bACI_ECBencodeStripe() returning wrong result

[mguc]

Hello everyone

It seems that there is a bug in the AES Coprocessor API which occurs sometimes (in most cases the result is correct). Below you can find a code snippet which encodes a plaintext (16 bytes) and returns the encoded data in the same array. The code runs on an JN5168-M05 module. To ensure that there is no "buffer" issue I used different tsReg128 structs for the result and the input data.

static void
encrypt(uint8_t *plaintext_and_result)
{
   HEXDUMP("aes_128_key", &jn516x_aes_128_key, sizeof(tsReg128));
   HEXDUMP("plaintext", plaintext_and_result, sizeof(tsReg128));
   memcpy(&jn516x_aes_128_plaintext, plaintext_and_result, sizeof(tsReg128));
   HEXDUMP("aes_128_plaintext", &jn516x_aes_128_plaintext, sizeof(tsReg128));
   if(!bACI_ECBencodeStripe(&jn516x_aes_128_key,
       jn516x_aes_128_key_changed,
       &jn516x_aes_128_plaintext,
       &jn516x_aes_128_result)) {
       ERROR("Could not encrypt...");
   }
   ERROR("new key: %u\n", jn516x_aes_128_key_changed);
   HEXDUMP("aes_128_result", &jn516x_aes_128_result, sizeof(tsReg128));
   memcpy(plaintext_and_result, &jn516x_aes_128_result, sizeof(tsReg128));
   HEXDUMP("result", plaintext_and_result, sizeof(tsReg128));
}‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍

Log output of node 1:

aes_128_key:
 0000 53 d6 00 56 6e 33 25 50 a3 62 db 17 f5 b2 a3 27 S..Vn3%P.b.....'
plaintext:
 0000 26 0c ab c0 9c b1 3b 71 e8 3d 47 f1 d5 22 ab 20 &.....;q.=G..". 
aes_128_plaintext:
 0000 26 0c ab c0 9c b1 3b 71 e8 3d 47 f1 d5 22 ab 20 &.....;q.=G..". 
new key: 0
aes_128_result:
 0000 cf b4 e9 86 c5 4b 9a 1f 6f b2 ff 49 5e 91 0a dc .....K..o..I^...
result:
 0000 cf b4 e9 86 c5 4b 9a 1f 6f b2 ff 49 5e 91 0a dc .....K..o..I^...
‍‍‍‍‍‍‍‍‍‍‍

Log output on node 2:

aes_128_key:
 0000 53 d6 00 56 6e 33 25 50 a3 62 db 17 f5 b2 a3 27 S..Vn3%P.b.....'
plaintext:
 0000 26 0c ab c0 9c b1 3b 71 e8 3d 47 f1 d5 22 ab 20 &.....;q.=G..".
result:
 0000 d0 95 5a 61 cc 17 3e 64 f6 7c 2d e2 01 f3 31 58 ..Za..>d.|-...1X‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍

Log output of another AES implementation:

Key:
 0000 53 d6 00 56 6e 33 25 50 a3 62 db 17 f5 b2 a3 27 S..Vn3%P.b.....'
Data:
 0000 26 0c ab c0 9c b1 3b 71 e8 3d 47 f1 d5 22 ab 20 &.....;q.=G..".
Result:
 0000 d0 95 5a 61 cc 17 3e 64 f6 7c 2d e2 01 f3 31 58 ..Za..>d.|-...1X‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍

Thanks for your help!

Marko

[mguc]

Hi Ovidiu

Thank you for your help. It indeed had something to do with jn516x_aes_128_key_changed  flag. Setting it to true all the time fixed the issue. The problem was that we used the bACI_CCMstar() function in another part of our application, which apparently uses the same key memory as bACI_ECBencodeStripe(). Even if the local key did not change, it might have changed by calling the other function.

With kind regards

Marko

[ovidiu_usturoi · Accepted solution]

Hi Marko,

Did you get the same result when the jn516x_aes_128_key_changed  is TRUE?

Could you also print the result of the function?

Regards,

Ovidiu