Talk:ISO/IEC 9797-1

Algorithm 5 and CMAC

I don't have a copy of 9797-1:2011, so I can't check, but it seems there is still a discrepancy (after this edit) between the descriptions in ISO/IEC 9797-1#MAC algorithm 5 and CMAC.

ISO/IEC 9797-1#MAC algorithm 5 says:

MAC algorithm 5 comprises two parallel instances of MAC algorithm 1. The first instance is operates on the original input data.

but the diagram on CMAC shows that the last encryption with k (the original key) is of m'_n, not m_n as would be implied by "The first instance [of MAC algorithm 1) operates on [all of] the original input data." (Note that n in CMAC is equivalent to q in ISO/IEC 9797-1, both being the number of blocks in the data/message.)

ISO/IEC 9797-1#MAC algorithm 5 also says:

The second instance [of MAC algorithm 1] operates on two key variants generated from the original key

but the diagram on CMAC shows that the key variants k₁ and k₂ are used to "tweak" m_n; there's nothing in the CMAC diagram the remotely looks like MAC algorithm 1 operating on (or with) the two key variants.

Finally ISO/IEC 9797-1#MAC algorithm 5 says:

The final MAC is computed by the bitwise exclusive-or of the MACs generated by each instance of algorithm 1

but the diagram on CMAC shows that the MAC is the result of E_k(...), not the result of an XOR.

Is the algorithm described by the CMAC article actually the same as ISO/IEC 9797-1 MAC algorithm 5? (The CMAC article does not mention 9797-1.) If so, ISO/IEC 9797-1#MAC algorithm 5 needs re-writing so that it is accurate. If not, we should not link to it. Possibly there are multiple algorithms referred to as "CMAC", in which case a disambiguation page might be required. Mitch Ames (talk) 12:53, 26 May 2016 (UTC)