crypto/sha3: Copy latest code from "golang.org/x/crypto/sha3"

Revision: 1f22c0103821b9390939b6776727195525381532

13 files changed, 1269 insertions, 0 deletions

diff --git a/crypto/sha3/LICENSE b/crypto/sha3/LICENSE
new file mode 100644
index 000000000..6a66aea5e
--- /dev/null
+++ b/crypto/sha3/LICENSE
@@ -0,0 +1,27 @@
+Copyright (c) 2009 The Go Authors. All rights reserved.
+
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+
+   * Redistributions of source code must retain the above copyright
+notice, this list of conditions and the following disclaimer.
+   * Redistributions in binary form must reproduce the above
+copyright notice, this list of conditions and the following disclaimer
+in the documentation and/or other materials provided with the
+distribution.
+   * Neither the name of Google Inc. nor the names of its
+contributors may be used to endorse or promote products derived from
+this software without specific prior written permission.
+
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
diff --git a/crypto/sha3/PATENTS b/crypto/sha3/PATENTS
new file mode 100644
index 000000000..733099041
--- /dev/null
+++ b/crypto/sha3/PATENTS
@@ -0,0 +1,22 @@
+Additional IP Rights Grant (Patents)
+
+"This implementation" means the copyrightable works distributed by
+Google as part of the Go project.
+
+Google hereby grants to You a perpetual, worldwide, non-exclusive,
+no-charge, royalty-free, irrevocable (except as stated in this section)
+patent license to make, have made, use, offer to sell, sell, import,
+transfer and otherwise run, modify and propagate the contents of this
+implementation of Go, where such license applies only to those patent
+claims, both currently owned or controlled by Google and acquired in
+the future, licensable by Google that are necessarily infringed by this
+implementation of Go.  This grant does not include claims that would be
+infringed only as a consequence of further modification of this
+implementation.  If you or your agent or exclusive licensee institute or
+order or agree to the institution of patent litigation against any
+entity (including a cross-claim or counterclaim in a lawsuit) alleging
+that this implementation of Go or any code incorporated within this
+implementation of Go constitutes direct or contributory patent
+infringement, or inducement of patent infringement, then any patent
+rights granted to you under this License for this implementation of Go
+shall terminate as of the date such litigation is filed.
diff --git a/crypto/sha3/doc.go b/crypto/sha3/doc.go
new file mode 100644
index 000000000..a0ee3ae72
--- /dev/null
+++ b/crypto/sha3/doc.go
@@ -0,0 +1,66 @@
+// Copyright 2014 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+// Package sha3 implements the SHA-3 fixed-output-length hash functions and
+// the SHAKE variable-output-length hash functions defined by FIPS-202.
+//
+// Both types of hash function use the "sponge" construction and the Keccak
+// permutation. For a detailed specification see http://keccak.noekeon.org/
+//
+//
+// Guidance
+//
+// If you aren't sure what function you need, use SHAKE256 with at least 64
+// bytes of output. The SHAKE instances are faster than the SHA3 instances;
+// the latter have to allocate memory to conform to the hash.Hash interface.
+//
+// If you need a secret-key MAC (message authentication code), prepend the
+// secret key to the input, hash with SHAKE256 and read at least 32 bytes of
+// output.
+//
+//
+// Security strengths
+//
+// The SHA3-x (x equals 224, 256, 384, or 512) functions have a security
+// strength against preimage attacks of x bits. Since they only produce "x"
+// bits of output, their collision-resistance is only "x/2" bits.
+//
+// The SHAKE-256 and -128 functions have a generic security strength of 256 and
+// 128 bits against all attacks, provided that at least 2x bits of their output
+// is used.  Requesting more than 64 or 32 bytes of output, respectively, does
+// not increase the collision-resistance of the SHAKE functions.
+//
+//
+// The sponge construction
+//
+// A sponge builds a pseudo-random function from a public pseudo-random
+// permutation, by applying the permutation to a state of "rate + capacity"
+// bytes, but hiding "capacity" of the bytes.
+//
+// A sponge starts out with a zero state. To hash an input using a sponge, up
+// to "rate" bytes of the input are XORed into the sponge's state. The sponge
+// is then "full" and the permutation is applied to "empty" it. This process is
+// repeated until all the input has been "absorbed". The input is then padded.
+// The digest is "squeezed" from the sponge in the same way, except that output
+// output is copied out instead of input being XORed in.
+//
+// A sponge is parameterized by its generic security strength, which is equal
+// to half its capacity; capacity + rate is equal to the permutation's width.
+// Since the KeccakF-1600 permutation is 1600 bits (200 bytes) wide, this means
+// that the security strength of a sponge instance is equal to (1600 - bitrate) / 2.
+//
+//
+// Recommendations
+//
+// The SHAKE functions are recommended for most new uses. They can produce
+// output of arbitrary length. SHAKE256, with an output length of at least
+// 64 bytes, provides 256-bit security against all attacks.  The Keccak team
+// recommends it for most applications upgrading from SHA2-512. (NIST chose a
+// much stronger, but much slower, sponge instance for SHA3-512.)
+//
+// The SHA-3 functions are "drop-in" replacements for the SHA-2 functions.
+// They produce output of the same length, with the same security strengths
+// against all attacks. This means, in particular, that SHA3-256 only has
+// 128-bit collision resistance, because its output length is 32 bytes.
+package sha3 // import "golang.org/x/crypto/sha3"
diff --git a/crypto/sha3/hashes.go b/crypto/sha3/hashes.go
new file mode 100644
index 000000000..2b51cf4e9
--- /dev/null
+++ b/crypto/sha3/hashes.go
@@ -0,0 +1,65 @@
+// Copyright 2014 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package sha3
+
+// This file provides functions for creating instances of the SHA-3
+// and SHAKE hash functions, as well as utility functions for hashing
+// bytes.
+
+import (
+	"hash"
+)
+
+// New224 creates a new SHA3-224 hash.
+// Its generic security strength is 224 bits against preimage attacks,
+// and 112 bits against collision attacks.
+func New224() hash.Hash { return &state{rate: 144, outputLen: 28, dsbyte: 0x06} }
+
+// New256 creates a new SHA3-256 hash.
+// Its generic security strength is 256 bits against preimage attacks,
+// and 128 bits against collision attacks.
+func New256() hash.Hash { return &state{rate: 136, outputLen: 32, dsbyte: 0x06} }
+
+// New384 creates a new SHA3-384 hash.
+// Its generic security strength is 384 bits against preimage attacks,
+// and 192 bits against collision attacks.
+func New384() hash.Hash { return &state{rate: 104, outputLen: 48, dsbyte: 0x06} }
+
+// New512 creates a new SHA3-512 hash.
+// Its generic security strength is 512 bits against preimage attacks,
+// and 256 bits against collision attacks.
