2 ms·
Good points. Far as code proofs, I think you should look at Microsoft's VerveOS (safety not security), seL4 (security), Astree, and the SPARK-based security pro
by nickpsecurity 11y ago
Good points. Far as code proofs, I think you should look at Microsoft's VerveOS (safety not security), seL4 (security), Astree, and the SPARK-based security projects. They all prove aspects of the code itself. The old A1 projects also typically mapped a low-level spec to code with many things at or near a 1-to-1 mapping. Very low risk of a misrepresentation or abstraction gap attack when done that way.
I still always encourage them to combine their efforts with thorough testing, code inspections, static analysis, and careful source-to-object code validation. Most of that is even required in the high assurance standards. Takes a lot more than proofs, code reviews, or tests in isolation to begin to make a strong security argument. As you know, it's so labor intensive and requires such specialist skills that almost all commercial and FOSS projects skip steps to get running code out the door. That's why penetrate and patch will always be the norm with highly secure systems always several versions worth of features behind mainstream. If they have them at all...