001/* 002 * Licensed to the Apache Software Foundation (ASF) under one or more 003 * contributor license agreements. See the NOTICE file distributed with 004 * this work for additional information regarding copyright ownership. 005 * The ASF licenses this file to You under the Apache License, Version 2.0 006 * (the "License"); you may not use this file except in compliance with 007 * the License. You may obtain a copy of the License at 008 * 009 * http://www.apache.org/licenses/LICENSE-2.0 010 * 011 * Unless required by applicable law or agreed to in writing, software 012 * distributed under the License is distributed on an "AS IS" BASIS, 013 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 014 * See the License for the specific language governing permissions and 015 * limitations under the License. 016 */ 017package org.apache.commons.codec.digest; 018 019import java.util.Arrays; 020import java.util.Objects; 021 022/** 023 * Implements the Blake3 algorithm providing a {@linkplain #initHash() hash function} with extensible output (XOF), a 024 * {@linkplain #initKeyedHash(byte[]) keyed hash function} (MAC, PRF), and a 025 * {@linkplain #initKeyDerivationFunction(byte[]) key derivation function} (KDF). Blake3 has a 128-bit security level 026 * and a default output length of 256 bits (32 bytes) which can extended up to 2<sup>64</sup> bytes. 027 * <h2>Hashing</h2> 028 * <p>Hash mode calculates the same output hash given the same input bytes and can be used as both a message digest and 029 * and extensible output function.</p> 030 * <pre>{@code 031 * Blake3 hasher = Blake3.initHash(); 032 * hasher.update("Hello, world!".getBytes(StandardCharsets.UTF_8)); 033 * byte[] hash = new byte[32]; 034 * hasher.doFinalize(hash); 035 * }</pre> 036 * <h2>Keyed Hashing</h2> 037 * <p>Keyed hashes take a 32-byte secret key and calculates a message authentication code on some input bytes. These 038 * also work as pseudo-random functions (PRFs) with extensible output similar to the extensible hash output. Note that 039 * Blake3 keyed hashes have the same performance as plain hashes; the key is used in initialization in place of a 040 * standard initialization vector used for plain hashing.</p> 041 * <pre>{@code 042 * SecureRandom random = SecureRandom.getInstanceStrong(); 043 * byte[] key = new byte[32]; 044 * random.nextBytes(key); 045 * Blake3 hasher = Blake3.initKeyedHash(key); 046 * hasher.update("Hello, Alice!".getBytes(StandardCharsets.UTF_8)); 047 * byte[] mac = new byte[32]; 048 * hasher.doFinalize(mac); 049 * }</pre> 050 * <h2>Key Derivation</h2> 051 * <p>A specific hash mode for deriving session keys and other derived keys in a unique key derivation context 052 * identified by some sequence of bytes. These context strings should be unique but do not need to be kept secret. 053 * Additional input data is hashed for key material which can be finalized to derive subkeys.</p> 054 * <pre>{@code 055 * String context = "org.apache.commons.codec.digest.Blake3Example"; 056 * byte[] sharedSecret = ...; 057 * byte[] senderId = ...; 058 * byte[] recipientId = ...; 059 * Blake3 kdf = Blake3.initKeyDerivationFunction(context.getBytes(StandardCharsets.UTF_8)); 060 * kdf.update(sharedSecret); 061 * kdf.update(senderId); 062 * kdf.update(recipientId); 063 * byte[] txKey = new byte[32]; 064 * byte[] rxKey = new byte[32]; 065 * kdf.doFinalize(txKey); 066 * kdf.doFinalize(rxKey); 067 * }</pre> 068 * <p> 069 * Adapted from the ISC-licensed O(1) Cryptography library by Matt Sicker and ported from the reference public domain 070 * implementation by Jack O'Connor. 