leocalm/lyra

lyra's Issues

Feature: Usage of KangarooTwelve to replace SHA3, or modding BLAKE3 to be like BlaMka

So Keccak upped their game in 2018 and BLAKE retaliated in 2019. They reduced the rounds and made things faster.
Throwing these hashes into Lyra does not hurt, I hope.

• KangarooTwelve (from the Keccak team) https://github.com/XKCP/XKCP/tree/master/lib/high/KangarooTwelve
  • https://keccak.team/files/KangarooTwelve.pdf
• BLAKE3 (from the BLAKE team) https://github.com/BLAKE3-team/BLAKE3
  • https://github.com/BLAKE3-team/BLAKE3-specs/blob/master/blake3.pdf

Potential Defect

Good morning!

I am a developer with the Vertcoin Cryptocurrency team. We are using your Lyra2 algorithm as the basis of our currency's hashing.

We believe a potential defect exists within your implementation. A community member had this remark:

"please check this code
//================================ Setup Phase =============================//
//Absorbing salt, password and basil: this is the only place in which the block length is hard-coded to 512 bits
ptrWord = &wholeMatrix[0];
for (i = 0; i < nBlocksInput; i++) {
absorbBlockBlake2Safe(state, ptrWord); //absorbs each block of pad(pwd || salt || basil)
ptrWord += BLOCK_LEN_BLAKE2_SAFE_BYTES; //goes to next block of pad(pwd || salt || basil)
-------------------------------------------------------
BLOCK_LEN_BLAKE2_SAFE_BYTES is 64
the problem is
ptrWord is a 64 bit pointer,but the line ,it add 64,means it add 64*64=4096 bit=512 byte,
so ,the next round ,all ptrWord is zero,
that means Concatenates the basil and padding are not used"

Let us know your thoughts.

Cheers,

Adam

Another (not that feature-rich) Lyra2 implementation in Java

Hello,

this is a friendly heads-up for a Java implementation of Lyra2 written by me that I've recently open-sourced here: https://github.com/all3fox/lyra2-java

To be honest, not much is implemented there: a single-threaded version of Lyra2 with the Blake2b-based sponge. That implementation produces the same hash values as your current Lyra2 C implementation (of course, if the compile-time and run-time parameters like sponge/blocks/password/hash/rounds/columns etc. are properly matched).

However, I will be adding other sponges and refactoring current implementation even more. Feel free to come by and have a look.

Segmentation fault (sse)

• ptrWordOut points to an invalid location.

Program received signal SIGSEGV, Segmentation fault.
[Switching to Thread 0x7ffff75e7700 (LWP 10889)]
reducedDuplexRow1and2 (rowOut=<optimized out>, rowIn=0x7ffff00068c0, state=0x7ffff000c920) at sse/Sponge.c:168
168             ptrWordOut[j] = _mm_xor_si128(state[j], ptrWordIn[j]);
(gdb) bt
#0  reducedDuplexRow1and2 (rowOut=<optimized out>, rowIn=0x7ffff00068c0, state=0x7ffff000c920) at sse/Sponge.c:168
#1  LYRA2._omp_fn.0.2482 (.omp_data_i=0x7fffffffdc90) at sse/Lyra2.c:433
#2  0x00007ffff75f034a in ?? () from /usr/lib/x86_64-linux-gnu/libgomp.so.1
#3  0x00007ffff7bc4182 in start_thread (arg=0x7ffff75e7700) at pthread_create.c:312
#4  0x00007ffff78f147d in clone () at ../sysdeps/unix/sysv/linux/x86_64/clone.S:111

(gdb) print ptrWordOut
$1 = (__m128i *) 0x5fa0
(gdb) print j
$2 = 6

(gdb) print state
$3 = (__m128i *) 0x7ffff000c920
(gdb) print ptrWordIn
$4 = (__m128i *) 0x7ffff00068c0
(gdb) 

---

Inputs:
Password: password
Password Length: 8
Salt: salt.string
Salt Length: 11
Output Length: 64

Parameters:
T: 1
R: 4
C: 256
Parallelism: 2
Sponge: Blake2
Sponge Blocks (bitrate): 12 = 768 bits
Memory: 98304 bytes