+func New512() hash.Hash { return &state{rate: 72, outputLen: 64, dsbyte: 0x06} }
+
+// Sum224 returns the SHA3-224 digest of the data.
+func Sum224(data []byte) (digest [28]byte) {
+	h := New224()
+	h.Write(data)
+	h.Sum(digest[:0])
+	return
+}
+
+// Sum256 returns the SHA3-256 digest of the data.
+func Sum256(data []byte) (digest [32]byte) {
+	h := New256()
+	h.Write(data)
+	h.Sum(digest[:0])
+	return
+}
+
+// Sum384 returns the SHA3-384 digest of the data.
+func Sum384(data []byte) (digest [48]byte) {
+	h := New384()
+	h.Write(data)
+	h.Sum(digest[:0])
+	return
+}
+
+// Sum512 returns the SHA3-512 digest of the data.
+func Sum512(data []byte) (digest [64]byte) {
+	h := New512()
+	h.Write(data)
+	h.Sum(digest[:0])
+	return
+}
diff --git a/crypto/sha3/keccakf.go b/crypto/sha3/keccakf.go
new file mode 100644
index 000000000..13e7058fa
--- /dev/null
+++ b/crypto/sha3/keccakf.go
@@ -0,0 +1,410 @@
+// Copyright 2014 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package sha3
+
+// rc stores the round constants for use in the ι step.
+var rc = [24]uint64{
+	0x0000000000000001,
+	0x0000000000008082,
+	0x800000000000808A,
+	0x8000000080008000,
+	0x000000000000808B,
+	0x0000000080000001,
+	0x8000000080008081,
+	0x8000000000008009,
+	0x000000000000008A,
+	0x0000000000000088,
+	0x0000000080008009,
+	0x000000008000000A,
+	0x000000008000808B,
+	0x800000000000008B,
+	0x8000000000008089,
+	0x8000000000008003,
+	0x8000000000008002,
+	0x8000000000000080,
+	0x000000000000800A,
+	0x800000008000000A,
+	0x8000000080008081,
+	0x8000000000008080,
+	0x0000000080000001,
+	0x8000000080008008,
+}
+
+// keccakF1600 applies the Keccak permutation to a 1600b-wide
+// state represented as a slice of 25 uint64s.
+func keccakF1600(a *[25]uint64) {
+	// Implementation translated from Keccak-inplace.c
+	// in the keccak reference code.
+	var t, bc0, bc1, bc2, bc3, bc4, d0, d1, d2, d3, d4 uint64
+
+	for i := 0; i < 24; i += 4 {
+		// Combines the 5 steps in each round into 2 steps.
+		// Unrolls 4 rounds per loop and spreads some steps across rounds.
+
+		// Round 1
+		bc0 = a[0] ^ a[5] ^ a[10] ^ a[15] ^ a[20]
+		bc1 = a[1] ^ a[6] ^ a[11] ^ a[16] ^ a[21]
+		bc2 = a[2] ^ a[7] ^ a[12] ^ a[17] ^ a[22]
+		bc3 = a[3] ^ a[8] ^ a[13] ^ a[18] ^ a[23]
+		bc4 = a[4] ^ a[9] ^ a[14] ^ a[19] ^ a[24]
+		d0 = bc4 ^ (bc1<<1 | bc1>>63)
+		d1 = bc0 ^ (bc2<<1 | bc2>>63)
+		d2 = bc1 ^ (bc3<<1 | bc3>>63)
+		d3 = bc2 ^ (bc4<<1 | bc4>>63)
+		d4 = bc3 ^ (bc0<<1 | bc0>>63)
+
+		bc0 = a[0] ^ d0
+		t = a[6] ^ d1
+		bc1 = t<<44 | t>>(64-44)
+		t = a[12] ^ d2
+		bc2 = t<<43 | t>>(64-43)
+		t = a[18] ^ d3
+		bc3 = t<<21 | t>>(64-21)
+		t = a[24] ^ d4
+		bc4 = t<<14 | t>>(64-14)
+		a[0] = bc0 ^ (bc2 &^ bc1) ^ rc[i]
+		a[6] = bc1 ^ (bc3 &^ bc2)
+		a[12] = bc2 ^ (bc4 &^ bc3)
+		a[18] = bc3 ^ (bc0 &^ bc4)
+		a[24] = bc4 ^ (bc1 &^ bc0)
+
+		t = a[10] ^ d0
+		bc2 = t<<3 | t>>(64-3)
+		t = a[16] ^ d1
+		bc3 = t<<45 | t>>(64-45)
+		t = a[22] ^ d2
+		bc4 = t<<61 | t>>(64-61)
+		t = a[3] ^ d3
+		bc0 = t<<28 | t>>(64-28)
+		t = a[9] ^ d4
+		bc1 = t<<20 | t>>(64-20)
+		a[10] = bc0 ^ (bc2 &^ bc1)
+		a[16] = bc1 ^ (bc3 &^ bc2)
+		a[22] = bc2 ^ (bc4 &^ bc3)
+		a[3] = bc3 ^ (bc0 &^ bc4)
+		a[9] = bc4 ^ (bc1 &^ bc0)
+
+		t = a[20] ^ d0
+		bc4 = t<<18 | t>>(64-18)
+		t = a[1] ^ d1
+		bc0 = t<<1 | t>>(64-1)
+		t = a[7] ^ d2
+		bc1 = t<<6 | t>>(64-6)
+		t = a[13] ^ d3
+		bc2 = t<<25 | t>>(64-25)
+		t = a[19] ^ d4
+		bc3 = t<<8 | t>>(64-8)
+		a[20] = bc0 ^ (bc2 &^ bc1)
+		a[1] = bc1 ^ (bc3 &^ bc2)
+		a[7] = bc2 ^ (bc4 &^ bc3)
+		a[13] = bc3 ^ (bc0 &^ bc4)
+		a[19] = bc4 ^ (bc1 &^ bc0)
+
+		t = a[5] ^ d0
+		bc1 = t<<36 | t>>(64-36)
+		t = a[11] ^ d1
+		bc2 = t<<10 | t>>(64-10)
+		t = a[17] ^ d2
+		bc3 = t<<15 | t>>(64-15)
+		t = a[23] ^ d3
+		bc4 = t<<56 | t>>(64-56)
+		t = a[4] ^ d4
+		bc0 = t<<27 | t>>(64-27)
+		a[5] = bc0 ^ (bc2 &^ bc1)
+		a[11] = bc1 ^ (bc3 &^ bc2)
