CRII: SaTC: Evolving I/O Protocols for Confidential Computing

Cloud computing has quickly become a standard approach for deploying online services due to its scalability and cost savings, despite the new threats it poses for data security and privacy. As the cloud service provider (CSP) operates the hardware and privileged software, a malicious CSP can readily access or alter a customer’s sensitive data, while a bug in the CSP’s software can allow an attacker to do the same. To improve cloud security, the major chip designers recently added support for confidential computing: running a workload in a secure hardware enclave where the workload’s memory is encrypted, thus shielding it from the CSP. This proposal advances the knowledge for extending confidential computing to distributed services, an increasingly common type of cloud computing. The proposed research will develop innovative, open-source systems software for organizations to make use of essential cloud services without sacrificing their privacy or the privacy of their users. To prepare the next generation of cybersecurity practitioners, planned activities include the mentoring undergraduates and outreach programs for K-12 students. The proposal’s central research questions are: (1) how must protocols handle authentication and authorization when some nodes in the system are enclaves, and (2) how can protocols safely encompass a mix of trusted and untrusted nodes, including nodes that an attacker compromised. The proposal investigates these questions for two important existing protocols: Transport Layer Security (TLS) and cloud object storage. Specifically, the proposal seeks to extend TLS to allow an enclave to authenticate itself using a set of identities (such as its launch measurement, runtime measurement, and domain name), and explores the use of joint signature schemes to make such authentication efficient and elegant. Having support for complex endpoint authentication, the proposal then develops extensions to TLS for authorizing how enclaves may share data, thus enforcing the flow of information through a distributed system of diverse parties. Finally, the proposal explores storage protocol extensions that mitigate the damage of key compromise in a shared, encrypted cloud object store by using efficient, enclave-assisted, key rotation and data re-encryption. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.