Tom, >>> > > s9.2 >>> > > is it the length of the key that gives it strength or its entropy? >>> > > Is abcdabcdabcdabcdabcdabcdabcdabcdabcdabcdabcdabcdabcdabcdabcd0004 >>> > > really stronger than !qaurk/99SS~ ? >> > >> > Strictly speaking, you are right, but it is common sense that > cryptographic keys should have maximum entropy, i.e., they >> > should be selected uniformly at random from bit string of that length. > Consequently, virtually all cryptographic papers >> > and text books use key the key length synonymous to its entropy. Thus, > I consider this distinction as unnecessary. > > Yes but, consider the audience. Cryptographic material is aimed at > cryptographers who know about entropy without being told, the audience > of this I-D is all sorts, including MIB module specialists whose > knowledge of cryptography is probably less than useless. So I think > 'length' is wrong, 'entropy' is probably too technical (especially if, > like me, Thermodynamics was part of your degree), RFC5310 uses 'size and > quality of the key' which is probably a good compromise. >
This section is about the influences of the cryptographic parameters of HMAC on the security. The entropy or quality (I don't like that term) of the key is not a parameter, thus, it is not appropriate to mention is along the size. I suggest the following wording: The security of each HMAC-SHA-2 authentication protocol depends on the parameters used in the corresponding HMAC computation, which are the length of the key (if the key has maximum entropy), the size of the [...] Rational: If the key's entropy is not maximal, the length does not necessarily influence the the security. abcdabcdabcdabcdabcd is not necessarily more secure than abcdabcdabcdabcd. At least, the key length is not a primary parameter in this case. On the other hand, it is best cryptographic practice to choose keys uniformly at random, so that their entropy is maximal. -- Johannes _______________________________________________ OPSAWG mailing list [email protected] https://www.ietf.org/mailman/listinfo/opsawg