+		a[17] = bc2 ^ (bc4 &^ bc3)
+		a[23] = bc3 ^ (bc0 &^ bc4)
+		a[4] = bc4 ^ (bc1 &^ bc0)
+
+		t = a[15] ^ d0
+		bc3 = t<<41 | t>>(64-41)
+		t = a[21] ^ d1
+		bc4 = t<<2 | t>>(64-2)
+		t = a[2] ^ d2
+		bc0 = t<<62 | t>>(64-62)
+		t = a[8] ^ d3
+		bc1 = t<<55 | t>>(64-55)
+		t = a[14] ^ d4
+		bc2 = t<<39 | t>>(64-39)
+		a[15] = bc0 ^ (bc2 &^ bc1)
+		a[21] = bc1 ^ (bc3 &^ bc2)
+		a[2] = bc2 ^ (bc4 &^ bc3)
+		a[8] = bc3 ^ (bc0 &^ bc4)
+		a[14] = bc4 ^ (bc1 &^ bc0)
+
+		// Round 2
+		bc0 = a[0] ^ a[5] ^ a[10] ^ a[15] ^ a[20]
+		bc1 = a[1] ^ a[6] ^ a[11] ^ a[16] ^ a[21]
+		bc2 = a[2] ^ a[7] ^ a[12] ^ a[17] ^ a[22]
+		bc3 = a[3] ^ a[8] ^ a[13] ^ a[18] ^ a[23]
+		bc4 = a[4] ^ a[9] ^ a[14] ^ a[19] ^ a[24]
+		d0 = bc4 ^ (bc1<<1 | bc1>>63)
+		d1 = bc0 ^ (bc2<<1 | bc2>>63)
+		d2 = bc1 ^ (bc3<<1 | bc3>>63)
+		d3 = bc2 ^ (bc4<<1 | bc4>>63)
+		d4 = bc3 ^ (bc0<<1 | bc0>>63)
+
+		bc0 = a[0] ^ d0
+		t = a[16] ^ d1
+		bc1 = t<<44 | t>>(64-44)
+		t = a[7] ^ d2
+		bc2 = t<<43 | t>>(64-43)
+		t = a[23] ^ d3
+		bc3 = t<<21 | t>>(64-21)
+		t = a[14] ^ d4
+		bc4 = t<<14 | t>>(64-14)
+		a[0] = bc0 ^ (bc2 &^ bc1) ^ rc[i+1]
+		a[16] = bc1 ^ (bc3 &^ bc2)
+		a[7] = bc2 ^ (bc4 &^ bc3)
+		a[23] = bc3 ^ (bc0 &^ bc4)
+		a[14] = bc4 ^ (bc1 &^ bc0)
+
+		t = a[20] ^ d0
+		bc2 = t<<3 | t>>(64-3)
+		t = a[11] ^ d1
+		bc3 = t<<45 | t>>(64-45)
+		t = a[2] ^ d2
+		bc4 = t<<61 | t>>(64-61)
+		t = a[18] ^ d3
+		bc0 = t<<28 | t>>(64-28)
+		t = a[9] ^ d4
+		bc1 = t<<20 | t>>(64-20)
+		a[20] = bc0 ^ (bc2 &^ bc1)
+		a[11] = bc1 ^ (bc3 &^ bc2)
+		a[2] = bc2 ^ (bc4 &^ bc3)
+		a[18] = bc3 ^ (bc0 &^ bc4)
+		a[9] = bc4 ^ (bc1 &^ bc0)
+
+		t = a[15] ^ d0
+		bc4 = t<<18 | t>>(64-18)
+		t = a[6] ^ d1
+		bc0 = t<<1 | t>>(64-1)
+		t = a[22] ^ d2
+		bc1 = t<<6 | t>>(64-6)
+		t = a[13] ^ d3
+		bc2 = t<<25 | t>>(64-25)
+		t = a[4] ^ d4
+		bc3 = t<<8 | t>>(64-8)
+		a[15] = bc0 ^ (bc2 &^ bc1)
+		a[6] = bc1 ^ (bc3 &^ bc2)
+		a[22] = bc2 ^ (bc4 &^ bc3)
+		a[13] = bc3 ^ (bc0 &^ bc4)
+		a[4] = bc4 ^ (bc1 &^ bc0)
+
+		t = a[10] ^ d0
+		bc1 = t<<36 | t>>(64-36)
+		t = a[1] ^ d1
+		bc2 = t<<10 | t>>(64-10)
+		t = a[17] ^ d2
+		bc3 = t<<15 | t>>(64-15)
+		t = a[8] ^ d3
+		bc4 = t<<56 | t>>(64-56)
+		t = a[24] ^ d4
+		bc0 = t<<27 | t>>(64-27)
+		a[10] = bc0 ^ (bc2 &^ bc1)
+		a[1] = bc1 ^ (bc3 &^ bc2)
+		a[17] = bc2 ^ (bc4 &^ bc3)
+		a[8] = bc3 ^ (bc0 &^ bc4)
+		a[24] = bc4 ^ (bc1 &^ bc0)
+
+		t = a[5] ^ d0
+		bc3 = t<<41 | t>>(64-41)
+		t = a[21] ^ d1
+		bc4 = t<<2 | t>>(64-2)
+		t = a[12] ^ d2
+		bc0 = t<<62 | t>>(64-62)
+		t = a[3] ^ d3
+		bc1 = t<<55 | t>>(64-55)
+		t = a[19] ^ d4
+		bc2 = t<<39 | t>>(64-39)
+		a[5] = bc0 ^ (bc2 &^ bc1)
+		a[21] = bc1 ^ (bc3 &^ bc2)
+		a[12] = bc2 ^ (bc4 &^ bc3)
+		a[3] = bc3 ^ (bc0 &^ bc4)
+		a[19] = bc4 ^ (bc1 &^ bc0)
+
+		// Round 3
+		bc0 = a[0] ^ a[5] ^ a[10] ^ a[15] ^ a[20]
+		bc1 = a[1] ^ a[6] ^ a[11] ^ a[16] ^ a[21]
+		bc2 = a[2] ^ a[7] ^ a[12] ^ a[17] ^ a[22]
+		bc3 = a[3] ^ a[8] ^ a[13] ^ a[18] ^ a[23]
+		bc4 = a[4] ^ a[9] ^ a[14] ^ a[19] ^ a[24]
+		d0 = bc4 ^ (bc1<<1 | bc1>>63)
+		d1 = bc0 ^ (bc2<<1 | bc2>>63)
+		d2 = bc1 ^ (bc3<<1 | bc3>>63)
+		d3 = bc2 ^ (bc4<<1 | bc4>>63)
+		d4 = bc3 ^ (bc0<<1 | bc0>>63)
+
+		bc0 = a[0] ^ d0
+		t = a[11] ^ d1
+		bc1 = t<<44 | t>>(64-44)
+		t = a[22] ^ d2
+		bc2 = t<<43 | t>>(64-43)
+		t = a[8] ^ d3
+		bc3 = t<<21 | t>>(64-21)
+		t = a[19] ^ d4
+		bc4 = t<<14 | t>>(64-14)
+		a[0] = bc0 ^ (bc2 &^ bc1) ^ rc[i+2]
+		a[11] = bc1 ^ (bc3 &^ bc2)
+		a[22] = bc2 ^ (bc4 &^ bc3)
+		a[8] = bc3 ^ (bc0 &^ bc4)
+		a[19] = bc4 ^ (bc1 &^ bc0)
+
+		t = a[15] ^ d0
+		bc2 = t<<3 | t>>(64-3)
+		t = a[1] ^ d1
+		bc3 = t<<45 | t>>(64-45)
+		t = a[12] ^ d2
+		bc4 = t<<61 | t>>(64-61)
+		t = a[23] ^ d3
+		bc0 = t<<28 | t>>(64-28)
+		t = a[9] ^ d4
+		bc1 = t<<20 | t>>(64-20)
+		a[15] = bc0 ^ (bc2 &^ bc1)
+		a[1] = bc1 ^ (bc3 &^ bc2)
+		a[12] = bc2 ^ (bc4 &^ bc3)
+		a[23] = bc3 ^ (bc0 &^ bc4)
+		a[9] = bc4 ^ (bc1 &^ bc0)
+
