MD5 is not what Apple will be using, since every different bitrate and/or encoder will yield different files and thus different md5 sums for the same song. AND you will never know from the perspective of apple, which song we are looking at, if there are no meta information. So Apple will use some kind of audio fingerprinting system like MusicDNS or whatever. Look it up in Wikipedia. Probably they can even reuse information gathered by iTunes genius.