The simplest error correction code is a repetition code. This has
   escaped many peoples attention.
   [1]https://eprint.iacr.org/2017/382
   " The frequency at which a key should be changed in order to maintain
   an minimum level of protection depending on the number of unrolled
   rounds computed per cycle is explored."
   Here some attacks were made against Simon and Speck.
   In my lay opinion, RC4 is more secure for the internet of things. The
   greatest vulnerability for computers is memory bound errors, not... uh.
   Malicious javascript sending a DDOS on an 100 megabit uplink to a
   server so a passive adversary can collect ciphertexts and do
   statistical analysis. Naturally everyone says that no one was fired for
   using AES, but who was fired for not putting a password on a database?
   I think if the RC4 round function was applied key length bytes more
   times (128-bit key, 16 more key schedule rounds), the first few bytes
   will have less bias, and the only related key recovery attacks apply to
   the first few bytes.
   An additional xor to mask the output or input of a byte lookup may
   improve things.
   Or use an ARX cipher as a NLFSR, like the Lex cipher.
   In any case, don't use error correction codes in cryptography.
   This is some kinda multidimensional chess.

References

   1. https://eprint.iacr.org/2017/382