|
The interaction between computer processors -- the hardware at the heart of our computers, tablets, and phones -- and software -- apps, web browsers, and other applications -- is governed by an Instruction Set Architecture (ISA). The ISA is the specification that defines how the processor will respond to commands from the software. It is large and complex, too large for a person to understand and reason about all the interactions between different parts completely. As a result, security vulnerabilities exist in the ISA. These vulnerabilities can sometimes be exploited by attackers to steal data or take control of the machine. This research is about detecting security vulnerabilities that exist in the ISA. Finding and removing these vulnerabilities will create a more secure foundation for all our computing activities. This will benefit government agencies that require high assurance environments, cloud providers that rely on hardware features for the security for their service, and users who, more and more, are relying on diverse hardware components from a variety of hardware design companies to handle their private and sensitive data.
The researchers posit that vulnerabilities in the ISA happen in one of two ways: 1) Erroneous specification: the ISA prescribes behavior that is dangerous; or 2) Nondeterminism in the specification: the ISA is incomplete and one of the possible behaviors allowed by the specification is dangerous. The hypothesis of this research is that it is possible to focus on a relatively small subset of the ISA for which these types of errors are likely to occur. The researchers are developing a practical methodology for discovering for which instructions vulnerabilities are most likely to occur. With that information, they are developing tools to detect and correct security-critical errors in the ISA. In addition to making a practical contribution, the research activities are improving understanding in the computer science community of what a vulnerability in an ISA looks like and where and under what conditions it is likely to occur. This will enable future verification efforts to concentrate on the most security-critical aspects of the ISA.
|
CRII: SaTC: Detecting Security Vulnerabilities in Instruction Set Architectures