英特尔软件防护扩展抗侧信道泄漏安全性研究

Data privacy and security are becoming one of the major concerns to process sensitive user data in a public cloud environment cost effectively. Secure remote computation is the problem of performing computation on a remote server owned and maintained by an untrusted party, providing integrity and privacy guarantees. Intel Software Guard Extensions (SGX) is the latest iteration in a long line of trusted computing designs which aim to solve the secure remote computation problem by leveraging trusted hardware in the remote server. However it has been shown recently that SGX is susceptible to a type of side channel attacks, which infer the sensitive data by observing operations on the resources shared with the SGX program..In face of the potential implications of side channel leakages, the project aims to understand how to protect users’ privacy data with SGX. Specifically the project plans to develop algorithm-specific side channel leakage resistant designs to perform privacy preserving genomic computations and oblivious SQL database queries. The project also plans to build efficient hybrid solutions to the secure remote computation problem by studying hardware enabled trusted execution environments (TEEs) based acceleration over fully homomorphic encryption schemes. Furthermore, considering even if the entire privacy guarantee is compromised, the project plans to investigate the problem of automatic partition and protection of binary code with the help of SGX’s integrity guarantee. The goal of the project is to enable the wide deployment of SGX in the cloud environment after key challenges to the side channel leakages are settled, and help to guide the designs of homemade processors in future with the understanding of both the advantages and limitations of the hardware enabled TEEs.

保护数据安全和用户隐私的前提下完成安全计算是云计算的一个重要问题，软件防护扩展(SGX)是英特尔处理器一项用于解决云环境下远程安全计算问题的硬件可信执行环境技术，具有广泛的应用前景。然而，研究表明SGX容易遭受基于微体系结构的侧信道攻击，如何从软件角度防止侧信道泄漏是研究人员需要解决的重要问题。.本项目旨在研究在面临侧信道威胁的情况下基于SGX保护已有应用代码，具体内容包括：以生物信息学算法和安全数据库算法为切入点，研究抗侧信道泄漏的算法设计；研究基于硬件可信执行环境的全同态加密技术，为远程安全计算提供软硬件协同的高效解决方案；基于SGX的完整性保证，研究二进制代码的自动分割保护技术。项目通过研究SGX应用中的侧信道泄漏问题，促进SGX在实际云环境下的部署和商业应用，有助于从根本上理解硬件可信执行环境的优势及局限性，对于我国安全处理器芯片的设计同样具有借鉴意义。

以英特尔软件防护扩展（SGX）为代表的可信执行环境（TEE）技术，通过处理器等硬件的支持，为应用程序创建基于硬件隔离的执行环境。本项目主要研究如何在面临侧信道威胁的情况下，基于英特尔SGX技术保护已有应用代码，具体内容包括英特尔SGX新型侧信道泄露方法研究、英特尔SGX抗侧信道泄露防御方法研究和英特尔SGX应用防护关键方法研究。在项目的资助下，取得的主要结果包括：提出基于二级方向预测器的侧信道攻击方法，通过对处理器分支预测组件的逆向工程分析，进一步拓展了SGX微体系结构侧信道泄露的范围；深化了对基于随机化缓存防御策略的理解，发现之前的基于随机化缓存策略的防御工作是不安全的，进一步提出基于末级缓存排出次数进行重新随机化，达到更高的安全性和效率；提出支持异构计算平台的TEE架构，具有可信计算基小、性能优、支持异构硬件、抗侧信道攻击等优势；针对基于TEE实现在线服务隐私保护时在多用户之间实现隔离的问题，提出一种轻量级、高效的解决方案，并进行了初步形式化验证；针对基于TEE的多用户协作场景下代码和数据隐私保护的问题，提出一种基于代理飞地的保护代码隐私的远程证明方法。