Key is bad (RFC2898 Password Based Key Derivation) about aes HOT 10 CLOSED

Unfortunately the files you have provided contain a number of errors. When these errors are corrected they then produce a good result (see the attached screenshot). I have attached a zip file that contains all the files that I used to produce a good result.

fcrypt.zip

from aes.

Strange ! I have compiled your fcrypt.zip on my linux machine and the results are different.

Same is the result when "pwd2key.c" is consider as the main source.

from aes.

Unfortunately I don't use Linux or GCC so I cannot help you directly. To see where the code is going wrong we need to check the values held in the two arrays uu and ux within the loop starting at line 83 in pwd2key.c. I have attached a zip archive that includes a modified version of pwd2key.c (which outputs these arrays in this loop) and output.txt that gives the result when I run it. If you use this version of pwd2key.c, it should allow us to see where the results are different for you. If you let me have your results I may be able to see what is going wrong for you.

looptest.zip

from aes.

Result is here

results.txt

The left one is my machine generated results and the right one is yours. The values changes at uu_0[0]. I have highlighted the region from where the values are changed.

from aes.

In fact the changes you have noticed do not matter because only the first four values are used. But you are right that the error occurs in the first execution in the loop and seems to occur in SHA1. I have a more recent implementation that you might like to try - I have attached the files in a zip archive.
fcrypt_new.zip

Please let me know if this also gives the errors.

from aes.

Just tested your latest files. The keys are good now.

Which file do you actually changed to get the correct result ? The problem was in SHA1 ?

from aes.

The issue was somewhere in the SHA1 code. The code you were using is very old so I just replaced it with my current SHA1 and SHA2 code.

from aes.

fileenc.c and fileenc.h are the old implementation in the fcrypt_new.zip which you had attached recently. You have changed hmac_sha1_begin(cx->auth_ctx) to hmac_sha_begin(2 arguments). It will be helpful if you could provide me the latest fileenc.c and fileenc.h

from aes.

My fileenc application is very old and I no longer maintain it so there is no later version. But it depends on my AES and SHA code, both of which are maintained (and can be found in my repository here). SInce the current SHA code can use both SHA1 and SHA2 I had to make changes in the fileenc code to accommodate this change (note also the addition of sha2.h and sha2.c).

If you wish to continue with the old code, you can now track down and find the issue that causes it to fail by running both the old and the new versions in debuggers to find the point at which the SHA1 code in the old version does not match the data in the new version. I may then be able to change the old SHA1 code to avoid the error that occurs with Linux/GCC.

from aes.

Thank you BrianGladman. I will let you know once i found the cause of the issue in previous SHA1 implementation.

from aes.