Serverless computing’s cost-efficiency and reliability drive its adoption in heterogeneous network data distribution, yet challenges persist in data leakage risks and multi-level consistency maintenance. Current privacy-preserving approaches for multilevel network structure consistency suffer from insufficient coupling of heterogeneous nodes to edge weights, unreasonable privacy budget allocation for differential privacy preservation of edge weights, and insufficient consideration of node sensitivity in maintaining multi-level structure consistency. Inspired by microkernel architecture, our design prioritizes lightweight task management and flexible inter-process communication mechanisms to meet diverse system requirements. Therefore, we propose a multi-level consistency efficient operational privacy protection framework for serverless computing. First, a hierarchical multi-type node similarity measure is designed to ensure the consistency of node features and edge weights. Second, a privacy budget allocation method is developed based on the feature values of subgraph structures to avoid the waste of privacy budget. Third, a spectral clustering method is introduced based on differential privacy combined with a graph reconstruction strategy to protect the privacy of community subgraphs and ensure community structural consistency during privacy perturbation. Finally, inter-community privacy protection is achieved by edge node diffusion and isomorphic substitution in a serverless runtime environment. Experimental evaluations on three real datasets of different sizes show that our framework not only effectively balances data privacy and availability, but also achieves efficient multi-level consistency maintenance in a serverless environment, demonstrating promising potential to address the diverse demands of ubiquitous computing.

Aiming at the deficiency of traditional data imputation methods in effectively using the label information and random characteristics of missing data, a particle swarm optimization based imputation method for mixed features was proposed. The value of continuous feature was modeled as Gaussian distribution, and the mean and standard deviation were used as optimization parameters. The value probability of categorical features was optimized as a parameter. The classification accuracy rate was used as the optimization target to make full use of random information of label information and missing data. Four statistical methods and two evolutionary algorithm based imputation methods were used to compare the results on six typical classification datasets. The results show that the proposed method significantly outperforms other comparison algorithms in terms of classification accuracy indicator, and has better time overhead at the same time, which can effectively solve the data missing problems of mixed features.