The purpose of this project is:
- provide an API that hashes bytes, not bits
- provide a simple reference implementation of a SHA-3 message digest algorithm, as defined in the FIPS 202 standard
- assist developers in the Ethereum blockchain ecosystem by providing the Keccak function used there
- implement the hashing API that employs the IUF paradigm (or
Init
,Update
,Finalize
style) - answer the design questions, such as:
- what does the state for IUF look like?
- how small can the state be (224 bytes on a 64-bit system for a unified SHA-3 algorithm)
- what is the incremental cost of adding e.g. SHA3-384 to a SHA3-256 implementation?
The implementation is written in C and uses uint64_t
types to manage the SHA-3 state. The code will compile and run on 64-bit and 32-bit architectures (gcc
and gcc -m32
on x86_64
were tested).
This work is licensed with a standard MIT license. I appreciate, but do not require, any attribution to this work if you used the code or ideas. I thank you for this in advance.
This is a clean-room implementation of IUF API for SHA3. The keccakf()
is based on the code from keccak.noekeon.org.
1600-bit message hashing test vectors are NIST test vectors.
Let's hash 'abc' with SHA3-256 using two methods: single buffer (but using IUF paradigm), and using the IUF API.
sha3_context c;
uint8_t *hash;
Single-buffer hashing:
sha3_Init256(&c);
sha3_Update(&c, "abc", 3);
hash = sha3_Finalize(&c);
// 'hash' points to a buffer inside 'c'
// with the value of SHA3-256
Alternatively, IUF hashing:
sha3_Init256(&c);
sha3_Update(&c, "a", 1);
sha3_Update(&c, "bc", 2);
hash = sha3_Finalize(&c);
// no free for 'c' is needed
The hash
points to the same 256/8=32
bytes in both cases.
There is also a single-call hashing API:
sha3_HashBuffer(256, SHA3_FLAGS_KECCAK, "abc", 3, out, sizeof(out));
// out contains 256 bits of Keccak256, or less if sizeof(out)<32
Call sha3_SetFlags(&c, SHA3_FLAGS_KECCAK)
immediately after sha3_InitX
or no later than sha3_Finalize
. This change cannot be undone for the given hash context.
$ make
See Makefile
for details. See also below for specific examples.
$ make test
Keccak-256 tests passed OK
SHA3-256, SHA3-384, SHA3-512 tests passed OK
or
$ make CFLAGS=-m32 LDFLAGS=-m32 test
Keccak-256 tests passed OK
SHA3-256, SHA3-384, SHA3-512 tests passed OK
There is also sha3sum
test program that takes following parameters:
sha3sum 256|384|512 file_path
or for Keccak version:
sha3sum 256|384|512 -k file_path
$ touch empty.txt
$ gcc -Wall sha3.c sha3sum.c -o sha3sum && ./sha3sum 256 empty.txt
a7ffc6f8bf1ed76651c14756a061d662f580ff4de43b49fa82d80a4b80f8434a empty.txt
Compare with Linux sha3sum
:
$ sha3sum -a 256 empty.txt
a7ffc6f8bf1ed76651c14756a061d662f580ff4de43b49fa82d80a4b80f8434a empty.txt
$ echo -n "abc" > abc
$ sha3sum 256 -k abc
4e03657aea45a94fc7d47ba826c8d667c0d1e6e33a64a036ec44f58fa12d6c45 abc
This corresponds to the result obtained in Solidity JavaScript test framework.
console.log(web3.utils.sha3('abc'));
// prints 0x4e03657aea45a94fc7d47ba826c8d667c0d1e6e33a64a036ec44f58fa12d6c45
- the same
sha3_context
object maintains the state for SHA3-256, SHA3-384, or SHA3-512 algorithm; - the hash algorithm used is determined by how the context was initialized with
sha3_InitX
, e.g.sha3_Init256
,sha3_Init384
, orsha3_Init512
call; sha3_Update
andsha3_Finalize
are the same for regardless the type of the algorithm (X
);- the buffer returned by
sha3_Finalize
will haveX
bits of hash; sha3_InitX
also works as Reset (zeroization) of the hash context; no Free function is needed;
See sha3.h
for the exact interface.
$ fuzz/run.sh
The fuzzing script expects clang installed.
Thanks to @ralight for moving the test code into separate sha3test.c
SHA3-224 is not supported, but can easily be added.
The code was written to work with the Microsoft Visual Studio compiler (under _MSC_VER
), but this build target was not tested.
This project was created to support SHA3 in OpenPGP work, but it applies to other protocols and formats, e.g. TLS.