GALOIS, INC. — Department of Defense SBIR Phase II: AF203-CSO1

Description

The objective of this study will be to develop technologies for recompiling legacy binaries to the secure CHERI ISA in support of DoD modernization priorities. Doing this would enable upgrading DoD systems processors and software by ruling out several important categories of security vulnerability that are possible on legacy architectures. However, recompilation is inherently a very difficult task, because CHERI binaries require metadata called capabilities, which are checked at run-time and alter the assembly code as well as the data layout. Recompiling a legacy binary to CHERI means reconstructing the logic of the original binary and inferring the capabilities necessary to ensure it runs safely.  The current state of technology is that binary recompilation to emerging platforms such as CHERI is impossible, even for the simplest binaries. The purpose of this study is to go from zero to one: that is, to establish that recompilation is possible, and to lay the groundwork for later studies that will show it can be done fully automatically at scale. We will do this leveraging technologies from binary lifting and recompilation, which exist in potential but have not been applied to the problem of recompiling to CHERI.