Yes, I agree with you on the part that it is not the most efficient network protocol, but you're overlooking the fact that blockchain is almost impossible to take down and is censorship-free. Even if they manage to locate the C&C; they simply couldn't shut it down, since BC is decentralised and a distributed database.

There's also the disadvantage of having every communication between the malware and the controller publicly logged, and the fact that that every command issued would have a small transaction free.

It's not expensive because you only need to post a message when you are changing IPs, and it will only cost a fraction of a cent. If botnets used to work with hardcoded IP addresses, it's clear that changing IP addresses does not happen very often, and usually triggered by a government action which is very expensive for the government.

It does not need to be CPU or network intensive because it can basically use a light wallet implementation.

On what basis are you saying this paper "isn't published"? Just curious.

"Published" papers are papers published in a scientific journal after peer review. This just appears to be an un-reviewed article.

This would be the fallback com-channel, not the primary. If an ip or board gets taken down, no problem.

Yes. But many botnets nowadays do bitcoin mining, the overhead in those cases might be small.