Search

Research on Cross-regional Disaster Data Sharing Model Based on Federated Learning

Author(s)

Xikun Li*, Zihan Chen, Tianqing Ma

Affiliation(s)

School of Civil and Transportation Engineering, Beijing University of Civil Engineering and Architecture, Beijing, China * Corresponding Author

Abstract

With the rapid development of information technology, disaster early warning and preparedness increasingly rely on the fusion and analysis of data from multiple sources. However, there is often a contradiction between privacy protection and sharing of disaster data among different regions and organizations, which limits the enhancement of cross-regional joint warning capability. In this paper, we propose a federated learning-based cross-regional disaster data sharing model that aims to address this critical issue. By constructing a federated learning framework for heterogeneous data, developing data desensitization and model parameter encryption techniques, and verifying the model's generalization performance in small- and medium-scale disasters, it provides new technical ideas and methods for cross-regional disaster warning. The experimental results show that the model can effectively improve the accuracy and timeliness of cross-regional disaster warning while protecting data privacy.

Keywords

Federated Learning; Cross-Regional Disaster Data Sharing; Privacy protection; Data Desensitization; Model Parameter Encryption; Generalization Performance

References

[1]Xi Enkang, Fan Jing, Jin Yadong, et al. Review of threats faced by federated learning in privacy and security field. Journal of Computer Applications,1-13[2025-05-27].https://kns-cnki-net.door.bucea.edu.cn/kcms/detail/ 51.1307.TP.20250526.1308.004.html. [2]Niu S, Kong W, Chen L, et al. Privacy-preserving and Verifiable Federated Learning with weighted average aggregation in edge computing. Network and Computer Applications, 2025, 240104201-104201. [3]Mou Yangcheng, Chen Aiwang, Chen Guirong, et al. Security defense strategies for federated learning based on data poisoning attacks.Systems Engineering and Electronics,1-14[2025-05-27].https://kns-cnki-net.door.bucea.edu.cn/kcms/detail/11.2422.TN.20250523.1048.051.html. [4]Su Linmao, Song Xuegui, Xiang Yi, et al. Application of Sensitive Data Identification and Desensitization Technology in Network Supervision under the Application of Sensitive Data Identification and Desensitization Technology in Network Supervision under the Integration of Big Data and Artificial Intelligence. China Science & Technology Resources Review, 2025, 57(02): 27-39. [5]Zhang Yu-Qing, Wang Xiao-Fei, Liu Xue-Feng, et al. Survey on Cloud Computing Security. Journal of Software, 2016, 27(06): 1328-1348. DOI:10.13328/ j.cnki.jos.005004. [6]Zhou Fei-Yan, Jin Lin-Peng, Dong Jun. Review of Convolutional Neural Network. Chinese Journal of Computers, 2017, 40(06): 1229-1251. [7]Jia Benyou, Li Dongzhou, Yang Fan, et al. Characteristics of heavy rainfall and floods in the Dawen River Basin from 2001 to 2022 and their relationship analysis. Hydro-Science and Engineering, 2025, (01): 76-86. [8]Zhao Lijie, Shi Yuzhi, Li Fulin, et al. Spatial-Temporal Variation Characteristics of Water System Connectivity in Dawenhe River Basin of the Lower Yellow River from 1990 to 2020. Journal of University of Jinan (Science and Technology), 1-8 [2025-05-29]. https://doi.org/10.13349/j.cnki.jdxbn.20250528.001. [9]Sun Xiaoming, Li Kexin. Research on Flood Forecasting in Dawen River Basin. Henan Science and Technology, 2024, 51(21): 43-48. DOI:10.19968/ j.cnki. hnkj.1003-5168.2024.21.009. [10]Xu Shan. Study on Variation of Hydrological Elements in Dawen River. Shaanxi Water Resources, 2024, (10): 44-46+50.DOI:10.16747/ j.cnki. cn61-1109/tv.2024.10.020.