基于喷泉码的隐私保护编码计算卸载方法  

Privacy-Preserving Coding Computation Offloading Method Based on Fountain Codes

作　　者：郭中孚 季新生[1,2] 游伟 赵宇[1] 巩小锐[1] GUO Zhongfu;JI Xinsheng;YOU Wei;ZHAO Yu;GONG Xiaorui(Information Engineering University,Zhengzhou 450001,China;Purple Mountain Laboratories,Nanjing 211111,China)

机构地区：[1]信息工程大学,河南郑州450001 [2]紫金山实验室,江苏南京211111

出　　处：《信息工程大学学报》2024年第5期559-566,共8页Journal of Information Engineering University

基　　金：国家重点研发计划(2022YFB2902204,2020YFB1806607)。

摘　　要：隐私保护编码计算研究未考虑节点之间的异构性及时变性,导致计算效率下降。针对此问题,提出一种基于喷泉码的隐私保护编码计算卸载方法。首先,为隐私分布式矩阵乘法运算设计多项式喷泉码编码器,该编码器支持持续输出任务直至成功解码,具有灵活的码率且能够适应不稳定的网络条件。其次,分析节点运行的时序特征,给出3种任务分配模式,从而丰富对评估分配策略的理解。最后,引入动态任务分发方法,为充分考虑节点的异构性将更多任务分配给性能更强的节点,并基于反馈自适应调节分配策略以适应系统的时变性。仿真结果验证了该方法的有效性,同时确保了信息论上的隐私保护。The prevailing research on privacy-preserving coded computation neglects the heterogeneity and time-varying nature of edge network,leading to compromised computational efficiency.To address this gap,a method for privacy-preserving coding computation offloading based on fountain codes is in⁃troduced.Firstly,a polynomial-fountain code encoder tailored for privacy-preserving distributed matrix multiplication is designed.This encoder supports ongoing task outputs until successful decoding,offer⁃ing an adaptive coding rate adept at accommodating fluctuating network conditions.Secondly,the tem⁃poral attributes of node computation is analyzed,and three distinct task allocation models are pro⁃posed,which enriches the understanding of allocation strategy performance.Finally,a dynamic task distribution method is subsequently presented,emphasizing node heterogeneity by allocating a larger share of tasks to high-performance nodes.This method adaptively fine-tunes the allocation strategy based on feedback,catering to the evolving system dynamics.Simulation outcomes validate the effi⁃cacy of this method,concurrently ensuring rigorous privacy protection as per information theory stan⁃dards.

关 键 词：边缘计算 编码计算 分布式计算 计算卸载 隐私保护 

分 类 号：TP399[自动化与计算机技术—计算机应用技术]