基于隐私信息检索的大规模用电增信查询方法  

Privacy Information Retrieval Based Credit Inquiry for Large-scale Electricity Users

作　　者：李辉 黄祖源 田园 毛正雄 赵鹏[2] 任雪斌[2] 李亚男[3] LI Hui;HUANG Zuyuan;TIAN Yuan;MAO Zhengxiong;ZHAO Peng;REN Xuebin;LI Ya'nan(Network Information Center,Yunnan State Grid,Kunming 650011,China;School of Computer Science and Technology,Xi'an Jiaotong University,Xi'an 710049,China;School of Software,Henan Polytechnic University,Jiaozuo 454003,China)

机构地区：[1]云南电网有限责任公司信息中心,云南昆明650011 [2]西安交通大学计算机科学与技术学院,陕西西安710049 [3]河南理工大学软件学院,河南焦作454003

出　　处：《山西大学学报（自然科学版）》2024年第6期1211-1220,共10页Journal of Shanxi University(Natural Science Edition)

基　　金：云南电网科技项目(YNKJXM20210141)。

摘　　要：用电信用报告已经成为企业增信的重要凭证,但现有电力金融服务平台在提供用电信用报告查询时存在未保护查询方的查询偏好隐私信息,并且难以支持大规模数据库查询两方面问题。针对上述两个问题,提出一种基于隐私信息检索和不经意多项式计算的安全高效检索方法Effi-Retrieval。具体地,使用Paillier同态加密和不经意多项式计算实现查询方的偏好隐私和电力金融平台的数据库安全。此外,基于k-匿名方法在实现查询方的个性化隐私需求同时,结合哈希映射设计了最优分桶策略,用以降低查询方和电力金融服务平台间的通信开销。综上两方面策略,Effi-Retrieval将传统隐私信息检索的复杂度由数据库规模的指数函数降低为匿名参数k。最后,给出了Effi-Retrieval的安全性分析和用户隐私需求和承担费用对通信开销影响的数值实验。本文使用python的paillier同态加密库编写实验代码并在单台主机上模拟客户端与服务器间的交互,在FATE开源联邦平台提供的Credit Card数据集上进行了实验。结果表明,在数据规模300及以上的数据库上使用8192密钥位数加密时,与不使用分桶策略的传统方法相比,Effi-Retrieval的密态多项式生成时间可降低50%以上,与不使用k-匿名方法的查询算法相比,Effi-Retrieval的检索时间可降低30%以上。Electricity usage-based credit report is an important way for enhancing the query accuracy of enterprises'credit conditions.However,there are two main problems in existing credit inquiry service.The first problem is that existing methods do not protect the inquiry preference,which is the private information.The second is that existing private information retrieval method cannot be extended to larger-scale database due to the poor efficiency.To address the above problems,a novel inquiry method named EffiRetrieval was proposed to simultaneously ensure the inquiry security and promote its efficiency.Particularly,Paillier homomorphic encryption and oblivious polynomial calculation were used to ensure the security of inquiry.Optimal binning strategy with kanonymity method and hash mapping were used to achieve the personalized privacy requirements of the query party in practice.The complexity of privacy information retrieval was reduced from the exponential function of the database size to the anonymity parameter k.Finally,the security analysis of Effi-Retrieval was presented and numerical experiments were conducted to show the impacts of user privacy requirements and affordable fee on communication overhead.We used the Paillier homomorphic encryption library of Python to simulate the interaction between the client and the server on the same host,and conducted experiments using the Credit Card dataset provided by the FATE open-source federated learning platform.Given database consists of at least 300 items encrpyted by 8192 key bits,experimental results showed that compared with the traditional method without binning strategy,the encryption polynomial construction time of Effi-Retrieval can be reduced by more than 50%,and compared with the query algorithm without k-anonymity method,the retrieval time can be reduced by at least 30%.

关 键 词：电力增信查询 隐私信息检索 不经意多项式计算 K-匿名 通信开销 

分 类 号：TP309.7[自动化与计算机技术—计算机系统结构]