+		t = a[5] ^ d0
+		bc4 = t<<18 | t>>(64-18)
+		t = a[16] ^ d1
+		bc0 = t<<1 | t>>(64-1)
+		t = a[2] ^ d2
+		bc1 = t<<6 | t>>(64-6)
+		t = a[13] ^ d3
+		bc2 = t<<25 | t>>(64-25)
+		t = a[24] ^ d4
+		bc3 = t<<8 | t>>(64-8)
+		a[5] = bc0 ^ (bc2 &^ bc1)
+		a[16] = bc1 ^ (bc3 &^ bc2)
+		a[2] = bc2 ^ (bc4 &^ bc3)
+		a[13] = bc3 ^ (bc0 &^ bc4)
+		a[24] = bc4 ^ (bc1 &^ bc0)
+
+		t = a[20] ^ d0
+		bc1 = t<<36 | t>>(64-36)
+		t = a[6] ^ d1
+		bc2 = t<<10 | t>>(64-10)
+		t = a[17] ^ d2
+		bc3 = t<<15 | t>>(64-15)
+		t = a[3] ^ d3
+		bc4 = t<<56 | t>>(64-56)
+		t = a[14] ^ d4
+		bc0 = t<<27 | t>>(64-27)
+		a[20] = bc0 ^ (bc2 &^ bc1)
+		a[6] = bc1 ^ (bc3 &^ bc2)
+		a[17] = bc2 ^ (bc4 &^ bc3)
+		a[3] = bc3 ^ (bc0 &^ bc4)
+		a[14] = bc4 ^ (bc1 &^ bc0)
+
+		t = a[10] ^ d0
+		bc3 = t<<41 | t>>(64-41)
+		t = a[21] ^ d1
+		bc4 = t<<2 | t>>(64-2)
+		t = a[7] ^ d2
+		bc0 = t<<62 | t>>(64-62)
+		t = a[18] ^ d3
+		bc1 = t<<55 | t>>(64-55)
+		t = a[4] ^ d4
+		bc2 = t<<39 | t>>(64-39)
+		a[10] = bc0 ^ (bc2 &^ bc1)
+		a[21] = bc1 ^ (bc3 &^ bc2)
+		a[7] = bc2 ^ (bc4 &^ bc3)
+		a[18] = bc3 ^ (bc0 &^ bc4)
+		a[4] = bc4 ^ (bc1 &^ bc0)
+
+		// Round 4
+		bc0 = a[0] ^ a[5] ^ a[10] ^ a[15] ^ a[20]
+		bc1 = a[1] ^ a[6] ^ a[11] ^ a[16] ^ a[21]
+		bc2 = a[2] ^ a[7] ^ a[12] ^ a[17] ^ a[22]
+		bc3 = a[3] ^ a[8] ^ a[13] ^ a[18] ^ a[23]
+		bc4 = a[4] ^ a[9] ^ a[14] ^ a[19] ^ a[24]
+		d0 = bc4 ^ (bc1<<1 | bc1>>63)
+		d1 = bc0 ^ (bc2<<1 | bc2>>63)
+		d2 = bc1 ^ (bc3<<1 | bc3>>63)
+		d3 = bc2 ^ (bc4<<1 | bc4>>63)
+		d4 = bc3 ^ (bc0<<1 | bc0>>63)
+
+		bc0 = a[0] ^ d0
+		t = a[1] ^ d1
+		bc1 = t<<44 | t>>(64-44)
+		t = a[2] ^ d2
+		bc2 = t<<43 | t>>(64-43)
+		t = a[3] ^ d3
+		bc3 = t<<21 | t>>(64-21)
+		t = a[4] ^ d4
+		bc4 = t<<14 | t>>(64-14)
+		a[0] = bc0 ^ (bc2 &^ bc1) ^ rc[i+3]
+		a[1] = bc1 ^ (bc3 &^ bc2)
+		a[2] = bc2 ^ (bc4 &^ bc3)
+		a[3] = bc3 ^ (bc0 &^ bc4)
+		a[4] = bc4 ^ (bc1 &^ bc0)
+
+		t = a[5] ^ d0
+		bc2 = t<<3 | t>>(64-3)
+		t = a[6] ^ d1
+		bc3 = t<<45 | t>>(64-45)
+		t = a[7] ^ d2
+		bc4 = t<<61 | t>>(64-61)
+		t = a[8] ^ d3
+		bc0 = t<<28 | t>>(64-28)
+		t = a[9] ^ d4
+		bc1 = t<<20 | t>>(64-20)
+		a[5] = bc0 ^ (bc2 &^ bc1)
+		a[6] = bc1 ^ (bc3 &^ bc2)
+		a[7] = bc2 ^ (bc4 &^ bc3)
+		a[8] = bc3 ^ (bc0 &^ bc4)
+		a[9] = bc4 ^ (bc1 &^ bc0)
+
+		t = a[10] ^ d0
+		bc4 = t<<18 | t>>(64-18)
+		t = a[11] ^ d1
+		bc0 = t<<1 | t>>(64-1)
+		t = a[12] ^ d2
+		bc1 = t<<6 | t>>(64-6)
+		t = a[13] ^ d3
+		bc2 = t<<25 | t>>(64-25)
+		t = a[14] ^ d4
+		bc3 = t<<8 | t>>(64-8)
+		a[10] = bc0 ^ (bc2 &^ bc1)
+		a[11] = bc1 ^ (bc3 &^ bc2)
+		a[12] = bc2 ^ (bc4 &^ bc3)
+		a[13] = bc3 ^ (bc0 &^ bc4)
+		a[14] = bc4 ^ (bc1 &^ bc0)
+
+		t = a[15] ^ d0
+		bc1 = t<<36 | t>>(64-36)
+		t = a[16] ^ d1
+		bc2 = t<<10 | t>>(64-10)
+		t = a[17] ^ d2
+		bc3 = t<<15 | t>>(64-15)
+		t = a[18] ^ d3
+		bc4 = t<<56 | t>>(64-56)
+		t = a[19] ^ d4
+		bc0 = t<<27 | t>>(64-27)
+		a[15] = bc0 ^ (bc2 &^ bc1)
+		a[16] = bc1 ^ (bc3 &^ bc2)
+		a[17] = bc2 ^ (bc4 &^ bc3)
+		a[18] = bc3 ^ (bc0 &^ bc4)
+		a[19] = bc4 ^ (bc1 &^ bc0)
+
+		t = a[20] ^ d0
+		bc3 = t<<41 | t>>(64-41)
+		t = a[21] ^ d1
+		bc4 = t<<2 | t>>(64-2)
+		t = a[22] ^ d2
+		bc0 = t<<62 | t>>(64-62)
+		t = a[23] ^ d3
+		bc1 = t<<55 | t>>(64-55)
+		t = a[24] ^ d4
+		bc2 = t<<39 | t>>(64-39)
+		a[20] = bc0 ^ (bc2 &^ bc1)
+		a[21] = bc1 ^ (bc3 &^ bc2)
+		a[22] = bc2 ^ (bc4 &^ bc3)
+		a[23] = bc3 ^ (bc0 &^ bc4)
+		a[24] = bc4 ^ (bc1 &^ bc0)
+	}
+}
diff --git a/crypto/sha3/register.go b/crypto/sha3/register.go
new file mode 100644
index 000000000..3cf6a22e0
--- /dev/null
+++ b/crypto/sha3/register.go
@@ -0,0 +1,18 @@
+// Copyright 2014 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+// +build go1.4
+
+package sha3
+
+import (
+	"crypto"
+)
+
+func init() {
+	crypto.RegisterHash(crypto.SHA3_224, New224)
+	crypto.RegisterHash(crypto.SHA3_256, New256)
+	crypto.RegisterHash(crypto.SHA3_384, New384)
