AES modification

Question

I have idea. Generate random 256 numbers or permute in random way numbers from 0 to 255. Add second password. Generate subkeys from second password same way they generated from first. Every round use generated subkey as indexes to array with randomly generated numbers and xor them with partialy ciphered text. Finly those 256 random number would need to be attached to ciphertext to deceupt.I would like to ask is this something already known/tried and does it make any sense.

Answer 1

(Psuedo-)Randomness don't have to be that complicated. All that need to make it computationally indistinguishable from is the combinition of linear layers over (not necessarily big or wide) non-linear s-boxes, (with round constants to remove symmetries), with many rounds to ensure sufficient confusion and diffusion.

That's the idea behind new designs such as Keccak, Xoodoo, Gimli, Ascon, etc. - Keccak and Ascon are chosen for standardization by NIST btw.

Linking a Q of ours to enrich the knowledge web.