Taking control the Byteball network with 10 USD a day ?

I have read the byteball white paper and I find this new concept really promising, in particular insofar as byteball does not need millions of asics processors running 24/24 to ensure network stability.

However I am wondering about its global stability strength.

I see on byteball.fr stats that witnesses post around 5000 units per day to ensure the "reality strenght" of the network.

The minimal cost (header + payload commission) of posting a unit is 501 bytes.

So, at the current byteball GB price on exchange markets, posting 5000 units costs 0,5 USD.

It seems that an attacker, bringing with him his own witnesses, could easily (with a budget of, say, less than 10 USD a day) completely flood the network and eject the current honest witnesses.

Will the "near conformity rule" prevent from such a take over ?

As far as I understand the rule, attacker can modify the witness list incrementally (one by one) so it will just take 12 units on the main chain to make the complete switch toward his bad witnesses list.

He could then impose his own "vision of reality", allowing double spend and so on...

I add that the required budget would actually be *even lower* as the attacker would get back most of the header and payload fees.

Did I misunderstand something or is it just a weakness of youth ?