+	crypto.RegisterHash(crypto.SHA3_512, New512)
+}
diff --git a/crypto/sha3/sha3.go b/crypto/sha3/sha3.go
new file mode 100644
index 000000000..c8fd31cb0
--- /dev/null
+++ b/crypto/sha3/sha3.go
@@ -0,0 +1,193 @@
+// Copyright 2014 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package sha3
+
+// spongeDirection indicates the direction bytes are flowing through the sponge.
+type spongeDirection int
+
+const (
+	// spongeAbsorbing indicates that the sponge is absorbing input.
+	spongeAbsorbing spongeDirection = iota
+	// spongeSqueezing indicates that the sponge is being squeezed.
+	spongeSqueezing
+)
+
+const (
+	// maxRate is the maximum size of the internal buffer. SHAKE-256
+	// currently needs the largest buffer.
+	maxRate = 168
+)
+
+type state struct {
+	// Generic sponge components.
+	a    [25]uint64 // main state of the hash
+	buf  []byte     // points into storage
+	rate int        // the number of bytes of state to use
+
+	// dsbyte contains the "domain separation" bits and the first bit of
+	// the padding. Sections 6.1 and 6.2 of [1] separate the outputs of the
+	// SHA-3 and SHAKE functions by appending bitstrings to the message.
+	// Using a little-endian bit-ordering convention, these are "01" for SHA-3
+	// and "1111" for SHAKE, or 00000010b and 00001111b, respectively. Then the
+	// padding rule from section 5.1 is applied to pad the message to a multiple
+	// of the rate, which involves adding a "1" bit, zero or more "0" bits, and
+	// a final "1" bit. We merge the first "1" bit from the padding into dsbyte,
+	// giving 00000110b (0x06) and 00011111b (0x1f).
+	// [1] http://csrc.nist.gov/publications/drafts/fips-202/fips_202_draft.pdf
+	//     "Draft FIPS 202: SHA-3 Standard: Permutation-Based Hash and
+	//      Extendable-Output Functions (May 2014)"
+	dsbyte  byte
+	storage [maxRate]byte
+
+	// Specific to SHA-3 and SHAKE.
+	fixedOutput bool            // whether this is a fixed-ouput-length instance
+	outputLen   int             // the default output size in bytes
+	state       spongeDirection // whether the sponge is absorbing or squeezing
+}
+
+// BlockSize returns the rate of sponge underlying this hash function.
+func (d *state) BlockSize() int { return d.rate }
+
+// Size returns the output size of the hash function in bytes.
+func (d *state) Size() int { return d.outputLen }
+
+// Reset clears the internal state by zeroing the sponge state and
+// the byte buffer, and setting Sponge.state to absorbing.
+func (d *state) Reset() {
+	// Zero the permutation's state.
+	for i := range d.a {
+		d.a[i] = 0
+	}
+	d.state = spongeAbsorbing
+	d.buf = d.storage[:0]
+}
+
+func (d *state) clone() *state {
+	ret := *d
+	if ret.state == spongeAbsorbing {
+		ret.buf = ret.storage[:len(ret.buf)]
+	} else {
+		ret.buf = ret.storage[d.rate-cap(d.buf) : d.rate]
+	}
+
+	return &ret
+}
+
+// permute applies the KeccakF-1600 permutation. It handles
+// any input-output buffering.
+func (d *state) permute() {
+	switch d.state {
+	case spongeAbsorbing:
+		// If we're absorbing, we need to xor the input into the state
+		// before applying the permutation.
+		xorIn(d, d.buf)
+		d.buf = d.storage[:0]
+		keccakF1600(&d.a)
+	case spongeSqueezing:
+		// If we're squeezing, we need to apply the permutatin before
+		// copying more output.
+		keccakF1600(&d.a)
+		d.buf = d.storage[:d.rate]
+		copyOut(d, d.buf)
+	}
+}
+
+// pads appends the domain separation bits in dsbyte, applies
+// the multi-bitrate 10..1 padding rule, and permutes the state.
+func (d *state) padAndPermute(dsbyte byte) {
+	if d.buf == nil {
+		d.buf = d.storage[:0]
+	}
+	// Pad with this instance's domain-separator bits. We know that there's
+	// at least one byte of space in d.buf because, if it were full,
+	// permute would have been called to empty it. dsbyte also contains the
+	// first one bit for the padding. See the comment in the state struct.
