DivA: Automated Generation of Logical Code Diversity

If a computer system suffers a cyber attack, any compromised component that is simply reinstalled will remain vulnerable to the attack that damaged it; so replacements should provide equivalent services implemented differently. This defense can be applied proactively, to create a"moving target"by periodic reconfiguration. Implementing these strategies requires an effective way to create diversity. Randomization has proven effective against many known attacks. ATC NY and Cornell University will develop DivA to provide a fundamentally different, and complementary, logical diversityto generate modules that provide equivalent services by different algorithms. DivA exploits the constructive logic principle of"proofs as programs."A developer creates an initial implementation of a component with a program extracted from a constructive proof; DivA uses heuristic methods to generate alternative proofs of the same proposition, and therefore alternative implementations of the component (with strong guarantees of equivalence). The Nuprl logical programming environment provides powerful support for both of these tasks. DivA will also provide a library interface through which a system integrator or automated recovery mechanism can discover modules and thereby create a vast number of logically distinct versions of the same system.