1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068
// Copyright (c) Meta Platforms, Inc. and affiliates.
//
// This source code is dual-licensed under either the MIT license found in the
// LICENSE-MIT file in the root directory of this source tree or the Apache
// License, Version 2.0 found in the LICENSE-APACHE file in the root directory
// of this source tree. You may select, at your option, one of the above-listed licenses.
//! Implementation of an auditable key directory
use crate::append_only_zks::{Azks, InsertMode};
use crate::ecvrf::{VRFKeyStorage, VRFPublicKey};
use crate::errors::{AkdError, DirectoryError, StorageError};
use crate::helper_structs::LookupInfo;
use crate::storage::manager::StorageManager;
use crate::storage::types::{DbRecord, ValueState, ValueStateRetrievalFlag};
use crate::storage::Database;
use crate::{
AkdLabel, AkdValue, AppendOnlyProof, AzksElement, Digest, EpochHash, HistoryProof, LookupProof,
NonMembershipProof, UpdateProof,
};
use crate::VersionFreshness;
use akd_core::configuration::Configuration;
use log::{error, info};
use std::collections::{HashMap, HashSet};
use std::marker::PhantomData;
use std::sync::Arc;
use tokio::sync::RwLock;
/// The representation of a auditable key directory
pub struct Directory<TC, S: Database, V> {
storage: StorageManager<S>,
vrf: V,
/// The cache lock guarantees that the cache is not
/// flushed mid-proof generation. We allow multiple proof generations
/// to occur (RwLock.read() operations can have multiple) but we want
/// to make sure no generations are underway when a cache flush occurs
/// (in this case we do utilize the write() lock which can only occur 1
/// at a time and gates further read() locks being acquired during write()).
cache_lock: Arc<RwLock<()>>,
tc: PhantomData<TC>,
}
// Manual implementation of Clone, see: https://github.com/rust-lang/rust/issues/41481
impl<TC, S: Database, V: VRFKeyStorage> Clone for Directory<TC, S, V> {
fn clone(&self) -> Self {
Self {
storage: self.storage.clone(),
vrf: self.vrf.clone(),
cache_lock: self.cache_lock.clone(),
tc: PhantomData,
}
}
}
impl<TC, S, V> Directory<TC, S, V>
where
TC: Configuration,
S: Database + 'static,
V: VRFKeyStorage,
{
/// Creates a new (stateless) instance of a auditable key directory.
/// Takes as input a pointer to the storage being used for this instance.
/// The state is stored in the storage.
pub async fn new(storage: StorageManager<S>, vrf: V) -> Result<Self, AkdError> {
let azks = Directory::<TC, S, V>::get_azks_from_storage(&storage, false).await;
if azks.is_err() {
// generate a new azks if one is not found
let azks = Azks::new::<TC, _>(&storage).await?;
// store it
storage.set(DbRecord::Azks(azks.clone())).await?;
}
Ok(Directory {
storage,
cache_lock: Arc::new(RwLock::new(())),
vrf,
tc: PhantomData,
})
}
/// Updates the directory to include the input label-value pairs.
///
/// Note that the vector of label-value pairs should not contain any entries with duplicate labels. This
/// condition is explicitly checked, and an error will be returned if this is the case.
pub async fn publish(&self, updates: Vec<(AkdLabel, AkdValue)>) -> Result<EpochHash, AkdError> {
// The guard will be dropped at the end of the publish
let _guard = self.cache_lock.read().await;
// Check for duplicate labels and return an error if any are encountered
let distinct_set: HashSet<AkdLabel> =
updates.iter().map(|(label, _)| label.clone()).collect();
if distinct_set.len() != updates.len() {
return Err(AkdError::Directory(DirectoryError::Publish(
"Cannot publish with a set of entries that contain duplicate labels".to_string(),
)));
}
let mut update_set = Vec::<AzksElement>::new();
let mut user_data_update_set = Vec::<ValueState>::new();
let mut current_azks = self.retrieve_azks().await?;
let current_epoch = current_azks.get_latest_epoch();
let next_epoch = current_epoch + 1;
let mut keys: Vec<AkdLabel> = updates
.iter()
.map(|(akd_label, _val)| akd_label.clone())
.collect();
// sort the keys, as inserting in primary-key order is more efficient for MySQL
keys.sort();
// we're only using the maximum "version" of the user's state at the last epoch
// they were seen in the directory. Therefore we've minimized the call to only
// return a hashmap of AkdLabel => u64 and not retrieving the other data which is not
// read (i.e. the actual _data_ payload).
