You have lots of answers here and they are accurate but they don't really explain why.
MD5 is a hashing algorithm. What a Hashing algorithm does, is take a long piece of data and analyse it cryptographically in a way that creates a smaller piece of data. So from ABCDEFGHIJKLMNOPQRSTUVWXYZ
with my custom hash algorithm I might create a single digit hash 5
.
When that is done, you lose information - ABCDEFGHIJKLMNOPQRSTUVWXYZ
contains far more information than 5
and there is no way to make the translation the other way.
The problem with hashing in a way that only allows an outcome of 0-9 ( this is effectively a Checksum ) is that if you take two pieces of text, the chances are quite high that they will have the same hash. So maybe with my algorithm ZZZZZZZZZ
will also produce a hash of 5
. This is what is termed a Hash Collision.
Now what happens if I take the hash of my hash? Well, my starting point is already very low information - the most it can possibly be is one of ten digits, so the chance of a collision is now exceedingly high. Supposing when my hash algorithm runs on numbers it returns 1
if it is odd and 0
if it is even- so if I have a hash of ABCDEFGHIJKLMNOPQRSTUVWXYZ
which comes to 5
then I have a 10% chance of a collision. But if I make a hash of that hash, I will now have a 50% chance of a collision.
The trick of cryptography is hiding information in such an enormous possible space that it is unbelievably hard to find. The more you shrink that possible space, the less well hidden your information is.