Factor out SHA3 from msrln16.

This commit is contained in:
Alex Parent 2016-11-25 12:29:26 -05:00
parent 8f706feb47
commit 4764de9963
5 changed files with 10 additions and 366 deletions

View File

@ -17,7 +17,7 @@
#include "LatticeCrypto_priv.h"
#include "oqs/rand.h"
#include "external/shake128.h"
#include <oqs/sha3.h>
extern const int32_t psi_rev_ntt1024_12289[1024];
extern const int32_t omegainv_rev_ntt1024_12289[1024];
@ -25,7 +25,6 @@ extern const int32_t omegainv10N_rev_ntt1024_12289;
extern const int32_t Ninv11_ntt1024_12289;
// import external code
#include "external/shake128.c"
#ifdef RLWE_ASM_AVX2
#include "AMD64/consts.c"
#include "AMD64/ntt_x64.c"
@ -296,10 +295,10 @@ CRYPTO_STATUS oqs_rlwe_msrln16_generate_a(uint32_t *a, const unsigned char *seed
unsigned int pos = 0, ctr = 0;
uint16_t val;
unsigned int nblocks = 16;
uint8_t buf[SHAKE128_RATE * 16];
unsigned char state[SHAKE128_STATE_SIZE] = { 0 };
FIPS202_SHAKE128_Absorb(seed, OQS_RLWE_MSRLN16_SEED_BYTES, state, sizeof(state));
FIPS202_SHAKE128_Squeeze(state, (unsigned char *)buf, nblocks * SHAKE128_RATE);
uint8_t buf[OQS_SHA3_SHAKE128_RATE * 16];
uint64_t state[OQS_SHA3_STATESIZE];
OQS_SHA3_shake128_absorb(state, seed, OQS_RLWE_MSRLN16_SEED_BYTES);
OQS_SHA3_shake128_squeezeblocks((unsigned char *)buf, nblocks, state);
while (ctr < OQS_RLWE_MSRLN16_PARAMETER_N) {
val = (buf[pos] | ((uint16_t)buf[pos + 1] << 8)) & 0x3fff;
@ -307,9 +306,9 @@ CRYPTO_STATUS oqs_rlwe_msrln16_generate_a(uint32_t *a, const unsigned char *seed
a[ctr++] = val;
}
pos += 2;
if (pos > SHAKE128_RATE * nblocks - 2) {
if (pos > OQS_SHA3_SHAKE128_RATE * nblocks - 2) {
nblocks = 1;
FIPS202_SHAKE128_Squeeze(state, (unsigned char *)buf, nblocks * SHAKE128_RATE);
OQS_SHA3_shake128_squeezeblocks((unsigned char *)buf, nblocks, state);
pos = 0;
}
}