let all_user_versions_retrieved = self
.storage
.get_user_state_versions(&keys, ValueStateRetrievalFlag::LeqEpoch(current_epoch))
.await?;
info!(
"Retrieved {} previous user versions of {} requested",
all_user_versions_retrieved.len(),
keys.len()
);
let vrf_computations = updates
.iter()
.flat_map(
|(akd_label, akd_value)| match all_user_versions_retrieved.get(akd_label) {
None => vec![(
akd_label.clone(),
VersionFreshness::Fresh,
1u64,
akd_value.clone(),
)],
Some((latest_version, existing_akd_value)) => {
if existing_akd_value == akd_value {
// Skip this because the user is trying to re-publish the same value
return vec![];
}
vec![
(
akd_label.clone(),
VersionFreshness::Stale,
*latest_version,
akd_value.clone(),
),
(
akd_label.clone(),
VersionFreshness::Fresh,
*latest_version + 1,
akd_value.clone(),
),
]
}
},
)
.collect::<Vec<_>>();
let vrf_map = self
.vrf
.get_node_labels::<TC>(&vrf_computations)
.await?
.into_iter()
.collect::<HashMap<_, _>>();
let commitment_key = self.derive_commitment_key().await?;
for ((akd_label, freshness, version, akd_value), node_label) in vrf_map {
let azks_value = match freshness {
VersionFreshness::Stale => TC::stale_azks_value(),
VersionFreshness::Fresh => {
TC::compute_fresh_azks_value(&commitment_key, &node_label, version, &akd_value)
}
};
update_set.push(AzksElement {
label: node_label,
value: azks_value,
});
if freshness == VersionFreshness::Fresh {
let latest_state =
ValueState::new(akd_label, akd_value, version, node_label, next_epoch);
user_data_update_set.push(latest_state);
}
}
if update_set.is_empty() {
info!("After filtering for duplicated user information, there is no publish which is necessary (0 updates)");
// The AZKS has not been updated/mutated at this point, so we can just return the root hash from before
let root_hash = current_azks.get_root_hash::<TC, _>(&self.storage).await?;
return Ok(EpochHash(current_epoch, root_hash));
}
if !self.storage.begin_transaction() {
error!("Transaction is already active");
return Err(AkdError::Storage(StorageError::Transaction(
"Transaction is already active".to_string(),
)));
}
info!("Starting inserting new leaves");
if let Err(err) = current_azks
.batch_insert_nodes::<TC, _>(&self.storage, update_set, InsertMode::Directory)
.await
{
// If we fail to do the batch-leaf insert, we should rollback the transaction so we can try again cleanly.
// Only fails if transaction is not currently active.
