I'm also researching the subject and reading in [crypto.SO] I found this interesting answer, so I'll fit it here, I hope it caters to us.
On page 3 of the LED Algorithm Proposal says:
"Note that for 64-bit key K
, all subkeys are equal to K
, while for 128-bit key K
, the subkeys are alternately equal to the left part K^1
and to the right part K^2
of K
. "
Translating:
Note that for a 64-bit k
key, all subkeys are equal to $ k $, while for a 128-bit k
key, the subkeys are alternately equal to the left of k^1
and the right part k^2
of k
.
Basically, the main key entry is divided into nibbles sorted lists, and when the algorithm needs material for subkeys it uses exactly the nibbles directly from the sorted list - moving each nibble to the end of the line, so that all nibbles are used in succession. Since the algorithm handles 64 bits of subkeys at one time, for a 64-bit master key each subkey is simply the primary key, and for 128-bit keys the algorithm will use the first 16 nibbles of the master key, and then the second part, the remaining 16 , and then the first 16 again, and so on. At the top of page 4 shows the diagram of how this works for an 80-bit master key.
Considering the original question in which this response was applied, the response author points out that unused subkeys in ' round ' as highlighted in the question, but that between each step , while each step is composed of 4 rounds. Each round consists of 4 operations, very similar to the one used in the AES algorithm - first you run a xor in one round, then you replace each nibble using a non- linear (the cipher s-box), you transpose the nibbles and then pass through the highest linear fuzzy permutation with a high branching factor (similar to MixColumns in AES , but optimized for nibbles ).
Based on response: link