View File

@ -1,332 +0,0 @@
#if defined(WINDOWS)
#define UNUSED
#else
#define UNUSED __attribute__ ((unused))
#endif
/*
Original implementation modified to allow spliting the absorb and squeeze
phases of Keccak.
*/
/*
Implementation by the Keccak, Keyak and Ketje Teams, namely, Guido Bertoni,
Joan Daemen, Michaël Peeters, Gilles Van Assche and Ronny Van Keer, hereby
denoted as "the implementer".
For more information, feedback or questions, please refer to our websites:
http://keccak.noekeon.org/
http://keyak.noekeon.org/
http://ketje.noekeon.org/
To the extent possible under law, the implementer has waived all copyright
and related or neighboring rights to the source code in this file.
http://creativecommons.org/publicdomain/zero/1.0/
*/
/*
================================================================
The purpose of this source file is to demonstrate a readable and compact
implementation of all the Keccak instances approved in the FIPS 202 standard,
including the hash functions and the extendable-output functions (XOFs).
We focused on clarity and on source-code compactness,
rather than on the performance.
The advantages of this implementation are:
+ The source code is compact, after removing the comments, that is. :-)
+ There are no tables with arbitrary constants.
+ For clarity, the comments link the operations to the specifications using
the same notation as much as possible.
+ There is no restriction in cryptographic features. In particular,
the SHAKE128 and SHAKE256 XOFs can produce any output length.
+ The code does not use much RAM, as all operations are done in place.
The drawbacks of this implementation are:
- There is no message queue. The whole message must be ready in a buffer.
- It is not optimized for peformance.
The implementation is even simpler on a little endian platform. Just define the
LITTLE_ENDIAN symbol in that case.
For a more complete set of implementations, please refer to
the Keccak Code Package at https://github.com/gvanas/KeccakCodePackage
For more information, please refer to:
* [Keccak Reference] http://keccak.noekeon.org/Keccak-reference-3.0.pdf
* [Keccak Specifications Summary] http://keccak.noekeon.org/specs_summary.html
This file uses UTF-8 encoding, as some comments use Greek letters.
================================================================
*/
/**
* Function to compute the Keccak[r, c] sponge function over a given input.
* @param rate The value of the rate r.
* @param capacity The value of the capacity c.
* @param input Pointer to the input message.
* @param inputByteLen The number of input bytes provided in the input message.
* @param delimitedSuffix Bits that will be automatically appended to the end
* of the input message, as in domain separation.
* This is a byte containing from 0 to 7 bits
* These <i>n</i> bits must be in the least significant bit positions
* and must be delimited with a bit 1 at position <i>n</i>
* (counting from 0=LSB to 7=MSB) and followed by bits 0
* from position <i>n</i>+1 to position 7.
* Some examples:
* - If no bits are to be appended, then @a delimitedSuffix must be 0x01.
* - If the 2-bit sequence 0,1 is to be appended (as for SHA3-*), @a delimitedSuffix must be 0x06.
* - If the 4-bit sequence 1,1,1,1 is to be appended (as for SHAKE*), @a delimitedSuffix must be 0x1F.
* - If the 7-bit sequence 1,1,0,1,0,0,0 is to be absorbed, @a delimitedSuffix must be 0x8B.
* @param output Pointer to the buffer where to store the output.
* @param outputByteLen The number of output bytes desired.
* @pre One must have r+c=1600 and the rate a multiple of 8 bits in this implementation.
*/
UNUSED static void Keccak(unsigned int rate, unsigned int capacity, const unsigned char *input, unsigned long long int inputByteLen, unsigned char delimitedSuffix, unsigned char *output, unsigned long long int outputByteLen);
/*
* Performs the Keccak absorb phase. Same parameters as the Keccak function, but a SHAKE128_STATE_SIZE-byte state must also be provided.
* The Keccak_squeeze function can be called successively to generate output.
*/
static void Keccak_absorb(unsigned int rate, unsigned int capacity, const unsigned char *input, unsigned long long int inputByteLen, unsigned char delimitedSuffix, unsigned char* state, unsigned int stateLen);
/*
* Performs the Keccak squeeze phase. Same parameters as the Keccak function, but a SHAKE128_STATE_SIZE-byte state must also be provided.
* The Keccak_absorb function must be called first.
*/
static void Keccak_squeeze(unsigned int rate, unsigned int capacity, unsigned char* state, unsigned char *output, unsigned long long int outputByteLen);
static void FIPS202_SHAKE128_Absorb(const unsigned char *input, unsigned int inputByteLen, unsigned char* state, unsigned int stateLen)
{
Keccak_absorb(1344, 256, input, inputByteLen, 0x1F, state, stateLen);
}
static void FIPS202_SHAKE128_Squeeze(unsigned char* state, unsigned char *output, int outputByteLen)
{
Keccak_squeeze(1344, 256, state, output, outputByteLen);
}
UNUSED static void FIPS202_SHAKE128(const unsigned char *input, unsigned int inputByteLen, unsigned char *output, int outputByteLen, unsigned int stateLen)
{
unsigned char state[200] = { 0 };
FIPS202_SHAKE128_Absorb(input, inputByteLen, state, stateLen);
FIPS202_SHAKE128_Squeeze(state, output, outputByteLen);
}
/*
================================================================
Technicalities
================================================================
*/
typedef unsigned char UINT8;
typedef unsigned long long int UINT64;
typedef UINT64 tKeccakLane;
#ifndef LITTLE_ENDIAN
