【作者】 王梅；

【导师】 陈晶；

【作者基本信息】 武汉大学 ， 网络空间安全， 2022， 博士

【摘要】 近年来,随着社会信息化的发展,人类工作、生活中的各种行为逐渐被数字化记录,从而引发了数据的爆发式增长。数据已经成为推动整个世界发展的主要生产要素之一,其中蕴含的价值日益彰显出来。数据资源的全球化成为已全球发展的重要趋势,其最核心的价值在于流通和共享,从而创造新价值、催生新场景。然而,目前数据共享的过程中普遍存在两个问题:第一,海量数据分布在独立运行的不同机构、系统、设备中,形成“数据孤岛”;第二,近年来关于隐私数据的泄露事件频繁发生,数据所有者之间互相不信任,数据共享的意愿越来越低。隐私计算技术的出现能够有效地解决上述两个问题,为促进数据的全球化进程提供了有力的技术支持。但是随着硬件技术、密码技术的飞速商业化,隐私计算领域也逐渐暴露出一些问题。首先,目前有许多隐私计算方案采用可信执行环境来保护数据运算过程中的隐私,但是资源受限和宿主机攻击等可信边界污染问题仍然是其实用化的障碍;其次,一些使用密码学隐私计算技术构造的方案需要参与方之间进行频繁交互和复杂计算,阻碍了其实用化的进程;最后,在保证隐私性和实用性的同时,还需考虑恶意攻击者存在的可能性,恶意攻击者的主动攻击将对系统中的数据隐私造成更为严重的威胁。针对可信执行环境的资源受限问题和宿主机攻击问题,本文在数据聚合场景中引入配备可信执行环境的聚合器以实现高效的多方协同计算,采用轻量级密码学操作对数据进行保护,从而降低系统的计算开销,并通过配备可信执行环境的密钥分发中心实现简单方便的密钥管理,提出凭证辅助的函数授权及验证机制,有效解决宿主机攻击问题。线上/线下的计算模式缓解了系统的线上计算压力,常数级的线上通信和计算开销表明该方案对资源受限环境同样适用。针对由密码学隐私计算技术构造方案的低效问题,本文在端到端的安全通信场景中提出了一种生物认证密钥协商框架,主要思想是将生物特征转化为生物特征私钥的方法实现生物认证,通过提出一种非对称模糊封装机制实现通信双方之间的单轮交互,利用用户自己的生物特征生成一个私钥和一个相应的公钥,用公钥对随机生成的秘密值进行封装用于协商会话密钥,只有与目标私钥类似的生物私钥才能解封装,从而协商会话密钥以保障后续通信数据的安全。本文用真实的生物特征(虹膜和指纹)对该框架进行实例化,安全性分析和实验结果有力证明了本文方案具有实用性和先进性。针对可能存在恶意攻击者的问题,本文在客户端/服务器模式的安全通信场景中提出了一个可更新的指纹认证密钥协商协议。考虑到恶意的攻击者有可能通过攻击服务器获取用户凭证,从而伪装身份,本文设计了一个基于指纹的盲凭证认证机制,将指纹模板隐藏在盲指纹中。当服务器的凭证数据库被攻击时,服务器可以独立地对凭证数据库进行更新,使得丢失的凭证失效,从而有效避免此类攻击。此外,本文采用的工具皆为计算较为轻量的密码学隐私计算技术,因此该协议仅需要单轮交互和线性的计算复杂度,具有较强的现实意义。本文主要从可信执行环境和密码学技术这两种不同的隐私计算技术所存在的难点问题出发,针对隐私保护数据聚合、端到端的安全通信和客户端/服务器模式的安全通信这三种具有现实意义的场景,设计低通信轮次、低计算开销、低存储开销的安全协议,并通过严谨的安全性分析和原型系统实验结果证明方案的安全性和实用性。本文研究内容虽不能覆盖隐私计算的所有方面,但仍将丰富安全协议设计理论,为隐私数据聚合和安全认证研究提供新的视野,为隐私计算长期可持续发展提供技术支持。更多还原

【Abstract】 In recent years,with the development of social information,various behaviors in human work and life are gradually recorded digitally,which leads to the explosive growth of data.The value contained in data is increasingly evident,making it one of the main factors of production that drives the development of the entire world.The core value of data is circulation and sharing,creating new value and enabling new scenarios.The globalization of data resources has become an important trend of global development.However,there are two common problems in the process of data sharing: first,a large amount of data is distributed in different institutions,systems,and devices that are not connected,forming ”data islands”;Second,in recent years,there have been frequent breaches of private data.Data owners distrust each other and are increasingly reluctant to share data.The emergence of private computation provides technical support for data globalization to solve the above two problems.With the rapid commercialization of hardware technology and cryptography,some problems are gradually exposed in this area.Firstly,many privacy computation schemes use a trusted execution environment to protect the privacy of data execution.However,resource limitations and host attacks are still obstacles to its practical application.Secondly,some schemes using cryptographic privacy computation technology require frequent interactions and complex calculations between participants,which hinders the process of their practical application.Finally,while ensuring privacy and practicality,we should also consider the possibility of malicious attackers,which will pose a more serious threat to data privacy in the system.Given the trust boundary pollution problems of the trusted execution environment,we employ the trusted execution environment to achieve practical multi-party collaborative computing,using lightweight cryptography operations to protect data,which reduces the computational cost of the system.Through the trusted execution environment of the key distribution center,we implement simple key management.In addition,we propose a credential-assisted function authorization and a verification mechanism to solve the host attack problem effectively.The online/offline mode relieves the online computing pressure of the system,and the constant online communication and computing costs are also suitable for resource-constrained environments.For the problem that the schemes based on cryptographic technologies are not practical,we present a biometric-authenticated key exchange framework in the end-to-end secure communication scenario.The main idea is to convert a biometric characteristic into a biometric secret key to implement biometric authentication.We propose a new asymmetric fuzzy encapsulation mechanism to implement a single-round communication protocol between two parties.Each party generates a biometric public key based on a corresponding biometric secret key,and encapsulates a secret value with the biometric public key to negotiate a secure session key.Only a party with a secret key that is very similar to the target secret key can decapsulate the message.Finally,we use the shared secret values to generate a secure session key.We use realistic biometrics(iris and fingerprint)databases to instantiate our framework.The security analysis and experimental results prove that the proposed scheme is practical and advanced.To solve the problem of malicious attackers,we proposed an updatable fingerprintbased authentication key agreement protocol in client/server secure communication scenarios.We consider that malicious attackers may obtain user credentials by corrupting the server to disguise their identity.In this paper,we design a fingerprint-based blind credential authentication scheme,which hides a fingerprint template into a blind fingerprint credential.When the server’s credential database is compromised,the server can update the credential database independently to invalidate the lost credentials,thus effectively avoiding subsequent attacks.In addition,we use lightweight private computation technologies in designing this protocol,which only needs single-round interaction and linear computing complexity.We focus on the difficult problems of two different private computation techniques in this paper,i.e.,the trusted execution environment and cryptographic techniques.This paper designs low-interaction,low-computation,and low-storage secure protocols in three promising practical scenarios,i.e.,the privacy-preserving data aggregation scenario,the end-to-end secure communication scenario,and the Client/Server secure communication scenario.We show the security and practicability of the designed protocols by giving rigorous secure analysis and conducting prototype experiments.Although this paper cannot cover all aspects of private computation,it will enrich the security protocol design theory.In addition,it provides a new perspective for privacy-preserving data aggregation and secure authentication research,and provides technical support for the long-term sustainable development of privacy computation.更多还原