Operating system crashes result in the interruption of vital services, productivity losses, and diminished system trustworthiness. A frequent cause of operating system crashes is a malfunction in the interaction between a hardware device and its driver. Such device/driver malfunctions are also a security threat which can be further exacerbated by backdoors planted in the drivers. This project will develop techniques to monitor device/driver interactions at run-time which can be used to enhance driver reliability and security.
This project explores hardware/software co-monitoring, i.e., simultaneous monitoring of hardware devices and software drivers. The research has four main goals. First, it develops techniques for run-time verification of properties at the device/driver interfaces. Second, it offers protection and adaptation mechanisms against both device and driver malfunctions. Third, it facilitates detection of (and protection against) transient hardware failures. These errors are particularly difficult to detect with traditional debugging methods, because of lack of replayability. Fourth, it provides a mechanism for early detection of security vulnerabilities, and system supports for on-the-fly protection and adaptation mechanisms. The outcomes of this project are tools for developing trustworthy computers in the face of a continual trend towards integrated systems with many disparate devices and drivers. The results will be disseminated by academic publications, publicly releasing the co-monitoring framework and its supporting tools, integration into undergraduate and graduate curricula, and technology transfer to the industry via active collaboration.