let _ = self.storage.rollback_transaction();
// bubble up the err
return Err(err);
}
// batch all the inserts into a single write to storage (in this case it insert's into the transaction log)
let mut updates = vec![DbRecord::Azks(current_azks.clone())];
for update in user_data_update_set.into_iter() {
updates.push(DbRecord::ValueState(update));
}
self.storage.batch_set(updates).await?;
// Commit the transaction
info!("Committing transaction");
match self.storage.commit_transaction().await {
Ok(num_records) => {
info!("Transaction committed ({} records)", num_records);
}
Err(err) => {
error!("Failed to commit transaction, rolling back");
let _ = self.storage.rollback_transaction();
return Err(AkdError::Storage(err));
}
};
let root_hash = current_azks
.get_root_hash_safe::<TC, _>(&self.storage, next_epoch)
.await?;
Ok(EpochHash(next_epoch, root_hash))
}
/// Provides proof for correctness of latest version
///
/// * `akd_label`: The target label to generate a lookup proof for
///
/// Returns [Ok((LookupProof, EpochHash))] upon successful generation for the latest version
/// of the target label's state. [Err(_)] otherwise
pub async fn lookup(&self, akd_label: AkdLabel) -> Result<(LookupProof, EpochHash), AkdError> {
// The guard will be dropped at the end of the proof generation
let _guard = self.cache_lock.read().await;
let current_azks = self.retrieve_azks().await?;
let current_epoch = current_azks.get_latest_epoch();
let lookup_info = self.get_lookup_info(akd_label, current_epoch).await?;
let root_hash = EpochHash(
current_epoch,
current_azks.get_root_hash::<TC, _>(&self.storage).await?,
);
let proof = self
.lookup_with_info(¤t_azks, lookup_info, false)
.await?;
Ok((proof, root_hash))
}
/// Generate a lookup proof with the provided target information
///
/// * `current_azks`: The current [Azks] element
/// * `lookup_info`: The information to target in the lookup request. Includes all
/// necessary information to build the proof
/// * `skip_preload`: Denotes if we should not preload as part of this optimization. Enabled
/// from bulk lookup proof generation, as it has its own preloading operation
///
/// Returns [Ok(LookupProof)] if the proof generation succeeded, [Err(_)] otherwise
async fn lookup_with_info(
&self,
current_azks: &Azks,
lookup_info: LookupInfo,
skip_preload: bool,
) -> Result<LookupProof, AkdError> {
if !skip_preload {
// Preload nodes needed for lookup.
#[cfg(feature = "greedy_lookup_preload")]
{
current_azks
.greedy_preload_lookup_nodes(&self.storage, lookup_info.clone())
.await?;
}
#[cfg(not(feature = "greedy_lookup_preload"))]
{
current_azks
.preload_lookup_nodes(&self.storage, &vec![lookup_info.clone()])
.await?;
}
}
let label = &lookup_info.value_state.username;
let current_version = lookup_info.value_state.version;
let commitment_key = self.derive_commitment_key().await?;
let plaintext_value = lookup_info.value_state.value;
let existence_vrf = self
.vrf
.get_label_proof::<TC>(label, VersionFreshness::Fresh, current_version)
.await?;
let commitment_label = self.vrf.get_node_label_from_vrf_proof(existence_vrf).await;
let lookup_proof = LookupProof {
epoch: lookup_info.value_state.epoch,
value: plaintext_value.clone(),
version: lookup_info.value_state.version,
existence_vrf_proof: existence_vrf.to_bytes().to_vec(),
existence_proof: current_azks
.get_membership_proof::<TC, _>(&self.storage, lookup_info.existent_label)
.await?,
marker_vrf_proof: self
.vrf
.get_label_proof::<TC>(label, VersionFreshness::Fresh, lookup_info.marker_version)
.await?
.to_bytes()
.to_vec(),
marker_proof: current_azks
.get_membership_proof::<TC, _>(&self.storage, lookup_info.marker_label)
.await?,
freshness_vrf_proof: self
.vrf
.get_label_proof::<TC>(label, VersionFreshness::Stale, current_version)
.await?
.to_bytes()
.to_vec(),
freshness_proof: current_azks
.get_non_membership_proof::<TC, _>(&self.storage, lookup_info.non_existent_label)
.await?,
commitment_nonce: TC::get_commitment_nonce(
&commitment_key,
&commitment_label,
lookup_info.value_state.version,
&plaintext_value,
)
.to_vec(),
};
Ok(lookup_proof)
}
// TODO(eoz): Call proof generations async
/// Allows efficient batch lookups by preloading necessary nodes for the lookups.
pub async fn batch_lookup(
&self,
akd_labels: &[AkdLabel],
) -> Result<(Vec<LookupProof>, EpochHash), AkdError> {
// The guard will be dropped at the end of the proof generation
let _guard = self.cache_lock.read().await;
let current_azks = self.retrieve_azks().await?;
let current_epoch = current_azks.get_latest_epoch();
// Take a union of the labels we will need proofs of for each lookup.
let mut lookup_infos = Vec::new();
for akd_label in akd_labels {
// Save lookup info for later use.