071 * </p> 072 * 073 * @see <a href="https://github.com/BLAKE3-team/BLAKE3">BLAKE3 hash function</a> 074 * @since 1.16 075 */ 076public final class Blake3 { 077 078 private static final int BLOCK_LEN = 64; 079 private static final int BLOCK_INTS = BLOCK_LEN / Integer.BYTES; 080 private static final int KEY_LEN = 32; 081 private static final int KEY_INTS = KEY_LEN / Integer.BYTES; 082 private static final int OUT_LEN = 32; 083 private static final int CHUNK_LEN = 1024; 084 private static final int CHAINING_VALUE_INTS = 8; 085 086 /** 087 * Standard hash key used for plain hashes; same initialization vector as Blake2s. 088 */ 089 private static final int[] IV = 090 { 0x6A09E667, 0xBB67AE85, 0x3C6EF372, 0xA54FF53A, 0x510E527F, 0x9B05688C, 0x1F83D9AB, 0x5BE0CD19 }; 091 092 // domain flags 093 private static final int CHUNK_START = 1; 094 private static final int CHUNK_END = 1 << 1; 095 private static final int PARENT = 1 << 2; 096 private static final int ROOT = 1 << 3; 097 private static final int KEYED_HASH = 1 << 4; 098 private static final int DERIVE_KEY_CONTEXT = 1 << 5; 099 private static final int DERIVE_KEY_MATERIAL = 1 << 6; 100 101 /** 102 * Pre-permuted for all 7 rounds; the second row (2,6,3,...) indicates the base permutation. 103 */ 104 private static final byte[][] MSG_SCHEDULE = { 105 { 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15 }, 106 { 2, 6, 3, 10, 7, 0, 4, 13, 1, 11, 12, 5, 9, 14, 15, 8 }, 107 { 3, 4, 10, 12, 13, 2, 7, 14, 6, 5, 9, 0, 11, 15, 8, 1 }, 108 { 10, 7, 12, 9, 14, 3, 13, 15, 4, 0, 11, 2, 5, 8, 1, 6 }, 109 { 12, 13, 9, 11, 15, 10, 14, 8, 7, 2, 5, 3, 0, 1, 6, 4 }, 110 { 9, 14, 11, 5, 8, 12, 15, 1, 13, 3, 0, 10, 2, 6, 4, 7 }, 111 { 11, 15, 5, 0, 1, 9, 8, 6, 14, 10, 2, 12, 3, 4, 7, 13 } 112 }; 113 114 private final EngineState engineState; 115 116 private Blake3(final int[] key, final int flags) { 117 engineState = new EngineState(key, flags); 118 } 119 120 /** 121 * Resets this instance back to its initial state when it was first constructed. 122 * @return this 123 */ 124 public Blake3 reset() { 125 engineState.reset(); 126 return this; 127 } 128 129 /** 130 * Updates this hash state using the provided bytes. 131 * 132 * @param in source array to update data from 133 * @return this 134 * @throws NullPointerException if in is null 135 */ 136 public Blake3 update(final byte[] in) { 137 return update(in, 0, in.length); 138 } 139 140 /** 141 * Updates this hash state using the provided bytes at an offset. 142 * 143 * @param in source array to update data from 144 * @param offset where in the array to begin reading bytes 145 * @param length number of bytes to update 146 * @return this 147 * @throws NullPointerException if in is null 148 * @throws IndexOutOfBoundsException if offset or length are negative or if offset + length is greater than the 149 * length of the provided array 150 */ 151 public Blake3 update(final byte[] in, final int offset, final int length) { 152 checkBufferArgs(in, offset, length); 153 engineState.inputData(in, offset, length); 154 return this; 155 } 156 157 /** 158 * Finalizes hash output data that depends on the sequence of updated bytes preceding this invocation and any 159 * previously finalized bytes. Note that this can finalize up to 2<sup>64</sup> bytes per instance. 