I did not fully verify the whole paper, but I feel pretty skeptical about usefullness of this for the small numbers. All of their formula provide a linear speed

I did not fully verify the whole paper, but I feel pretty skeptical about usefullness of this for the small numbers. All of their formula provide a linear speed-up vs naive approach, which means that the big-O complexity is not going to be changing. So the best thing that this paper can do for determining if a number is a prime is to speed it up by some small constant factor (at most 210)... That's less than 8 bits of 2000+ bits that is currently used in RSA signatures, so basically noise.