+	d.buf = append(d.buf, dsbyte)
+	zerosStart := len(d.buf)
+	d.buf = d.storage[:d.rate]
+	for i := zerosStart; i < d.rate; i++ {
+		d.buf[i] = 0
+	}
+	// This adds the final one bit for the padding. Because of the way that
+	// bits are numbered from the LSB upwards, the final bit is the MSB of
+	// the last byte.
+	d.buf[d.rate-1] ^= 0x80
+	// Apply the permutation
+	d.permute()
+	d.state = spongeSqueezing
+	d.buf = d.storage[:d.rate]
+	copyOut(d, d.buf)
+}
+
+// Write absorbs more data into the hash's state. It produces an error
+// if more data is written to the ShakeHash after writing
+func (d *state) Write(p []byte) (written int, err error) {
+	if d.state != spongeAbsorbing {
+		panic("sha3: write to sponge after read")
+	}
+	if d.buf == nil {
+		d.buf = d.storage[:0]
+	}
+	written = len(p)
+
+	for len(p) > 0 {
+		if len(d.buf) == 0 && len(p) >= d.rate {
+			// The fast path; absorb a full "rate" bytes of input and apply the permutation.
+			xorIn(d, p[:d.rate])
+			p = p[d.rate:]
+			keccakF1600(&d.a)
+		} else {
+			// The slow path; buffer the input until we can fill the sponge, and then xor it in.
+			todo := d.rate - len(d.buf)
+			if todo > len(p) {
+				todo = len(p)
+			}
+			d.buf = append(d.buf, p[:todo]...)
+			p = p[todo:]
+
+			// If the sponge is full, apply the permutation.
+			if len(d.buf) == d.rate {
+				d.permute()
+			}
+		}
+	}
+
+	return
+}
+
+// Read squeezes an arbitrary number of bytes from the sponge.
+func (d *state) Read(out []byte) (n int, err error) {
+	// If we're still absorbing, pad and apply the permutation.
+	if d.state == spongeAbsorbing {
+		d.padAndPermute(d.dsbyte)
+	}
+
+	n = len(out)
+
+	// Now, do the squeezing.
+	for len(out) > 0 {
+		n := copy(out, d.buf)
+		d.buf = d.buf[n:]
+		out = out[n:]
+
+		// Apply the permutation if we've squeezed the sponge dry.
+		if len(d.buf) == 0 {
+			d.permute()
+		}
+	}
+
+	return
+}
+
+// Sum applies padding to the hash state and then squeezes out the desired
+// number of output bytes.
+func (d *state) Sum(in []byte) []byte {
+	// Make a copy of the original hash so that caller can keep writing
+	// and summing.
+	dup := d.clone()
+	hash := make([]byte, dup.outputLen)
+	dup.Read(hash)
+	return append(in, hash...)
+}
diff --git a/crypto/sha3/sha3_test.go b/crypto/sha3/sha3_test.go
new file mode 100644
index 000000000..caf72f279
--- /dev/null
+++ b/crypto/sha3/sha3_test.go
@@ -0,0 +1,306 @@
+// Copyright 2014 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package sha3
+
+// Tests include all the ShortMsgKATs provided by the Keccak team at
+// https://github.com/gvanas/KeccakCodePackage
+//
+// They only include the zero-bit case of the bitwise testvectors
+// published by NIST in the draft of FIPS-202.
+
+import (
+	"bytes"
+	"compress/flate"
+	"encoding/hex"
+	"encoding/json"
+	"hash"
+	"os"
+	"strings"
+	"testing"
+)
+
+const (
+	testString  = "brekeccakkeccak koax koax"
+	katFilename = "testdata/keccakKats.json.deflate"
+)
+
+// Internal-use instances of SHAKE used to test against KATs.
+func newHashShake128() hash.Hash {
+	return &state{rate: 168, dsbyte: 0x1f, outputLen: 512}
+}
+func newHashShake256() hash.Hash {
+	return &state{rate: 136, dsbyte: 0x1f, outputLen: 512}
+}
+
+// testDigests contains functions returning hash.Hash instances
+// with output-length equal to the KAT length for both SHA-3 and
+// SHAKE instances.
+var testDigests = map[string]func() hash.Hash{
+	"SHA3-224": New224,
+	"SHA3-256": New256,
+	"SHA3-384": New384,
+	"SHA3-512": New512,
+	"SHAKE128": newHashShake128,
+	"SHAKE256": newHashShake256,
+}
+
+// testShakes contains functions that return ShakeHash instances for
+// testing the ShakeHash-specific interface.
+var testShakes = map[string]func() ShakeHash{
+	"SHAKE128": NewShake128,
+	"SHAKE256": NewShake256,
+}
+
+// decodeHex converts a hex-encoded string into a raw byte string.
+func decodeHex(s string) []byte {
+	b, err := hex.DecodeString(s)
+	if err != nil {
+		panic(err)
+	}
+	return b
+}
+
+// structs used to marshal JSON test-cases.
+type KeccakKats struct {
+	Kats map[string][]struct {
+		Digest  string `json:"digest"`
+		Length  int64  `json:"length"`
+		Message string `json:"message"`
+	}
+}
+
+func testUnalignedAndGeneric(t *testing.T, testf func(impl string)) {
+	xorInOrig, copyOutOrig := xorIn, copyOut
+	xorIn, copyOut = xorInGeneric, copyOutGeneric
+	testf("generic")
+	if xorImplementationUnaligned != "generic" {
+		xorIn, copyOut = xorInUnaligned, copyOutUnaligned
+		testf("unaligned")
+	}
+	xorIn, copyOut = xorInOrig, copyOutOrig
+}
+
+// TestKeccakKats tests the SHA-3 and Shake implementations against all the
+// ShortMsgKATs from https://github.com/gvanas/KeccakCodePackage
+// (The testvectors are stored in keccakKats.json.deflate due to their length.)
+func TestKeccakKats(t *testing.T) {
+	testUnalignedAndGeneric(t, func(impl string) {
+		// Read the KATs.
+		deflated, err := os.Open(katFilename)
+		if err != nil {
+			t.Errorf("error opening %s: %s", katFilename, err)
+		}
+		file := flate.NewReader(deflated)
+		dec := json.NewDecoder(file)
+		var katSet KeccakKats
+		err = dec.Decode(&katSet)
+		if err != nil {
+			t.Errorf("error decoding KATs: %s", err)
+		}
+
+		// Do the KATs.
+		for functionName, kats := range katSet.Kats {
+			d := testDigests[functionName]()
+			for _, kat := range kats {
+				d.Reset()
+				in, err := hex.DecodeString(kat.Message)
+				if err != nil {
+					t.Errorf("error decoding KAT: %s", err)
+				}
+				d.Write(in[:kat.Length/8])
+				got := strings.ToUpper(hex.EncodeToString(d.Sum(nil)))
+				if got != kat.Digest {
+					t.Errorf("function=%s, implementation=%s, length=%d\nmessage:\n  %s\ngot:\n  %s\nwanted:\n %s",
+						functionName, impl, kat.Length, kat.Message, got, kat.Digest)
+					t.Logf("wanted %+v", kat)
+					t.FailNow()
+				}
+				continue
+			}
+		}
+	})
+}
+
+// TestUnalignedWrite tests that writing data in an arbitrary pattern with
+// small input buffers.
