We describe the challenges of analyzing information flow in today's computing systems. The enforcement of information flow policies in modern computer systems is one of the most important aspects of modern computer security, yet it is one of the hardest to get correct in implementation. The security of a system depends on managing the multitudes of interacting devices propagating various types of information. Designers are asked to think about new properties of the system while ensuring that unchecked data does not subvert the integrity of the system at large. The scale and complexity of interaction between different components in hardware and software makes it difficult to reason about how system properties cut across the different layers. As data propagates from one layer to another it changes along the way, generating new data and influencing system behavior.

We propose that the visualization of information flow at the hardware level can be used to gain insight into the interaction of components and the non-trivial, non-intuitive dependencies between data and design. To realize this vision we propose a three-step approach: First, we track and collect information on events of interest at the hardware layer and across layers of the software stack. At the hardware layer we present the technique of Backtracking on data to identify the conditions that caused one or more violations. Understanding how information is processed at the gate level gives a microscopic picture of how it is manipulated by the system. To move beyond the physical bounds of hardware and gain a macroscopic view of system behavior, we present Data Flow Tomography. With this tool we capture the footprint of information from the network and registers to software libraries and user space applications. Second, we classify low level events into comprehensible categories.

Low level signal changes are mapped to higher order abstractions of the various components. Third, we provide visual representations of these abstractions that enable designers to investigate the behavior of a system at multiple levels. The research in this dissertation focuses on how visualization of the properties of data can aid in understanding the structure and soundness of embedded systems during the design process.