let lookup_info = self
.get_lookup_info(akd_label.clone(), current_epoch)
.await?;
lookup_infos.push(lookup_info.clone());
}
// Load nodes needed using the lookup infos.
current_azks
.preload_lookup_nodes(&self.storage, &lookup_infos)
.await?;
// Ensure we have got all lookup infos needed.
assert_eq!(akd_labels.len(), lookup_infos.len());
let root_hash = EpochHash(
current_epoch,
current_azks.get_root_hash::<TC, _>(&self.storage).await?,
);
let mut lookup_proofs = Vec::new();
for info in lookup_infos.into_iter() {
lookup_proofs.push(self.lookup_with_info(¤t_azks, info, true).await?);
}
Ok((lookup_proofs, root_hash))
}
async fn build_lookup_info(&self, latest_st: &ValueState) -> Result<LookupInfo, AkdError> {
let akd_label = &latest_st.username;
// Need to account for the case where the latest state is
// added but the database is in the middle of an update
let version = latest_st.version;
let marker_version = 1 << get_marker_version(version);
let existent_label = self
.vrf
.get_node_label::<TC>(akd_label, VersionFreshness::Fresh, version)
.await?;
let marker_label = self
.vrf
.get_node_label::<TC>(akd_label, VersionFreshness::Fresh, marker_version)
.await?;
let non_existent_label = self
.vrf
.get_node_label::<TC>(akd_label, VersionFreshness::Stale, version)
.await?;
Ok(LookupInfo {
value_state: latest_st.clone(),
marker_version,
existent_label,
marker_label,
non_existent_label,
})
}
async fn get_lookup_info(
&self,
akd_label: AkdLabel,
epoch: u64,
) -> Result<LookupInfo, AkdError> {
match self
.storage
.get_user_state(&akd_label, ValueStateRetrievalFlag::LeqEpoch(epoch))
.await
{
Err(_) => {
// Need to throw an error
match std::str::from_utf8(&akd_label) {
Ok(name) => Err(AkdError::Storage(StorageError::NotFound(format!(
"User {name} at epoch {epoch}"
)))),
_ => Err(AkdError::Storage(StorageError::NotFound(format!(
"User {akd_label:?} at epoch {epoch}"
)))),
}
}
Ok(latest_st) => self.build_lookup_info(&latest_st).await,
}
}
/// Takes in the current state of the server and a label.
/// If the label is present in the current state,
/// this function returns all the values ever associated with it,
/// and the epoch at which each value was first committed to the server state.
/// It also returns the proof of the latest version being served at all times.
pub async fn key_history(
&self,
akd_label: &AkdLabel,
params: HistoryParams,
) -> Result<(HistoryProof, EpochHash), AkdError> {
// The guard will be dropped at the end of the proof generation
let _guard = self.cache_lock.read().await;
let current_azks = self.retrieve_azks().await?;
let current_epoch = current_azks.get_latest_epoch();
let mut user_data = self.storage.get_user_data(akd_label).await?.states;
// reverse sort from highest epoch to lowest
user_data.sort_by(|a, b| b.epoch.cmp(&a.epoch));
// apply filters specified by HistoryParams struct
user_data = match params {
HistoryParams::Complete => user_data,
HistoryParams::MostRecent(n) => user_data.into_iter().take(n).collect::<Vec<_>>(),
HistoryParams::SinceEpoch(epoch) => {
user_data = user_data
.into_iter()
.filter(|val| val.epoch >= epoch)
.collect::<Vec<_>>();
// Ordering should be maintained after filtering, but let's re-sort just in case
user_data.sort_by(|a, b| b.epoch.cmp(&a.epoch));
user_data
}
};
if user_data.is_empty() {
let msg = if let Ok(username_str) = std::str::from_utf8(akd_label) {
format!("User {username_str}")
} else {
format!("User {akd_label:?}")
};
return Err(AkdError::Storage(StorageError::NotFound(msg)));
}
#[cfg(feature = "preload_history")]
{
let mut lookup_infos = vec![];
for ud in user_data.iter() {
if let Ok(lo) = self.build_lookup_info(ud).await {
lookup_infos.push(lo);
}
}
current_azks
.preload_lookup_nodes(&self.storage, &lookup_infos)
.await?;
}
let mut update_proofs = Vec::<UpdateProof>::new();
let mut last_version = 0;
for user_state in user_data {