+func testUnalignedWrite(t *testing.T) {
+	testUnalignedAndGeneric(t, func(impl string) {
+		buf := sequentialBytes(0x10000)
+		for alg, df := range testDigests {
+			d := df()
+			d.Reset()
+			d.Write(buf)
+			want := d.Sum(nil)
+			d.Reset()
+			for i := 0; i < len(buf); {
+				// Cycle through offsets which make a 137 byte sequence.
+				// Because 137 is prime this sequence should exercise all corner cases.
+				offsets := [17]int{1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 1}
+				for _, j := range offsets {
+					if v := len(buf) - i; v < j {
+						j = v
+					}
+					d.Write(buf[i : i+j])
+					i += j
+				}
+			}
+			got := d.Sum(nil)
+			if !bytes.Equal(got, want) {
+				t.Errorf("Unaligned writes, implementation=%s, alg=%s\ngot %q, want %q", impl, alg, got, want)
+			}
+		}
+	})
+}
+
+// TestAppend checks that appending works when reallocation is necessary.
+func TestAppend(t *testing.T) {
+	testUnalignedAndGeneric(t, func(impl string) {
+		d := New224()
+
+		for capacity := 2; capacity <= 66; capacity += 64 {
+			// The first time around the loop, Sum will have to reallocate.
+			// The second time, it will not.
+			buf := make([]byte, 2, capacity)
+			d.Reset()
+			d.Write([]byte{0xcc})
+			buf = d.Sum(buf)
+			expected := "0000DF70ADC49B2E76EEE3A6931B93FA41841C3AF2CDF5B32A18B5478C39"
+			if got := strings.ToUpper(hex.EncodeToString(buf)); got != expected {
+				t.Errorf("got %s, want %s", got, expected)
+			}
+		}
+	})
+}
+
+// TestAppendNoRealloc tests that appending works when no reallocation is necessary.
+func TestAppendNoRealloc(t *testing.T) {
+	testUnalignedAndGeneric(t, func(impl string) {
+		buf := make([]byte, 1, 200)
+		d := New224()
+		d.Write([]byte{0xcc})
+		buf = d.Sum(buf)
+		expected := "00DF70ADC49B2E76EEE3A6931B93FA41841C3AF2CDF5B32A18B5478C39"
+		if got := strings.ToUpper(hex.EncodeToString(buf)); got != expected {
+			t.Errorf("%s: got %s, want %s", impl, got, expected)
+		}
+	})
+}
+
+// TestSqueezing checks that squeezing the full output a single time produces
+// the same output as repeatedly squeezing the instance.
+func TestSqueezing(t *testing.T) {
+	testUnalignedAndGeneric(t, func(impl string) {
+		for functionName, newShakeHash := range testShakes {
+			d0 := newShakeHash()
+			d0.Write([]byte(testString))
+			ref := make([]byte, 32)
+			d0.Read(ref)
+
+			d1 := newShakeHash()
+			d1.Write([]byte(testString))
+			var multiple []byte
+			for _ = range ref {
+				one := make([]byte, 1)
+				d1.Read(one)
+				multiple = append(multiple, one...)
+			}
+			if !bytes.Equal(ref, multiple) {
+				t.Errorf("%s (%s): squeezing %d bytes one at a time failed", functionName, impl, len(ref))
+			}
+		}
+	})
+}
+
+// sequentialBytes produces a buffer of size consecutive bytes 0x00, 0x01, ..., used for testing.
+func sequentialBytes(size int) []byte {
+	result := make([]byte, size)
+	for i := range result {
+		result[i] = byte(i)
+	}
+	return result
+}
+
+// BenchmarkPermutationFunction measures the speed of the permutation function
+// with no input data.
+func BenchmarkPermutationFunction(b *testing.B) {
+	b.SetBytes(int64(200))
+	var lanes [25]uint64
+	for i := 0; i < b.N; i++ {
+		keccakF1600(&lanes)
+	}
+}
+
+// benchmarkHash tests the speed to hash num buffers of buflen each.
+func benchmarkHash(b *testing.B, h hash.Hash, size, num int) {
+	b.StopTimer()
+	h.Reset()
+	data := sequentialBytes(size)
+	b.SetBytes(int64(size * num))
+	b.StartTimer()
+
+	var state []byte
+	for i := 0; i < b.N; i++ {
+		for j := 0; j < num; j++ {
+			h.Write(data)
+		}
+		state = h.Sum(state[:0])
+	}
+	b.StopTimer()
+	h.Reset()
+}
+
+// benchmarkShake is specialized to the Shake instances, which don't
+// require a copy on reading output.
+func benchmarkShake(b *testing.B, h ShakeHash, size, num int) {
+	b.StopTimer()
+	h.Reset()
+	data := sequentialBytes(size)
+	d := make([]byte, 32)
+
+	b.SetBytes(int64(size * num))
+	b.StartTimer()
+
+	for i := 0; i < b.N; i++ {
+		h.Reset()
+		for j := 0; j < num; j++ {
+			h.Write(data)
+		}
+		h.Read(d)
+	}
+}
+
+func BenchmarkSha3_512_MTU(b *testing.B) { benchmarkHash(b, New512(), 1350, 1) }
+func BenchmarkSha3_384_MTU(b *testing.B) { benchmarkHash(b, New384(), 1350, 1) }
+func BenchmarkSha3_256_MTU(b *testing.B) { benchmarkHash(b, New256(), 1350, 1) }
+func BenchmarkSha3_224_MTU(b *testing.B) { benchmarkHash(b, New224(), 1350, 1) }
+
+func BenchmarkShake128_MTU(b *testing.B)  { benchmarkShake(b, NewShake128(), 1350, 1) }
+func BenchmarkShake256_MTU(b *testing.B)  { benchmarkShake(b, NewShake256(), 1350, 1) }
+func BenchmarkShake256_16x(b *testing.B)  { benchmarkShake(b, NewShake256(), 16, 1024) }
+func BenchmarkShake256_1MiB(b *testing.B) { benchmarkShake(b, NewShake256(), 1024, 1024) }
+
+func BenchmarkSha3_512_1MiB(b *testing.B) { benchmarkHash(b, New512(), 1024, 1024) }
+
+func Example_sum() {
+	buf := []byte("some data to hash")
+	// A hash needs to be 64 bytes long to have 256-bit collision resistance.