160 * 161 * @param out destination array to finalize bytes into 162 * @return this 163 * @throws NullPointerException if out is null 164 */ 165 public Blake3 doFinalize(final byte[] out) { 166 return doFinalize(out, 0, out.length); 167 } 168 169 /** 170 * Finalizes an arbitrary number of bytes into the provided output array that depends on the sequence of previously 171 * updated and finalized bytes. Note that this can finalize up to 2<sup>64</sup> bytes per instance. 172 * 173 * @param out destination array to finalize bytes into 174 * @param offset where in the array to begin writing bytes to 175 * @param length number of bytes to finalize 176 * @return this 177 * @throws NullPointerException if out is null 178 * @throws IndexOutOfBoundsException if offset or length are negative or if offset + length is greater than the 179 * length of the provided array 180 */ 181 public Blake3 doFinalize(final byte[] out, final int offset, final int length) { 182 checkBufferArgs(out, offset, length); 183 engineState.outputHash(out, offset, length); 184 return this; 185 } 186 187 /** 188 * Squeezes and returns an arbitrary number of bytes dependent on the sequence of previously absorbed and squeezed bytes. 189 * 190 * @param nrBytes number of bytes to finalize 191 * @return requested number of finalized bytes 192 * @throws IllegalArgumentException if nrBytes is negative 193 */ 194 public byte[] doFinalize(final int nrBytes) { 195 if (nrBytes < 0) { 196 throw new IllegalArgumentException("Requested bytes must be non-negative"); 197 } 198 final byte[] hash = new byte[nrBytes]; 199 doFinalize(hash); 200 return hash; 201 } 202 203 /** 204 * Constructs a fresh Blake3 hash function. The instance returned functions as an arbitrary length message digest. 205 * 206 * @return fresh Blake3 instance in hashed mode 207 */ 208 public static Blake3 initHash() { 209 return new Blake3(IV, 0); 210 } 211 212 /** 213 * Constructs a fresh Blake3 keyed hash function. The instance returned functions as a pseudorandom function (PRF) or as a 214 * message authentication code (MAC). 215 * 216 * @param key 32-byte secret key 217 * @return fresh Blake3 instance in keyed mode using the provided key 218 * @throws NullPointerException if key is null 219 * @throws IllegalArgumentException if key is not 32 bytes 220 */ 221 public static Blake3 initKeyedHash(final byte[] key) { 222 Objects.requireNonNull(key); 223 if (key.length != KEY_LEN) { 224 throw new IllegalArgumentException("Blake3 keys must be 32 bytes"); 225 } 226 return new Blake3(unpackInts(key, KEY_INTS), KEYED_HASH); 227 } 228 229 /** 230 * Constructs a fresh Blake3 key derivation function using the provided key derivation context byte string. 231 * The instance returned functions as a key-derivation function which can further absorb additional context data 232 * before squeezing derived key data. 233 * 234 * @param kdfContext a globally unique key-derivation context byte string to separate key derivation contexts from each other 235 * @return fresh Blake3 instance in key derivation mode 236 * @throws NullPointerException if kdfContext is null 237 */ 238 public static Blake3 initKeyDerivationFunction(final byte[] kdfContext) { 239 Objects.requireNonNull(kdfContext); 240 final EngineState kdf = new EngineState(IV, DERIVE_KEY_CONTEXT); 241 kdf.inputData(kdfContext, 0, kdfContext.length); 242 final byte[] key = new byte[KEY_LEN]; 243 kdf.outputHash(key, 0, key.length); 244 return new Blake3(unpackInts(key, KEY_INTS), DERIVE_KEY_MATERIAL); 245 } 246 247 /** 248 * Calculates the Blake3 hash of the provided data. 249 * 250 * @param data source array to absorb data from 251 * @return 32-byte hash squeezed from the provided data 252 * @throws NullPointerException if data is null 253 */ 254 public static byte[] hash(final byte[] data) { 255 return Blake3.initHash().update(data).doFinalize(OUT_LEN); 256 } 257 258 /** 259 * Calculates the Blake3 keyed hash (MAC) of the provided data. 260 * 261 * @param key 32-byte secret key 262 * @param data source array to absorb data from 263 * @return 32-byte mac squeezed from the provided data 264 * @throws NullPointerException if key or data are null 265 */ 266 public static byte[] keyedHash(final byte[] key, final byte[] data) { 267 return Blake3.initKeyedHash(key).update(data).doFinalize(OUT_LEN); 268 } 269 270 private static void checkBufferArgs(final byte[] buffer, final int offset, final int length) { 271 Objects.requireNonNull(buffer); 272 if (offset < 0) { 273 throw new IndexOutOfBoundsException("Offset must be non-negative"); 274 } 275 if (length < 0) { 276 throw new IndexOutOfBoundsException("Length must be non-negative"); 277 } 278 final int bufferLength = buffer.length; 279 if (offset > bufferLength - length) { 280 throw new IndexOutOfBoundsException( 281 "Offset " + offset + " and length " + length + " out of bounds with buffer length " + bufferLength); 282 } 283 } 284 285 private static void packInt(final int value, final byte[] dst, final int off, final int len) { 286 for (int i = 0; i < len; i++) { 287 dst[off + i] = (byte) (value >>> i * Byte.SIZE); 288 } 289 } 290 291 private static int unpackInt(final byte[] buf, final int off) { 292 return buf[off] & 0xFF | (buf[off + 1] & 0xFF) << 8 | (buf[off + 2] & 0xFF) << 16 | (buf[off + 3] & 0xFF) << 24; 293 } 294 295 private static int[] unpackInts(final byte[] buf, final int nrInts) { 296 final int[] values = new int[nrInts]; 297 for (int i = 0, off = 0; i < nrInts; i++, off += Integer.BYTES) { 298 values[i] = unpackInt(buf, off); 299 } 300 return values; 301 } 302 303 /** 304 * The mixing function, G, which mixes either a column or a diagonal. 305 */ 306 private static void g( 307 final int[] state, final int a, final int b, final int c, final int d, final int mx, final int my) { 308 state[a] += state[b] + mx; 309 state[d] = Integer.rotateRight(state[d] ^ state[a], 16); 310 state[c] += state[d]; 311 state[b] = Integer.rotateRight(state[b] ^ state[c], 12); 312 state[a] += state[b] + my; 313 state[d] = Integer.rotateRight(state[d] ^ state[a], 8); 314 state[c] += state[d]; 315 state[b] = Integer.rotateRight(state[b] ^ state[c], 7); 316 } 317 318 private static void round(final int[] state, final int[] msg, final byte[] schedule) { 319 // Mix the columns. 320 g(state, 0, 4, 8, 12, msg[schedule[0]], msg[schedule[1]]); 321 g(state, 1, 5, 9, 13, msg[schedule[2]], msg[schedule[3]]); 322 g(state, 2, 6, 10, 14, msg[schedule[4]], msg[schedule[5]]); 323 g(state, 3, 7, 11, 15, msg[schedule[6]], msg[schedule[7]]); 324 325 // Mix the diagonals. 326 g(state, 0, 5, 10, 15, msg[schedule[8]], msg[schedule[9]]); 327 g(state, 1, 6, 11, 12, msg[schedule[10]], msg[schedule[11]]); 328 g(state, 2, 7, 8, 13, msg[schedule[12]], msg[schedule[13]]); 329 g(state, 3, 4, 9, 14, msg[schedule[14]], msg[schedule[15]]); 330 } 331 332 private static int[] compress( 333 final int[] chainingValue, final int[] blockWords, final int blockLength, final long counter, 334 final int flags) { 335 final int[] state = Arrays.copyOf(chainingValue, BLOCK_INTS); 336 System.arraycopy(IV, 0, state, 8, 4); 337 state[12] = (int) counter; 338 state[13] = (int) (counter >> Integer.SIZE); 339 state[14] = blockLength; 340 state[15] = flags; 341 for (int i = 0; i < 7; i++) { 342 final byte[] schedule = MSG_SCHEDULE[i]; 343 