My master's thesis project, advised by Dr. Yeongjin Jang at Oregon State University. For further information, see my paper and slides.
Abstract:
As the software engineering industry adopts memory-safe programming languages for security- and performance-sensitive systems programming, developers face a need for interoperability between memory-safe and memory-unsafe codebases, without compromising on either security or performance. We present a technique for sandboxing unsafe dependencies within a safe program, using hardware features such as Intel's Memory Protection Keys to prevent memory corruption bugs within unsafe code from affecting the safe portions of the program. We use the type system and metaprogramming capabilities of the Rust programming language to create a low-overhead, ergonomic, and type- and memory-safe interface allowing the safe portions of the program to freely interact with the unsafe portions, without risk of inducing undefined behavior in memory-safe code.