Mobile Networks and Applications最新文献

The number of terminal devices has skyrocketed along with the quick growth of cognitive radio networks. Massive equipment produce a lot of data that should not be shared, often WiFi signals. The radio frequency (RF) fingerprint identification approach for WiFi signals proposed in this research is based on federated learning and trains a collaborative model to complete RF fingerprint without transferring privacy-sensitive data. Aiming at the lack of labeled data and heterogeneous distribution of labeled data in actual situations, a federated transfer learning mechanism is designed. The technique suggested in this paper increases the accuracy of RF fingerprint at various sizes and assures that data privacy is not compromised, according to experimental results on real-world datasets.

Accurate positioning in the constrained environment of Global Navigation satellite Systems (GNSS) is a challenging problem, especially in resource-constrained urban canyon environments. In order to incentivize collaborative agency, this paper, grounded in an economic framework, proposes the utilization of auction mechanisms to address issues pertaining to collaboration and power allocation among agents. For different types of agents, different auction methods are designed according to their own resources for collaborative positioning. Firstly, an Iterative Bidirectional Auction (IBA) cooperative localization algorithm is proposed to solve the problem of cooperation and power allocation among agents in resource-constrained environments. Secondly, in order to ensure the fairness of power distribution, the auction reserve price is introduced, and the relationship between the auction reserve price and power distribution is deduced. Then, considering that there are different types of agents in the actual scenario, One-Shot Auction (OSA) algorithm is proposed to realize the cooperation between user agents and vehicle agents. Finally, analysis and numerical results demonstrate that under the proposed collaborative strategy, agents with better network conditions are more likely to participate in cooperation. Compared to non-cooperative positioning (NC), each agent experiences an improvement in position accuracy of over 60%. The performance of the proposed algorithm is approximately 43% better than uniform power allocation (UPA), and the position accuracy approaches that of the full power allocation (FPA) algorithm. Our algorithm outperforms OSA, PAR and BACL in positioning accuracy with the same agent nodes, and is the most power-efficient. This is pivotal for collaborative positioning under resource constraints.

In addition to a large amount of text, there are also many emoticons in the comment data on social media platforms. The multimodal nature of online comment data increases the difficulty of sentiment analysis. A big data sentiment analysis technology for social online multimodal (SOM) comments has been proposed. This technology uses web scraping technology to obtain SOM comment big data from the internet, including text data and emoji data, and then extracts and segments the text big data, preprocess part of speech tagging. Using the attention mechanism-based feature extraction method for big SOM comment data and the correlation based expression feature extraction method for SOM comment, the emotional features of SOM comment text and expression package data were obtained, respectively. Using the extracted two emotional features as inputs and the ELMO pre-training model as the basis, a GE-Bi LSTM model for SOM comment sentiment analysis is established. This model combines the ELMO pre training model with the Glove model to obtain the emotional factors of social multimodal big data. After recombining them, the GE-Bi LSTM model output layer is used to output the sentiment analysis of big SOM comment data. The experiment shows that this technology has strong extraction and segmentation capabilities for SOM comment text data, which can effectively extract emotional features contained in text data and emoji packet data, and obtain accurate emotional analysis results for big SOM comment data.

In the rapidly evolving landscape of wireless communication systems, the vulnerability of automatic modulation classification (AMC) models to adversarial attacks presents a significant security challenge. This study introduces a pruning and training methodology tailored to address the nuances of signal processing within these systems. Leveraging a pruning method based on channel activation contributions, our approach optimizes adversarial training potential, enhancing the model’s capacity to improve robustness against attacks. Additionally, the approach constructs a resilient training method based on a composite strategy, integrating balanced adversarial training, soft target regularization, and gradient masking. This combination effectively broadens the model’s uncertainty space and obfuscates gradients, thereby enhancing the model’s defenses against a wide spectrum of adversarial tactics. The training regimen is carefully adjusted to retain sensitivity to adversarial inputs while maintaining accuracy on original data. Comprehensive evaluations conducted on the RML2016.10A dataset demonstrate the effectiveness of our method in defending against both gradient-based and optimization-based attacks within the realm of wireless communication. This research offers insightful and practical approaches to improving the security and performance of AMC models against the complex and evolving threats present in modern wireless communication environments.

The Handwritten Mathematical Expression Recognition (HMER) task aims to generate corresponding LATEX sequences from images of handwritten mathematical expressions. Currently, the encoder-decoder architecture has made significant progress in this task. However, the architecture based on the DenseNet encoder fails to adequately consider the unique features of handwritten mathematical expressions (HME) and the similarity between different characters. Additionally, the decoder, with its small receptive field during the decoding process, fails to effectively capture the spatial positional information of the targets, resulting in a lack of global contextual information during decoding. To address these issues, this paper proposes a neural network called DGNet based on deformable convolution and global contextual attention. Our network takes into full consideration the sparse nature of handwritten mathematical formulas and utilizes the properties of deformable convolution, allowing the convolution kernel to deform based on the content of the neighborhood. This enables our model to better adapt to geometric changes and other deformations in handwritten mathematical expressions. Simultaneously, we introduce GCAttention in optimizing the feature part to fully aggregate global contextual features of both position and channel. In experiments, our model achieved accuracies of 58.51%, 56.32%, and 56.1% on the CROHME 2014, 2016, and 2019 datasets, respectively.

