nesdev.org

My browser have eaten my previous mega-post on this, so here we go again... :P ----- I have done a bunch of tests these days, with very interesting results: To start with, i calculated the "mutual information" coefficients associated to the changes of every pair of output bits when changin...

Cita: I don't have much knowledge of block ciphers to back this up, but my intuition leads me to ask: Is this an unbalanced Feistel block cipher which breaks the input into 4 bytes? There is no way i can answer that question without knowing the algorithm, but i don't consider that tests as a solid ...

If there is a map taking any N bit input to another N bit output, and the mapping is one-to-one and therefore invertable, then if we construct the table (as above) for this, then the property that all the values in a collumn must be either all odd or all even can be said for the rows as well. This ...

EDIT2: Doh! I see now, that property of the table must be true for any algorithm. So it constrains nothing.

True. I'm sorry for not having explained it before... it could have saved you the work.

Yeah, I already started with the old list. I might as well let that finish, and rerun with the new list next if you want. Not still; firstly send me the file you are doing; i will decide based on what i see in that. I'm not sure what this table is representing. Can you describe it a bit more? Of co...

Ok, I have repeated the test with the 4dc ID. the result would be this: input bits: 0 4 5 6 7 9 a c d e f 16 17 1a 1b 1d output bits: 0 2 4 6 8 9 f 12 13 14 15 16 18 1a 1e 1f and, when mixing both, this: input bits: 0 2 4 5 6 7 e f 14 15 16 17 1a 1b 1d 1e output bits: 0 1 2 4 6 8 9 c f 12 13 14 18 1...

For every challenge I've seen so far, the following ID's give the same output: ID004DC ID004DD ID0001C ID0001D The only other devices I own, fit in another group. These ID's give the same output: ID0007C ID0007D ID001EC ID001ED ID0006C ID0006D Well, maybe this could give some info about the key-sch...

OK; i have got some time to do tests on the last set of data. Trying to see if this could be a Feistel network, i did what i said you: i have tried to distinguish two 16-bits set of bits (both in the input and in the ouput) that show clear differences in "randomness". I have seem big diffe...

What data would be the most useful at this point? Repeating these for another "ID"? Probably; when trying to make statistics for characteristics than only depend on the "structure" (the topology, etc) of a network, mixing data from different keys can help to cancel out key-relat...

Would it be helpful if I held the lower word fairly constant and "counted" the upper word instead? Yes, that's another one i would like to ask. While a evenly-spaced set is better for the tests that try to group bits, that could be more useable for doing tests changing isolated bits in th...

Another commentary: Due to the fact that the big blocks of data we have are not homogeneusly spatiated within F2^32, i'm fearing some of the statistics could be distorted (it's almost impossible to say if this would be the case, but a priori it's possible). In that line, i would like to get a set of...

That's amazing, F doesn't even need to be invertible. What a neat structure. Yeah. Indeed, if F were invertible, the schema would become weaker. So then is the main difference between a Feistel network and a Substitution-permutation network, that the Feistel network uses non-invertible functions in...

Yeah. After taking a first look at some statistics, it really seems a block cipher. I'm guessing it could be a multi-round scheme, most probably a SP network or a Feistel network with a relatively low number of rounds. If we got a full table, i could confirm o discard the second case; in our case, i...

Thanks for the heads-up, Lord Nightmare. Hi, neviksti; i'm glad to hear from you again. :) I'm interested on this; indeed, you are doing what i was desiring to do with *another* encryption device, so i feel motivated to this. ;) I have just read the thread, and i still have to take a look at the fil...