round(state, blockWords, schedule); 344 } 345 for (int i = 0; i < state.length / 2; i++) { 346 state[i] ^= state[i + 8]; 347 state[i + 8] ^= chainingValue[i]; 348 } 349 return state; 350 } 351 352 private static Output parentOutput( 353 final int[] leftChildCV, final int[] rightChildCV, final int[] key, final int flags) { 354 final int[] blockWords = Arrays.copyOf(leftChildCV, BLOCK_INTS); 355 System.arraycopy(rightChildCV, 0, blockWords, 8, CHAINING_VALUE_INTS); 356 return new Output(key.clone(), blockWords, 0, BLOCK_LEN, flags | PARENT); 357 } 358 359 private static int[] parentChainingValue( 360 final int[] leftChildCV, final int[] rightChildCV, final int[] key, final int flags) { 361 return parentOutput(leftChildCV, rightChildCV, key, flags).chainingValue(); 362 } 363 364 /** 365 * Represents the state just prior to either producing an eight word chaining value or any number of output bytes 366 * when the ROOT flag is set. 367 */ 368 private static class Output { 369 private final int[] inputChainingValue; 370 private final int[] blockWords; 371 private final long counter; 372 private final int blockLength; 373 private final int flags; 374 375 private Output( 376 final int[] inputChainingValue, final int[] blockWords, final long counter, final int blockLength, 377 final int flags) { 378 this.inputChainingValue = inputChainingValue; 379 this.blockWords = blockWords; 380 this.counter = counter; 381 this.blockLength = blockLength; 382 this.flags = flags; 383 } 384 385 private int[] chainingValue() { 386 return Arrays 387 .copyOf(compress(inputChainingValue, blockWords, blockLength, counter, flags), CHAINING_VALUE_INTS); 388 } 389 390 private void rootOutputBytes(final byte[] out, int offset, int length) { 391 int outputBlockCounter = 0; 392 while (length > 0) { 393 int chunkLength = Math.min(OUT_LEN * 2, length); 394 length -= chunkLength; 395 final int[] words = 396 compress(inputChainingValue, blockWords, blockLength, outputBlockCounter++, flags | ROOT); 397 int wordCounter = 0; 398 while (chunkLength > 0) { 399 final int wordLength = Math.min(Integer.BYTES, chunkLength); 400 packInt(words[wordCounter++], out, offset, wordLength); 401 offset += wordLength; 402 chunkLength -= wordLength; 403 } 404 } 405 } 406 } 407 408 private static class ChunkState { 409 private int[] chainingValue; 410 private final long chunkCounter; 411 private final int flags; 412 413 private final byte[] block = new byte[BLOCK_LEN]; 414 private int blockLength; 415 private int blocksCompressed; 416 417 private ChunkState(final int[] key, final long chunkCounter, final int flags) { 418 chainingValue = key; 419 this.chunkCounter = chunkCounter; 420 this.flags = flags; 421 } 422 423 private int length() { 424 return BLOCK_LEN * blocksCompressed + blockLength; 425 } 426 427 private int startFlag() { 428 return blocksCompressed == 0 ? CHUNK_START : 0; 429 } 430 431 private void update(final byte[] input, int offset, int length) { 432 while (length > 0) { 433 if (blockLength == BLOCK_LEN) { 434 // If the block buffer is full, compress it and clear it. More 435 // input is coming, so this compression is not CHUNK_END. 436 final int[] blockWords = unpackInts(block, BLOCK_INTS); 437 chainingValue = Arrays.copyOf( 438 compress(chainingValue, blockWords, BLOCK_LEN, chunkCounter, flags | startFlag()), 439 CHAINING_VALUE_INTS); 440 blocksCompressed++; 441 blockLength = 0; 442 Arrays.fill(block, (byte) 0); 443 } 444 445 final int want = BLOCK_LEN - blockLength; 446 final int take = Math.min(want, length); 447 