+	h := make([]byte, 64)
+	// Compute a 64-byte hash of buf and put it in h.
+	ShakeSum256(h, buf)
+}
+
+func Example_mac() {
+	k := []byte("this is a secret key; you should generate a strong random key that's at least 32 bytes long")
+	buf := []byte("and this is some data to authenticate")
+	// A MAC with 32 bytes of output has 256-bit security strength -- if you use at least a 32-byte-long key.
+	h := make([]byte, 32)
+	d := NewShake256()
+	// Write the key into the hash.
+	d.Write(k)
+	// Now write the data.
+	d.Write(buf)
+	// Read 32 bytes of output from the hash into h.
+	d.Read(h)
+}
diff --git a/crypto/sha3/shake.go b/crypto/sha3/shake.go
new file mode 100644
index 000000000..841f9860f
--- /dev/null
+++ b/crypto/sha3/shake.go
@@ -0,0 +1,60 @@
+// Copyright 2014 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package sha3
+
+// This file defines the ShakeHash interface, and provides
+// functions for creating SHAKE instances, as well as utility
+// functions for hashing bytes to arbitrary-length output.
+
+import (
+	"io"
+)
+
+// ShakeHash defines the interface to hash functions that
+// support arbitrary-length output.
+type ShakeHash interface {
+	// Write absorbs more data into the hash's state. It panics if input is
+	// written to it after output has been read from it.
+	io.Writer
+
+	// Read reads more output from the hash; reading affects the hash's
+	// state. (ShakeHash.Read is thus very different from Hash.Sum)
+	// It never returns an error.
+	io.Reader
+
+	// Clone returns a copy of the ShakeHash in its current state.
+	Clone() ShakeHash
+
+	// Reset resets the ShakeHash to its initial state.
+	Reset()
+}
+
+func (d *state) Clone() ShakeHash {
+	return d.clone()
+}
+
+// NewShake128 creates a new SHAKE128 variable-output-length ShakeHash.
+// Its generic security strength is 128 bits against all attacks if at
+// least 32 bytes of its output are used.
+func NewShake128() ShakeHash { return &state{rate: 168, dsbyte: 0x1f} }
+
+// NewShake256 creates a new SHAKE128 variable-output-length ShakeHash.
+// Its generic security strength is 256 bits against all attacks if
+// at least 64 bytes of its output are used.
+func NewShake256() ShakeHash { return &state{rate: 136, dsbyte: 0x1f} }
+
+// ShakeSum128 writes an arbitrary-length digest of data into hash.
+func ShakeSum128(hash, data []byte) {
+	h := NewShake128()
+	h.Write(data)
+	h.Read(hash)
+}
+
+// ShakeSum256 writes an arbitrary-length digest of data into hash.
+func ShakeSum256(hash, data []byte) {
+	h := NewShake256()
+	h.Write(data)
+	h.Read(hash)
+}
diff --git a/crypto/sha3/testdata/keccakKats.json.deflate b/crypto/sha3/testdata/keccakKats.json.deflate
new file mode 100644
index 000000000..62e85ae24
--- /dev/null
+++ b/crypto/sha3/testdata/keccakKats.json.deflate
diff --git a/crypto/sha3/xor.go b/crypto/sha3/xor.go
new file mode 100644
index 000000000..d622979c1
--- /dev/null
+++ b/crypto/sha3/xor.go
@@ -0,0 +1,16 @@
+// Copyright 2015 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+// +build !amd64,!386 appengine
+
+package sha3
+
+var (
+	xorIn            = xorInGeneric
+	copyOut          = copyOutGeneric
+	xorInUnaligned   = xorInGeneric
+	copyOutUnaligned = copyOutGeneric
+)
+
+const xorImplementationUnaligned = "generic"
diff --git a/crypto/sha3/xor_generic.go b/crypto/sha3/xor_generic.go
new file mode 100644
index 000000000..fd35f02ef
--- /dev/null
+++ b/crypto/sha3/xor_generic.go
@@ -0,0 +1,28 @@
+// Copyright 2015 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package sha3
+
+import "encoding/binary"
+
+// xorInGeneric xors the bytes in buf into the state; it
+// makes no non-portable assumptions about memory layout
+// or alignment.
+func xorInGeneric(d *state, buf []byte) {
+	n := len(buf) / 8
+
+	for i := 0; i < n; i++ {
+		a := binary.LittleEndian.Uint64(buf)
+		d.a[i] ^= a
+		buf = buf[8:]
+	}
+}
+
+// copyOutGeneric copies ulint64s to a byte buffer.
+func copyOutGeneric(d *state, b []byte) {
+	for i := 0; len(b) >= 8; i++ {
+		binary.LittleEndian.PutUint64(b, d.a[i])
+		b = b[8:]
+	}
+}
diff --git a/crypto/sha3/xor_unaligned.go b/crypto/sha3/xor_unaligned.go
new file mode 100644
index 000000000..c7851a1d8
--- /dev/null
+++ b/crypto/sha3/xor_unaligned.go
@@ -0,0 +1,58 @@
+// Copyright 2015 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+// +build amd64 386
+// +build !appengine
+
+package sha3
+
+import "unsafe"
+
+func xorInUnaligned(d *state, buf []byte) {
+	bw := (*[maxRate / 8]uint64)(unsafe.Pointer(&buf[0]))
+	n := len(buf)
+	if n >= 72 {
+		d.a[0] ^= bw[0]
+		d.a[1] ^= bw[1]
+		d.a[2] ^= bw[2]
+		d.a[3] ^= bw[3]
+		d.a[4] ^= bw[4]
+		d.a[5] ^= bw[5]
+		d.a[6] ^= bw[6]
+		d.a[7] ^= bw[7]
+		d.a[8] ^= bw[8]
+	}
+	if n >= 104 {
+		d.a[9] ^= bw[9]
+		d.a[10] ^= bw[10]
+		d.a[11] ^= bw[11]
+		d.a[12] ^= bw[12]
+	}
+	if n >= 136 {
+		d.a[13] ^= bw[13]
+		d.a[14] ^= bw[14]
+		d.a[15] ^= bw[15]
+		d.a[16] ^= bw[16]
+	}
+	if n >= 144 {
+		d.a[17] ^= bw[17]
+	}
+	if n >= 168 {
+		d.a[18] ^= bw[18]
+		d.a[19] ^= bw[19]
+		d.a[20] ^= bw[20]
+	}
+}
+
+func copyOutUnaligned(d *state, buf []byte) {
+	ab := (*[maxRate]uint8)(unsafe.Pointer(&d.a[0]))
+	copy(buf, ab[:])
+}
+
+var (
+	xorIn   = xorInUnaligned
+	copyOut = copyOutUnaligned
+)
+
+const xorImplementationUnaligned = "unaligned"