Federated Learning (FL) revolutionizes data processing in mobile networks by enabling collaborative learning without data exchange. This not only reduces latency and enhances computational efficiency but also enables the system to adapt, learn and optimize the performance from the user’s context in real-time. Nevertheless, FL faces challenges in training and generalization due to statistical heterogeneity, stemming from the diverse data nature across varying user contexts. To address these challenges, we propose (textsf {WAFL}), a robust FL framework grounded in Wasserstein distributionally robust optimization, aimed at enhancing model generalization against all adversarial distributions within a predefined Wasserstein ambiguity set. We approach (textsf {WAFL}) by formulating it as an empirical surrogate risk minimization problem, which is then solved using a novel federated algorithm. Experimental results demonstrate that (textsf {WAFL}) outperforms other robust FL baselines in non-i.i.d settings, showcasing superior generalization and robustness to significant distribution shifts.

Cognitive Internet of Things (CIoT) is a new subfield of the Internet of Things (IoT) that aims to integrate cognition into the IoT's architecture and design. Various CIoT applications require techniques to inevitably extract machine-understandable concepts from unprocessed sensory data to provide value-added insights about CIoT devices and their users. The time series classification, which is used for the concept's extraction poses challenges to many applications across various domains, i.e., dimensionality reduction strategies have been suggested as an effective method to decrease the dimensionality of time series. The most common approach for time-series classification is the symbolic aggregate approximation (SAX). However, its main drawback is that it does not select the most significant point from the segment during the piecewise aggregate approximation (PAA) stage. The situation is cumbersome when data is heterogeneous and massive. Therefore, this research presents a novel technique for the selection of the most significant point from a segment during the PAA stage in SAX. The proposed technique chooses the maximum informative point as the most significant point using the probabilistic interpretation of sensory data with an appropriate copula design. The appropriate copula is selected using the minimum akaike information criteria (AIC) value. Subsequently, the modified SAX considers the maximum informative points instead of a selection of mean/max/extreme data points on a given segment during the PAA stage. The experimental evaluation of the environmental dataset reveals that the proposed method is more accurate and computationally efficient than classic SAX. Also, for cross-validation it computes the entropy of the information point (i-value) from each dataset to verify the successful transformation of normal data points to information points.

In millimeter wave (mmWave) communication, the utilization of multi-hop relay technology has been regarded as a promising approach to overcome the significant path loss encountered during signal transmission. However, the traditional active relay network suffers from low energy efficiency (EE) and uneven resource distribution. To address these challenges, we introduce Reconfigurable Intelligent Surface (RIS) as a passive relay to the mmWave communication system and create a hybrid relay system that combines passive and active relays, which aims to improve EE through the multi-hop relay. Additionally, with the development of Artificial Intelligence, deep Q-learning (DQN) is applied to optimize the hybrid relay system in this paper, where every transmitted signal of the base station (BS) is considered an agent. In this approach, the network is trained based on the interaction between the collected environment information and the users’ relay allocation strategy. Considering competition-cooperation relationships of multiple users, we propose a multi-agent DQN (MADQN) algorithm to allocate the relay resource where the primary goal is maximizing EE. Simulation results demonstrate that our proposed scheme can effectively converge to the optimal relay link, further improving EE and reducing energy consumption in comparison with conventional schemes.

With the development and adoption of cloud platforms in various fields, malware attacks have become a serious threat to the Internet cloud ecosystem. However, the pooling process of existing graph classification techniques for malware variant detection uses only a serial and single strategy, resulting in localized malicious behaviors of malware that may be overlooked. In this paper, we propose MalDMTP, a malware detection framework based on multilevel graph classification learning, which implements the graph pooling process for malware classification in parallel and performs graph instance-based discrimination. In particular, MalDMTP first constructs an API call graph based on results obtained from dynamic execution of malware. Then it combines multiple graph neural network learning strategies through multi-level pooling to learn the global importance of nodes in the pooled graph and extract node representations from multiple perspectives for heterogeneous graphs. After that, MalDMTP is aggregated into graph representations by the graph-level pooling function GMT based on a multi-head attention mechanism, which goes through a classifier in order to obtain malware prediction labels. Experimental results show that the proposed MalDMTP can achieve 96.53% accuracy on the Alibaba cloud malware dataset, which improves 1.9% 7.6% over the previous single-graph pooling methods on the graph classification task of malware detection.

To enhance task completion efficiency and quality, the coordination of Unmanned Surface Vehicle (USV) formations in complex environmental situations often requires user intervention. This paper proposes a human-machine collaborative approach for USV mission planning and explores a method for identifying user intervention intentions. A method for recognizing user intention based on intervention style was proposed. The method utilizes the Improved Particle Swarm Optimization-Support Vector Machine (IPSO-SVM) model to recognize intervention style and emphasizes human intention recognition to enhance the ability of USV in complex environments. The method involves modeling continuous intervention operations and incorporating intervention style features to accurately identify user intent. The study proposes a fusion method that combines feature attention, self-attention, and Fusion of Long Short-Term Memory Networks (FLSTMS) to achieve its purpose. Furthermore, it suggests a cooperative mission planning method based on prospect theory, which integrates user risk propensity and identified intentions to optimize planning. Simulation experiments confirm the effectiveness of this approach, highlighting its advantages over traditional methods.