Assume ibm cares as much about your privacy as it cared about the welfare of south Africans in 1960. Also assume its building nsa utah spying center. Also assume, as with todays secret orders to such firms, back in 1980 it designed the combination of des and des key management “with a view to certain secret orders” (of the day).
What would they have been?
Folks are learning that cryptanakysis is half search (for text with correlations) and half search for correlations of particular types.
For ibm, one wants to satisfy both customers : the banks wanting crypto for their transactions and nsa wanting to spy on the transactions. And ibm wants to sell what ut has special skills in: search hardware.
Back in 1980, cryptanalysis needed to let nsa search out wrapped keys from encrypted keys fro encrypted data.
Parity bits are critical, in 56bit des (with its 8×8 block length).
Parity bits are critical when a des deciphered block is used itself as a key. A wrapped key upon unwrapping shows the correct parity (to the enforcing hardware). An encrypted key does not, so hardware must take a different path key loading. And, the point being, it does so when instructed that this is expected – given the phase of the key distribution Protocol that needs key encryption (vs key wrapping recall).
When searching for correlations between ciphertexts, in 1980 one was still essentially looking for “depths” – in order to then exploit this set as the more likely candidates to feature the required detectable correlations.
One has to remember that the feistel net only assures that, given one bit of input change, half the output bits change ON AVERAGE. Which means one is interested in the cases that only change a small number (for related messages). Here is your “depth collection” search. Find them!