Bolted: Securing Bare Metal Clouds

Speaker/Bio

Jason Hennessey is a postdoc with the Mass Open Cloud at BU (an effort to foster open, multi-vendor, secure clouds), where he leads the Hardware Isolation Layer and Secure Cloud/Bolted efforts. His interests and works include improving security & privacy (especially side-channels and defensive techniques), cloud and operating systems, analyzing the use of online scholarly resources and bioinformatics.

Prior to completing his PhD in Computational Sciences and Statistics at South Dakota State University, Jason spent 12 years in the computer industry, doing Operating Systems at Digital/Compaq/Hewlett Packard and Cloud Computing at VMware.

Abstract

This work introduces Bolted, an architecture for increasing a tenant's trust of the firmware and software installed on a leased bare metal server. In general, when executing in a cloud environment, the tenant must trust all the software on top of which it executes. When tenants make use of bare metal nodes, there is less software to trust, but more opportunities for previous tenants to compromise the privileged software and firmware, introducing persistent malware to a machine.

To increase a tenant's trust in the bare metal node, we propose an architecture and implementation that uses an exokernel-like approach to enable a tenant to select the components that meet their security and efficiency requirements by mixing components. Bolted makes use of: Heads (a hardened, minimal Linux packaged as a Coreboot payload) as the main firmware infrastructure; Keylime as a distributed, tenant-operated attestation infrastructure that leverages a node's TPM; HIL, a network isolation and maximal tenant control; and BMI, for very fast imagine provisioning under tenant's control. In this way, tenants can have a level of confidence in the components based on two mechanisms: they can be inspected for correct functionality; and attestations can be provided based on their measurements to show that no modifications had been made between boots.

A collaboration between BU, Northeastern, Two Sigma and MIT Lincoln Labs.