// Ignore states in storage that are ahead of current directory epoch
if user_state.epoch <= current_epoch {
let proof = self
.create_single_update_proof(akd_label, &user_state)
.await?;
update_proofs.push(proof);
last_version = if user_state.version > last_version {
user_state.version
} else {
last_version
};
}
}
let next_marker = get_marker_version(last_version) + 1;
let final_marker = get_marker_version(current_epoch);
let mut until_marker_vrf_proofs = Vec::<Vec<u8>>::new();
let mut non_existence_until_marker_proofs = Vec::<NonMembershipProof>::new();
for ver in last_version + 1..(1 << next_marker) {
let label_for_ver = self
.vrf
.get_node_label::<TC>(akd_label, VersionFreshness::Fresh, ver)
.await?;
let non_existence_of_ver = current_azks
.get_non_membership_proof::<TC, _>(&self.storage, label_for_ver)
.await?;
non_existence_until_marker_proofs.push(non_existence_of_ver);
until_marker_vrf_proofs.push(
self.vrf
.get_label_proof::<TC>(akd_label, VersionFreshness::Fresh, ver)
.await?
.to_bytes()
.to_vec(),
);
}
let mut future_marker_vrf_proofs = Vec::<Vec<u8>>::new();
let mut non_existence_of_future_marker_proofs = Vec::<NonMembershipProof>::new();
for marker_power in next_marker..final_marker + 1 {
let ver = 1 << marker_power;
let label_for_ver = self
.vrf
.get_node_label::<TC>(akd_label, VersionFreshness::Fresh, ver)
.await?;
let non_existence_of_ver = current_azks
.get_non_membership_proof::<TC, _>(&self.storage, label_for_ver)
.await?;
non_existence_of_future_marker_proofs.push(non_existence_of_ver);
future_marker_vrf_proofs.push(
self.vrf
.get_label_proof::<TC>(akd_label, VersionFreshness::Fresh, ver)
.await?
.to_bytes()
.to_vec(),
);
}
let root_hash = EpochHash(
current_epoch,
current_azks.get_root_hash::<TC, _>(&self.storage).await?,
);
Ok((
HistoryProof {
update_proofs,
until_marker_vrf_proofs,
non_existence_until_marker_proofs,
future_marker_vrf_proofs,
non_existence_of_future_marker_proofs,
},
root_hash,
))
}
/// Poll for changes in the epoch number of the AZKS struct
/// stored in the storage layer. If an epoch change is detected,
/// the object cache (if present) is flushed immediately so
/// that new objects are retrieved from the storage layer against
/// the "latest" epoch. There is a "special" flow in the storage layer
/// to do a storage-layer retrieval which ignores the cache
pub async fn poll_for_azks_changes(
&self,
period: tokio::time::Duration,
change_detected: Option<tokio::sync::mpsc::Sender<()>>,
) -> Result<(), AkdError> {
// Retrieve the same AZKS that all the other calls see (i.e. the version that could be cached
// at this point). We'll compare this via an uncached call when a change is notified
let mut last = Directory::<TC, S, V>::get_azks_from_storage(&self.storage, false).await?;
loop {
// loop forever polling for changes
tokio::time::sleep(period).await;
let latest = Directory::<TC, S, V>::get_azks_from_storage(&self.storage, true).await?;
if latest.latest_epoch > last.latest_epoch {
{
// acquire a singleton lock prior to flushing the cache to assert that no
// cache accesses are underway (i.e. publish/proof generations/etc)
let _guard = self.cache_lock.write().await;
// flush the cache in its entirety
self.storage.flush_cache().await;
// re-fetch the azks to load it into cache so when we release the cache lock
// others will see the new AZKS loaded up and ready
last =
Directory::<TC, S, V>::get_azks_from_storage(&self.storage, false).await?;
// notify change occurred
if let Some(channel) = &change_detected {
channel.send(()).await.map_err(|send_err| {
AkdError::Storage(StorageError::Connection(format!(
"Tokio MPSC sender failed to publish notification with error {send_err:?}"
)))
})?;
}
// drop the guard
}
}
}
#[allow(unreachable_code)]
Ok(())
}
/// Returns an [AppendOnlyProof] for the leaves inserted into the underlying tree between
/// the epochs `audit_start_ep` and `audit_end_ep`.
