md5 hashing

[ninjarider]

anybody know how it works

[Ghirai]

http://www.google.com/search?q=md5+hash

First result.

[ninjarider]

im currently reading
http://www.ietf.org/rfc/rfc1321.txt

was wondering if anybody has actually worked with it. or had any projects to work with.

[Shantanu Gadgil]

im currently reading

The source code at the end of the RFC works OK.

Used the code to hash large file too ... work fine!

Regards,
Shantanu

[Tedd]

It's quite straightforward to implement, just as long as you follow the instructions correctly (and check your results with the examples given.)
The RFC gives all of the information you need, and source code; and there are a few examples on the forum :wink

How it works (without detail): take the input data, segment it into equal size blocks (apply padding to the last block), apply a transformation to each block, combine the blocks progressively, produce the resulting hash value. Easy :lol

[drhowarddrfine]

The web site I am creating, I use it to generate a customer id based on his ip, date/time, and a wildcard value. 

[evlncrn8]

The web site I am creating, I use it to generate a customer id based on his ip, date/time, and a wildcard value. 

i hope you're aware of md5 collision problems... :)

[Tedd]

i hope you're aware of md5 collision problems... :)

Any kind of hash is going to have collisions (unless you know all of the values to be hashed in advance - then you can cheat) - you're trying to map a large number range onto a smaller one, obviously some values will have to 'collide.'

[drhowarddrfine]

The collision problem is avoided because the wildcard number is known only to us, at least as far as any security flaws are concerned, plus the id is only valid for 30 minutes and ids are checked for duplicates before being issued.

[AkinforASM]

i hope you're aware of md5 collision problems... :)

While using pascal, I even sometimes encounter collisions with SHA256  :dazzled:.

[ninjarider]

ok. i understand most of it except 1 thing

at this site http://en.wikipedia.org/wiki/MD5

they use the variable w[]

they have it listed on the site as a word. did they mistype that and it needs to be double word.

[Tedd]

'word' simply means "the word size of the machine" - so for a 32-bit machine the word-size is 32-bits ('dword') and for a 16-bit machine it is 16-bits ('word').

It causes some confusion as we tend to use 'word' (not 'word-size') to mean a 16-bit word -- the reason for this is historical, but unfortunately it's stuck.
Anyway, no they haven't made a mistake, it says "sixteen 32-bit little-endian words" which means: 16 DWORDs, little-endian byte-order.

[ninjarider]

 k := floor(abs(sin(i + 1)) × (2 pow 32))

what in the world is floor() suppose to do.

[Ian_B]

I hate to see people totally wasting time re-inventing the wheel.

I posted fast MD5 code on the old forum, and I was prompted to update it slightly recently. You'll find the original MD5 Rivest specification in this zip to see how the code follows the spec exactly. Note it is not neatly tidied into Macros like some simple versions you may see around, as that doesn't allow full optimisation.

http://www.i-asm.com/MD5v2a.zip

EDIT: minor update in zipped code

IanB

[ninjarider]

dont know why i want to know. but how do you go about using that function on the windows calculator. and for some reason i cant remember what abs does.