Federated Tensor Decomposition-Based Feature Extraction Approach for Industrial IoT

Abstract : Data in modern industrial applications and data science presents multidimensional progressively, the dimension and the structural complexity of these data are becoming extremely high, which renders existing data analysis methods and machine learning algorithms inadequate to the extent. In addition, high-dimensional data in actual scenarios often share some common latent components and patterns, it is necessary and significant to analyze such data in an associative manner, rather than treating them independently. Considering the problem of data islands and data privacy that is prevalent in the industry. In this paper, we propose the first joint orthogonal iterative algorithm (J-HOOI) for simultaneous tensor decomposition and federated tensor decomposition model (FTD) for feature extraction and dimension reduction of high-dimensional industrial data under the federated learning framework. Moreover, we also develop a secure federated computation process based on the joint orthogonal iteration method (J-HOOI). Using this method, multiple participants iteratively calculate the local factor matrices and transfer the local information to the parameter server, which aggregates the local information to generate the globally updated factor matrices. Finally, each client generates globally compressed features by projecting local data onto these common potential spaces. We have demonstrated with real-world industrial data that our approach is similar to a centralized training model in decomposition accuracy and classification accuracy while respecting privacy.

 EXISTING SYSTEM :

 ? Existing IoT vehicular systems are based on centralization of data in which each vehicle uploads its data to a central cloud which puts a limit on amount on data being shared as there is a risk of data leakage. ? Storing QoI indicators along with existing industry metadata of participating organizations can further enhance the ability to build and update appropriate FL models. ? However, existing security mechanisms do not focus on designing the effective framework and blockchain’s reliability issue. ? The revised FL model and the implemented adversarial learning loss function, the proposed strategy can metabolise DNNs to successfully handle existing adversarial threats while guaranteeing privacy protection among various IIoT devices.

 DISADVANTAGE :

 ? An FL-empowered task offloading problem with mobile edge computing (MEC) has been also analyzed in. ? To be clear, each client selects a learning strategy for solving its local sub-problem to ensure desired accuracy with lowest participation costs, while the central server builds a utility function by averaging local updates to offer reward to the clients. ? The potential of FL has been investigated in solving distributed binary supervised classification problems to estimate hospitalizations for cardiac events. ? FL can help solve these problems by allocating intelligence to robotic devices, instead of relying on a remote server for data processing.

 PROPOSED SYSTEM :

 • A cross domain sharing based scheduling scheme was proposed for providing edge services with higher intelligence. • An optimization technique using particle swarm optimization (PSO) and FL is proposed in focusing on the hyper-parameter tuning for the DL models that are available locally in the smart city applications. • This work proposed a model which used distributed and hierarchical FL for offering optimized resource communication and robustness. • This work proposed a novel deep federated reinforcement learning (DFRL) scheme for providing dynamic resource allocation and network management for the IIoT networks.

 ADVANTAGE :

 ? The federation of mobile devices helps eliminate the need for a centralized data processing architecture, instead of performing data training locally using their own dataset without degrading learning performances and compromising user privacy values. ? Based on the proposed FL scheme, the offloading performance can be improved in terms of better accuracy compared to the heterogeneity-unaware equal task allocation approach. ? Implementation results indicate a good performance in terms of high accuracy without much testing loss for unlabeled smart city data. ? Moreover, by cooperating data centers over different urban areas, the prediction model can achieve a high learning performance with better accuracy rate, compared to centralized learning solutions at a single server.