pub async fn audit(
&self,
audit_start_ep: u64,
audit_end_ep: u64,
) -> Result<AppendOnlyProof, AkdError> {
// The guard will be dropped at the end of the proof generation
let _guard = self.cache_lock.read().await;
let current_azks = self.retrieve_azks().await?;
let current_epoch = current_azks.get_latest_epoch();
if audit_start_ep >= audit_end_ep {
Err(AkdError::Directory(DirectoryError::InvalidEpoch(format!(
"Start epoch {audit_start_ep} is greater than or equal the end epoch {audit_end_ep}"
))))
} else if current_epoch < audit_end_ep {
Err(AkdError::Directory(DirectoryError::InvalidEpoch(format!(
"End epoch {audit_end_ep} is greater than the current epoch {current_epoch}"
))))
} else {
self.storage.disable_cache_cleaning();
let result = current_azks
.get_append_only_proof::<TC, _>(&self.storage, audit_start_ep, audit_end_ep)
.await;
self.storage.enable_cache_cleaning();
result
}
}
/// Retrieves the [Azks]
pub(crate) async fn retrieve_azks(&self) -> Result<Azks, crate::errors::AkdError> {
Directory::<TC, S, V>::get_azks_from_storage(&self.storage, false).await
}
async fn get_azks_from_storage(
storage: &StorageManager<S>,
ignore_cache: bool,
) -> Result<Azks, crate::errors::AkdError> {
let got = if ignore_cache {
storage
.get_direct::<Azks>(&crate::append_only_zks::DEFAULT_AZKS_KEY)
.await?
} else {
storage
.get::<Azks>(&crate::append_only_zks::DEFAULT_AZKS_KEY)
.await?
};
match got {
DbRecord::Azks(azks) => Ok(azks),
_ => {
error!("No AZKS can be found. You should re-initialize the directory to create a new one");
Err(AkdError::Storage(StorageError::NotFound(
"AZKS not found".to_string(),
)))
}
}
}
/// HELPERS ///
/// Use this function to retrieve the [VRFPublicKey] for this AKD.
pub async fn get_public_key(&self) -> Result<VRFPublicKey, AkdError> {
Ok(self.vrf.get_vrf_public_key().await?)
}
async fn create_single_update_proof(
&self,
akd_label: &AkdLabel,
user_state: &ValueState,
) -> Result<UpdateProof, AkdError> {
let epoch = user_state.epoch;
let value = &user_state.value;
let version = user_state.version;
let label_at_ep = self
.vrf
.get_node_label::<TC>(akd_label, VersionFreshness::Fresh, version)
.await?;
let current_azks = self.retrieve_azks().await?;
let existence_vrf = self
.vrf
.get_label_proof::<TC>(akd_label, VersionFreshness::Fresh, version)
.await?;
let existence_vrf_proof = existence_vrf.to_bytes().to_vec();
let existence_label = self.vrf.get_node_label_from_vrf_proof(existence_vrf).await;
let existence_proof = current_azks
.get_membership_proof::<TC, _>(&self.storage, label_at_ep)
.await?;
let mut previous_version_proof = Option::None;
let mut previous_version_vrf_proof = Option::None;
if version > 1 {
let prev_label_at_ep = self
.vrf
.get_node_label::<TC>(akd_label, VersionFreshness::Stale, version - 1)
.await?;
previous_version_proof = Option::Some(
current_azks
.get_membership_proof::<TC, _>(&self.storage, prev_label_at_ep)
.await?,
);
previous_version_vrf_proof = Option::Some(
self.vrf
.get_label_proof::<TC>(akd_label, VersionFreshness::Stale, version - 1)
.await?