System.arraycopy(input, offset, block, blockLength, take); 448 blockLength += take; 449 offset += take; 450 length -= take; 451 } 452 } 453 454 private Output output() { 455 final int[] blockWords = unpackInts(block, BLOCK_INTS); 456 final int outputFlags = flags | startFlag() | CHUNK_END; 457 return new Output(chainingValue, blockWords, chunkCounter, blockLength, outputFlags); 458 } 459 } 460 461 private static class EngineState { 462 private final int[] key; 463 private final int flags; 464 // Space for 54 subtree chaining values: 2^54 * CHUNK_LEN = 2^64 465 // No more than 54 entries can ever be added to this stack (after updating 2^64 bytes and not finalizing any) 466 // so we preallocate the stack here. This can be smaller in environments where the data limit is expected to 467 // be much lower. 468 private final int[][] cvStack = new int[54][]; 469 private int stackLen; 470 private ChunkState state; 471 472 private EngineState(final int[] key, final int flags) { 473 this.key = key; 474 this.flags = flags; 475 state = new ChunkState(key, 0, flags); 476 } 477 478 private void inputData(final byte[] in, int offset, int length) { 479 while (length > 0) { 480 // If the current chunk is complete, finalize it and reset the 481 // chunk state. More input is coming, so this chunk is not ROOT. 482 if (state.length() == CHUNK_LEN) { 483 final int[] chunkCV = state.output().chainingValue(); 484 final long totalChunks = state.chunkCounter + 1; 485 addChunkCV(chunkCV, totalChunks); 486 state = new ChunkState(key, totalChunks, flags); 487 } 488 489 // Compress input bytes into the current chunk state. 490 final int want = CHUNK_LEN - state.length(); 491 final int take = Math.min(want, length); 492 state.update(in, offset, take); 493 offset += take; 494 length -= take; 495 } 496 } 497 498 private void outputHash(final byte[] out, final int offset, final int length) { 499 // Starting with the Output from the current chunk, compute all the 500 // parent chaining values along the right edge of the tree, until we 501 // have the root Output. 502 Output output = state.output(); 503 int parentNodesRemaining = stackLen; 504 while (parentNodesRemaining-- > 0) { 505 final int[] parentCV = cvStack[parentNodesRemaining]; 506 output = parentOutput(parentCV, output.chainingValue(), key, flags); 507 } 508 output.rootOutputBytes(out, offset, length); 509 } 510 511 private void reset() { 512 stackLen = 0; 513 Arrays.fill(cvStack, null); 514 state = new ChunkState(key, 0, flags); 515 } 516 517 // Section 5.1.2 of the BLAKE3 spec explains this algorithm in more detail. 518 private void addChunkCV(final int[] firstCV, final long totalChunks) { 519 // This chunk might complete some subtrees. For each completed subtree, 520 // its left child will be the current top entry in the CV stack, and 521 // its right child will be the current value of `newCV`. Pop each left 522 // child off the stack, merge it with `newCV`, and overwrite `newCV` 523 // with the result. After all these merges, push the final value of 524 // `newCV` onto the stack. The number of completed subtrees is given 525 // by the number of trailing 0-bits in the new total number of chunks. 526 int[] newCV = firstCV; 527 long chunkCounter = totalChunks; 528 while ((chunkCounter & 1) == 0) { 529 newCV = parentChainingValue(popCV(), newCV, key, flags); 530 chunkCounter >>= 1; 531 } 532 pushCV(newCV); 533 } 534 535 private void pushCV(final int[] cv) { 536 cvStack[stackLen++] = cv; 537 } 538 539 private int[] popCV() { 540 return cvStack[--stackLen]; 541 } 542 } 543 544}