topic Re: Rijndael CTR random access in Intel® Integrated Performance Primitives

Rijndael CTR random access

mickaelpic — Mon, 07 Jul 2008 20:16:28 GMT

Hi, I am trying to use the IPPS Rijndael capability.

So far everything is ok for encrypting and decrypting complete files.
But now I am trying to partially decrypt a file.
Given a block number I want to be able to start to decrypt the file from there instead of starting from the beginning.

The AES (Rijndael) CTR algorithm support such functionallity according to http://en.wikipedia.org/wiki/Cipher_block_chaining
In the algorithm, if I set the counter to the block index, and I provide the correct block, it will decrypt properlyfrom that block directly.

But the IPPs does not provide access to the counter.
At least no way I could find.

So am I missing something, or this functionnality was not provided in the IPPs?
I use IPPs5.3 Update 3 build 85.25, [5.3.471.85]

Here is a sample of the code I use to decrypt:

#define bufferSize 24576 // 24KB
#define aesBlkSize 32 // size of Rijndael-256 algorithm block is equal to 32

int ctxSize;
ippsRijndael256GetSize(&ctxSize);
// and allocate one
IppsRijndael256Spec* pCtx = (IppsRijndael256Spec*)( new Ipp8u [ctxSize] );
// and prepare the context for Rijndael256 usage
ippsRijndael256Init(key, IppsRijndaelKey256, pCtx);
Ipp8u crt[aesBlkSize];
int ctrNumBitSize = 64;
// init the counter
memcpy_s(crt, aesBlkSize, IV, aesBlkSize);
IppStatus status = ippStsNoErr;
len = (unsigned long)fread((char*)inBuffer, 1, bufferSize, fin);
status = ippsRijndael256DecryptCTR(inBuffer, outBuffer, len, pCtx, crt, ctrNumBitSize);

Re: Rijndael CTR random access

mickaelpic — Wed, 09 Jul 2008 19:57:52 GMT

I found my own answer.
The crt is actually the Nonce+counter
The documentation is not very clear about this.

Anyway, I now modify the crt directly to change the counter value and I can achieve partial decryption.

for (int i = 0; i < sizeof(unsigned); ++i)
{
crt[aesBlkSize - 2 - i] = ((bufferIndex * 3) >> (8 * i)) & 0xFF;
}

One thing I do not understand yet is why I have multiply the bufferIndex value by 3.

I have monitored the crt counter value while debugging and for each buffer I decrypt the counter is increased by 3.
As my code above shown I am also increasing by 3, but I would like to understand why.

Another thing, I can not set a counter value higher than 0xFFFFFFFF (4294967295).
If I try 0xFFFFFFFF01 for example, I end up with a counter value of 1 instead of 1099511627521.
I guess my code to convert the int to hex is not the best.
Any idea how to improve it?

Re: Rijndael CTR random access

mickaelpic — Fri, 11 Jul 2008 14:27:47 GMT

I answered one of my own question.

The increment of 3 is actually actually the number of 8K buffers that fits in a 24K buffersize

For a 64K buffersize you can fit 8,...

Re: Rijndael CTR random access

Vladimir_Dudnik — Tue, 29 Jul 2008 19:29:53 GMT

Glad to see you can figure oout how to use IPP functions

Regards,
Vladimir