"is extended to the edge of the network.Then,in order to meet the application requirements of high mobility scenes,and support all kinds of terminal devices to access the edge cloud to enjoy"cloud"and"edge"information services to carry out relevant businesses,this paper proposes the terminal on-demand access to edge cloud method under highly mobile conditions of the edge.This technology can provide a highly available terminal adaptive discovery access to edge cloud method for highly mobile communication environments,and can support terminal applications on edge cloud platforms.
Plain-ciphertext recognition is the basis of cryptanalysis and cryptographic supervision.Thus,this paper proposes an information entropy-based plain-ciphertext recognition algorithm.N-truncated entropy is used to calculate the confidence interval,and by inputting the file to be detected,it is possible to determine whether the file is encrypted or not.Experimental results indicate that the accuracy of the algorithm reaches 99%.It can be applied in any scenario that requires judgment of plain-ciphertext.
As computing power and other resources of wireless sensor network are limited,it is always a research hotspot that using less computing to achieve authentication and key agreement between nodes to ensure communication security.To address the problems of poor network scalability and difficult key updating of symmetric cryptography-based authentication schemes,as well as the high computational resource overhead of asymmetric cryptography-based authentication schemes,an identity-based non-two-party authentication and key agreement scheme is proposed to meet the security requirements of WSNs.The scheme is based on the ECC(Elliptic Curve Cryptography)algorithm which realizes two-party authentication,session key agreement,confirmation and update between communication nodes.The analysis indicates that the scheme can satisfy several typical security attributes required for the key agreement process of WSN nodes and is optimized in terms of resource consumption.
