ABSTRACT It is impossible to fix security for everyone in all cases. Nothing can be made foolproof, because fools are too ingenious.
However, currently, even users who "do everything right" are remarkably unsafe, because important software fails to be secure even when properly configured.
I present the case that, surprisingly, enough is finally known to dramatically improve this problem.
I will argue that, if appropriate and (entirely affordable) methods are adopted by the authors of important systems, slow incremental improvements could at last fix the infrastructure problem, and possibly even before computer security issues destroy our civilization.
I will discuss the necessary techniques, and present a plan for their adoption.