OSDev.org

To clarify: That paper does not detail how to use radiation of any kind to deterministically penetrate system security measures. All it does is showing how random memory errors of any kind could be exploited theoretically.

There's a world of difference between "hey, look here, I found a way how random memory errors could be exploited" and "this is the X-Bug 5000. Aim it at some computer and get a root login".

Well, I never claimed that we're at the stage where cosmic radiation can let us remotely control a machine, if that will ever be possible at all. What that paper showed was that given a malicious program running in a virtual machine (that under normal circumstances can't do anything bad) and radiation provided by the attacker, there is a success rate of around 70% for taking over said virtual machine (that's slightly less than random). I think it's pretty impressive.