The methodology can be more or less the same

• Use industry-standard scores to assess the security of a certain installation.
• Improve configuration via evolutionary (or other) algorithms, focusing on certain services or overall configuration.

But once that's said and done, we need to

• See if we focus on security, or also speed, for a multimodal optimization algorithm.
• Focus on server hardening or other services: there's been a lot of work on SDN, for instance.
• Even within servers, expand the possible configurations (use several different web servers, for instance), or more services (ssh servers, others).

Expand the state of the art with relevant bibliography.

Check test coverage and maybe expand it.

Incluso aunque sean scripts, se debe testear cada una de las funciones.

Check test coverage and maybe expand it.

Details missing:

• Cost
• Details on the evolutionary algorithm and what's peculiar to it.

This ad-hoc one, for instance, includes only 1-point crossover, which is not the most convenient, even more so in this case. Generally, 2-point crossover works better.

this is a test

No tengo muy claro el proceso de selección que se usa, pero me da la impresión de que la probabilidad de reproducción no depende del fitness. Generalmente, no es una forma buena de explorar el espacio de búsqueda, porque van a aparecer con la misma probabilidad en el pool reproductivo todos los individuos por encima de un fitness determinado.