/** Function to load a 64-bit value using the little-endian (LE) convention.
* On a LE platform, this could be greatly simplified using a cast.
*/
static UINT64 load64(const UINT8 *x)
{
int i;
UINT64 u=0;
for(i=7; i>=0; --i) {
u <<= 8;
u |= x[i];
}
return u;
}
/** Function to store a 64-bit value using the little-endian (LE) convention.
* On a LE platform, this could be greatly simplified using a cast.
*/
static void store64(UINT8 *x, UINT64 u)
{
unsigned int i;
for(i=0; i<8; ++i) {
x[i] = u;
u >>= 8;
}
}
/** Function to XOR into a 64-bit value using the little-endian (LE) convention.
* On a LE platform, this could be greatly simplified using a cast.
*/
static void xor64(UINT8 *x, UINT64 u)
{
unsigned int i;
for(i=0; i<8; ++i) {
x[i] ^= u;
u >>= 8;
}
}
#endif
/*
================================================================
A readable and compact implementation of the Keccak-f[1600] permutation.
================================================================
*/
#define ROL64(a, offset) ((((UINT64)a) << offset) ^ (((UINT64)a) >> (64-offset)))
#define i(x, y) ((x)+5*(y))
#ifdef LITTLE_ENDIAN
#define readLane(x, y) (((tKeccakLane*)state)[i(x, y)])
#define writeLane(x, y, lane) (((tKeccakLane*)state)[i(x, y)]) = (lane)
#define XORLane(x, y, lane) (((tKeccakLane*)state)[i(x, y)]) ^= (lane)
#else
#define readLane(x, y) load64((UINT8*)state+sizeof(tKeccakLane)*i(x, y))
#define writeLane(x, y, lane) store64((UINT8*)state+sizeof(tKeccakLane)*i(x, y), lane)
#define XORLane(x, y, lane) xor64((UINT8*)state+sizeof(tKeccakLane)*i(x, y), lane)
#endif
/**
* Function that computes the linear feedback shift register (LFSR) used to
* define the round constants (see [Keccak Reference, Section 1.2]).
*/
static int LFSR86540(UINT8 *LFSR)
{
int result = ((*LFSR) & 0x01) != 0;
if (((*LFSR) & 0x80) != 0)
// Primitive polynomial over GF(2): x^8+x^6+x^5+x^4+1
(*LFSR) = ((*LFSR) << 1) ^ 0x71;
else
(*LFSR) <<= 1;
return result;
}
/**
* Function that computes the Keccak-f[1600] permutation on the given state.
*/
static void KeccakF1600_StatePermute(void *state)
{
unsigned int round, x, y, j, t;
UINT8 LFSRstate = 0x01;
for(round=0; round<24; round++) {
{ // === θ step (see [Keccak Reference, Section 2.3.2]) ===
tKeccakLane C[5], D;
// Compute the parity of the columns
for(x=0; x<5; x++)
C[x] = readLane(x, 0) ^ readLane(x, 1) ^ readLane(x, 2) ^ readLane(x, 3) ^ readLane(x, 4);
for(x=0; x<5; x++) {
// Compute the θ effect for a given column
D = C[(x+4)%5] ^ ROL64(C[(x+1)%5], 1);
// Add the θ effect to the whole column
for (y=0; y<5; y++)
XORLane(x, y, D);
}
}
{ // === ρ and π steps (see [Keccak Reference, Sections 2.3.3 and 2.3.4]) ===
tKeccakLane current, temp;
// Start at coordinates (1 0)
x = 1; y = 0;
current = readLane(x, y);
// Iterate over ((0 1)(2 3))^t * (1 0) for 0 ≤ t ≤ 23
for(t=0; t<24; t++) {
// Compute the rotation constant r = (t+1)(t+2)/2
unsigned int r = ((t+1)*(t+2)/2)%64;
// Compute ((0 1)(2 3)) * (x y)
unsigned int Y = (2*x+3*y)%5; x = y; y = Y;
// Swap current and state(x,y), and rotate
temp = readLane(x, y);
writeLane(x, y, ROL64(current, r));
current = temp;
}
}
{ // === χ step (see [Keccak Reference, Section 2.3.1]) ===
tKeccakLane temp[5];
for(y=0; y<5; y++) {
// Take a copy of the plane
for(x=0; x<5; x++)
temp[x] = readLane(x, y);
// Compute χ on the plane
for(x=0; x<5; x++)
writeLane(x, y, temp[x] ^((~temp[(x+1)%5]) & temp[(x+2)%5]));
}
}
{ // === ι step (see [Keccak Reference, Section 2.3.5]) ===
for(j=0; j<7; j++) {
unsigned int bitPosition = (1<<j)-1; //2^j-1
if (LFSR86540(&LFSRstate))
XORLane(0, 0, (tKeccakLane)1<<bitPosition);
}
}
}
}
/*
================================================================
A readable and compact implementation of the Keccak sponge functions
that use the Keccak-f[1600] permutation.
================================================================
*/
#include <string.h>
#define MIN(a, b) ((a) < (b) ? (a) : (b))
void Keccak_absorb(unsigned int rate, unsigned int capacity, const unsigned char *input, unsigned long long int inputByteLen, unsigned char delimitedSuffix, unsigned char* state, unsigned int stateLen)
{
unsigned int rateInBytes = rate/8;
unsigned int blockSize = 0;
unsigned int i;
if (((rate + capacity) != 1600) || ((rate % 8) != 0))
return;
// === Initialize the state ===
memset(state, 0, stateLen);
// === Absorb all the input blocks ===
while(inputByteLen > 0) {
blockSize = MIN(inputByteLen, rateInBytes);
for(i=0; i<blockSize; i++)
state[i] ^= input[i];
input += blockSize;
inputByteLen -= blockSize;
if (blockSize == rateInBytes) {
KeccakF1600_StatePermute(state);
blockSize = 0;
}
}
// === Do the padding and switch to the squeezing phase ===
// Absorb the last few bits and add the first bit of padding (which coincides with the delimiter in delimitedSuffix)
state[blockSize] ^= delimitedSuffix;
// If the first bit of padding is at position rate-1, we need a whole new block for the second bit of padding
if (((delimitedSuffix & 0x80) != 0) && (blockSize == (rateInBytes-1)))
KeccakF1600_StatePermute(state);
// Add the second bit of padding
state[rateInBytes-1] ^= 0x80;
}
void Keccak_squeeze(unsigned int rate, UNUSED unsigned int capacity, unsigned char* state, unsigned char *output, unsigned long long int outputByteLen)
{
unsigned int blockSize = 0;
unsigned int rateInBytes = rate / 8;
// Switch to the squeezing phase
KeccakF1600_StatePermute(state);
// === Squeeze out all the output blocks ===
while (outputByteLen > 0) {
blockSize = MIN(outputByteLen, rateInBytes);
memcpy(output, state, blockSize);
output += blockSize;
outputByteLen -= blockSize;
if (outputByteLen > 0)
KeccakF1600_StatePermute(state);
}
}

