The security guarantees of formally verified compilers for compartmentalized programs in unsafe languages like C typically stop at a high level of abstraction. According to the arXiv abstract, they assume a built-in protection mechanism and a block-based memory model, leaving a gap between these assumptions and the low-level realities of compartment isolation.
The paper proposes 'blame-aware recomposition' as a way to construct secure compilation proofs that operate at a lower level. The title suggests a focus on assigning blame to components when isolation fails, but the abstract does not provide further detail on the method or its evaluation.
The significance lies in addressing the mismatch between high-level proof assumptions and actual low-level mechanisms. By making proofs more aware of the real protection and memory layout, the approach could strengthen guarantees for compartmentalized C programs, though the abstract leaves the specifics open.