Autonomously Secure Disks

Disks increasingly support multi-boot systems, are accessed remotely, attached to mobile devices, or used in other previously unanticipated operating environments. Because these uses further separate the storage from the data owners, they often lead to new security vulnerabilities. Such concerns call for a change to traditional disk security models. This talk introduce and explores autonomously secure disks (ASDs). ASDs enforce security at their interface–thus treating all systems as untrusted entities. Such architectures provide a means to protect user data from rogue or penetrated systems, and significantly reduces the trusted computing base (TCB) for enforcing data security. We introduce
three example ASD security policy systems and develop an architecture based on recently introduced hybrid hard drives. Further, we augment the DiskSim simulator with NVRAM services to evaluate the costs associated with managing security contexts within confidentiality and integrity protected secure storage. These experiments show that proper tuning of system parameters can eliminate much of costs associated with managing security meta-data. Finally, we consider the challenges involved with the integration of ASDs into current operating systems.