In this study, to alleviate the inconsistency between dimensionality reduction and feature retention in anomaly detection, we propose a Variational Long Short-Term Memory (VLSTM) learning model for intelligent anomaly detection based on reconstructed feature representation. A compressed network associated with a variational reparameterization scheme is designed to learn the low-dimensional feature representation from high-dimensional raw data. Three loss functions are defined and quantified to constrain the reconstructed hidden variable into a more explicit and meaningful way. A lightweight estimation network is fed with the refined feature representation to identify anomalies in IBD. Experiments using a public IBD dataset UNSW-NB15 demonstrate that the proposed VLSTM model can significantly improve F1, AUC, and FAR scores in anomaly detection.

Intelligent anomaly detection for identifying cyber/physical attacks to guarantee the work efficiency and safety is still a challenging issue, especially when dealing with few labeled data for cyber-physical security protection. In this study, we propose a few-shot learning model based on Siamese Convolution Neural Network (FS-SCNN), to alleviate the over-fitting issue and enhance the accuracy for intelligent anomaly detection in industrial CPSs. A Siamese CNN encoding network is constructed to measure distances of input samples based on their optimized feature representations. A robust cost function design including three specific losses is then proposed to enhance the efficiency of training process. An intelligent anomaly detection algorithm is developed finally. Experiment results based on a fully labeled public dataset and a few labeled dataset demonstrate that our proposed FS-SCNN can significantly improve FAR and F1 scores when detecting intrusion signals for industrial CPS security protection.

With the flourish of digital technologies and rapid development of 5G and beyond networks, Metaverse has become an increasingly hotly discussed topic, which offers users with multiple roles for diversified experience interacting with virtual services. How to capture and model users’ multi-platform or cross-space data/behaviors become essential to enrich people with more realistic and immersed experience in Metaverse-enabled smart applications over 5G and beyond networks. In this study, we propose a Personalized Federated Learning with Model-Contrastive Learning (PFL-MCL) framework, which may efficiently enhance the communication and interaction in human-centric Metaverse environments by making use of the large-scale, heterogeneous, and multi-modal Metaverse data. Differing from the conventional Federated Learning (FL) architecture, a multi-center aggregation structure to learn multiple global models based on the changes of dynamically updated local model weights, is developed in global, while a hierarchical neural network structure which includes a personalized module and a federated module to tackle both issues on data heterogeneity and model heterogeneity, is designed in local, so as to enhance the performance of PFL with unique characteristics of Metaverse data. In particular, a two-stage iterative clustering algorithm with a more precise initialization is developed to facilitate the personalized global aggregation with dynamically updated multiple aggregation centers. A personalized multi-modal fusion network is constructed to greatly reduce the computational cost and feature dimensions from the high-dimensional heterogeneous inputs for more efficient cross-modal fusion, based on a hierarchical shift-window attention mechanism and a newly designed bridge attention mechanism. A MCL scheme is then incorporated to speed up the model convergence with less communication overload between the local federated module and global model, while an embedding layer which effectively enables the delivered global model to better adapt to the local personality in each client is further integrated. Compared with five baseline methods, experiment and evaluation results based on two different real-world datasets demonstrate the excellent performance of our proposed PFL-MCL model in a fine-grain personalized training strategy, toward more efficient communication and networking among human-centric Metaverse enabled smart applications.

Vehicular Social Networks (VSNs) have attracted the research community due to its diverse applications ranging from safety to entertainment. Social vehicles standing for private cars and floating cars standing for taxis are two important components of VSN. However, the lack of social vehicles data causes some factors are neglected including social aspects and macroscopic features, that blocks researching social attributes of vehicles in VSN. Generating a realistic mobility dataset for VSN validation has been a great challenge. In this paper, we present the detailed procedure to generate social vehicular mobility dataset from the view of floating car data, which has the advantage of wide universality. Firstly, through the deep analysis and modeling of the dataset of floating cars and combining with the official data, we predict the Origin-Destination (OD) Matrix of social vehicles with the Gravity Model, and then calibrate the OD matrix with the Average Growth Factor Method. Secondly, we construct network description after editing the road network. Thirdly, we use Simulation of Urban Mobility (SUMO) to reproduce the scenario in view of micro-simulation by generating the mobility dataset of social vehicles based on floating car data and urban functional areas. At last, we prove the effectiveness of our method by comparing with real traffic situation in Beijing. The generated mobility model may not accurately represent the mobility of social vehicles in few spots, such as train station or airport, however, exploiting figures and facts of transportation in the city have been considered in the study to calibrate the model up to maximum possible realization.

In this study, a 2-Dimensional Federated Learning (2DFL) framework, including the vertical and horizontal federated learning phases, is designed to cope with the insufficient training data and insecure data sharing issues in CPSS during a secure distributed learning process. Considering a specific application of Human Activity Recognition (HAR) across a variety of different devices from multiple individual users, the vertical federated learning scheme is developed to integrate shareable features from heterogeneous data across different devices into a full feature space, and the horizontal federated learning scheme is developed to effectively aggregate the encrypted local models among multiple individual users to achieve a high-quality global HAR model. A computationally efficient somewhat homomorphic encryption (SWHE) scheme is then improved and applied to support the parameter aggregation without giving access to it, which enables heterogeneous data sharing with privacy protection across different personal devices and multiple users in building a more precise personalized HAR model. Experiments are conducted based on two public datasets. Comparing with three conventional machine learning methods, evaluation results demonstrate the usefulness and effectiveness of our proposed model in achieving faster and smoother convergence, and better precision, recall, and F1 scores for HAR applications in CPSS.