View File

@ -1,24 +0,0 @@
#ifndef SHAKE128_H
#define SHAKE128_H
#define SHAKE128_STATE_SIZE 200
#define SHAKE128_RATE 168
/*
* Computes SHAKE128 on the array "input" of length "inputByteLen", resulting in "outputByteLen" bytes stored in "output".
*/
static void FIPS202_SHAKE128(const unsigned char *input, unsigned int inputByteLen, unsigned char *output, int outputByteLen, unsigned int stateLen);
/*
* Performs the absorb phase of SHAKE128: ingests the "inputByteLen" bytes stored in "input"; storing the internal
* SHAKE128 state of length SHAKE128_STATE_SIZE in "state".
*/
static void FIPS202_SHAKE128_Absorb(const unsigned char *input, unsigned int inputByteLen, unsigned char* state, unsigned int stateLen);
/*
* Performs the squeeze phase of SHAKE128: generates "outputByteLen" bytes stored in "output" from the "state" of length
* SHAKE128_STATE_SIZE. Must be preceeded by a call to FIPS202_SHAKE128_Absorb.
*/
static void FIPS202_SHAKE128_Squeeze(unsigned char* state, unsigned char *output, int outputByteLen);
#endif

View File

@ -142,9 +142,9 @@ static void poly_tobytes(unsigned char *r, const poly *p) {
static void poly_uniform(poly *a, const unsigned char *seed) {
unsigned int pos = 0, ctr = 0;
uint16_t val;
uint64_t state[25];
uint64_t state[OQS_SHA3_STATESIZE];
unsigned int nblocks = 16;
uint8_t buf[2688]; // SHAKE128_RATE * nblocks
uint8_t buf[OQS_SHA3_SHAKE128_RATE * 16];
OQS_SHA3_shake128_absorb(state, seed, NEWHOPE_SEEDBYTES);

View File

@ -6,6 +6,7 @@
#ifndef __OQS_SHA3_H
#define __OQS_SHA3_H
#define OQS_SHA3_STATESIZE 25
#define OQS_SHA3_SHAKE128_RATE 168
#define OQS_SHA3_SHA3_256_RATE 136