Hash function like MD5 or MD4 in CCS code for 18F2620

[evsource]

Hi,

I need a function that can take two relatively short IDs (about 10 digits each), and produce some output that is unique. If I understand correctly, this is what a MD5 hash will do.

Is there a computationally less intensive way to do this?

Does anyone have the MD5, MD4, or suitable hash function they could provide the code for?

[FvM]

A MD4/MD5 implementation can be found in RFC 1321 because it's used for the CHAP password algorithm. But's it's probably a bit extensive for a PIC18 project. Unintentionally, it has become my tool to reveal PCD arithmetic bugs. I hope, it compiles correctly with PCH, but I didn't yet try.

MD5 purpose is to make the signature unpredictable. If you don't need this strong cryptographic feature, simple CRC or XOR algorithms can do.

[FvM]

To add an update, the RFC 1321 test code is consuming 13% ROM and 25% RAM of PIC18F2620, but it works correctly, different from PCD with PIC24.

[ckielstra]

If you want to create a not (easily) predictable code word generated from the two IDs I would simply use a CRC32 routine. This is much lighter on the PIC resources than an MD4 or MD5 algorithm (which are not recommended for security anyway).

The CRC routines are defined by a set of six parameters, see chapter 15 in http://www.cs.waikato.ac.nz/~312/crc.txt.
Changing one or two of these parameters to not common used values will make it very difficult to guess the coding algorithm used.

[pattousai]

[FvM]

I simply compiled the code published in RF1321. It worked correctly with PIC18F2620 and PCH.

As reported, the MD5 code also has (unintentionally) become a testcase to check for PCD bugs. I wish, CCS would use it too, before releasing new versions.