.to_bytes()
.to_vec(),
);
}
let commitment_key = self.derive_commitment_key().await?;
let commitment_nonce =
TC::get_commitment_nonce(&commitment_key, &existence_label, version, value).to_vec();
Ok(UpdateProof {
epoch,
version,
value: value.clone(),
existence_vrf_proof,
existence_proof,
previous_version_vrf_proof,
previous_version_proof,
commitment_nonce,
})
}
/// Gets the root hash at the current epoch.
pub async fn get_epoch_hash(&self) -> Result<EpochHash, AkdError> {
let current_azks = self.retrieve_azks().await?;
let latest_epoch = current_azks.get_latest_epoch();
let root_hash = current_azks.get_root_hash::<TC, _>(&self.storage).await?;
Ok(EpochHash(latest_epoch, root_hash))
}
// We simply hash the VRF private key to derive the commitment key
async fn derive_commitment_key(&self) -> Result<Digest, AkdError> {
let raw_key = self.vrf.retrieve().await?;
let commitment_key = TC::hash(&raw_key);
Ok(commitment_key)
}
}
/// A thin newtype which offers read-only interactivity with a [Directory].
#[derive(Clone)]
pub struct ReadOnlyDirectory<TC, S, V>(Directory<TC, S, V>)
where
TC: Configuration,
S: Database + Sync + Send,
V: VRFKeyStorage;
impl<TC, S, V> ReadOnlyDirectory<TC, S, V>
where
TC: Configuration,
S: Database + 'static,
V: VRFKeyStorage,
{
/// Constructs a new instance of [ReadOnlyDirectory]. In the event that an [Azks]
/// does not exist in the storage, or we're unable to retrieve it from storage, then
/// a [DirectoryError] will be returned.
pub async fn new(storage: StorageManager<S>, vrf: V) -> Result<Self, AkdError> {
let azks = Directory::<TC, S, V>::get_azks_from_storage(&storage, false).await;
if azks.is_err() {
return Err(AkdError::Directory(DirectoryError::ReadOnlyDirectory(
format!(
"Cannot start directory in read-only mode when AZKS is missing, error: {:?}",
azks.err().take()
),
)));
}
Ok(Self(Directory {
storage,
cache_lock: Arc::new(RwLock::new(())),
vrf,
tc: PhantomData,
}))
}
/// Read-only access to [Directory::lookup](Directory::lookup).
pub async fn lookup(&self, uname: AkdLabel) -> Result<(LookupProof, EpochHash), AkdError> {
self.0.lookup(uname).await
}
/// Read-only access to [Directory::batch_lookup](Directory::batch_lookup).
pub async fn batch_lookup(
&self,
unames: &[AkdLabel],
) -> Result<(Vec<LookupProof>, EpochHash), AkdError> {
self.0.batch_lookup(unames).await
}
/// Read-only access to [Directory::key_history](Directory::key_history).
pub async fn key_history(
&self,
uname: &AkdLabel,
params: HistoryParams,
) -> Result<(HistoryProof, EpochHash), AkdError> {
self.0.key_history(uname, params).await
}
/// Read-only access to [Directory::poll_for_azks_changes](Directory::poll_for_azks_changes).
pub async fn poll_for_azks_changes(
&self,
period: tokio::time::Duration,
change_detected: Option<tokio::sync::mpsc::Sender<()>>,
) -> Result<(), AkdError> {
self.0.poll_for_azks_changes(period, change_detected).await
}
/// Read-only access to [Directory::audit](Directory::audit).
pub async fn audit(
&self,
audit_start_ep: u64,
audit_end_ep: u64,
) -> Result<AppendOnlyProof, AkdError> {
self.0.audit(audit_start_ep, audit_end_ep).await
}
/// Read-only access to [Directory::get_epoch_hash].
pub async fn get_epoch_hash(&self) -> Result<EpochHash, AkdError> {
self.0.get_epoch_hash().await
}
/// Read-only access to [Directory::get_public_key](Directory::get_public_key).
pub async fn get_public_key(&self) -> Result<VRFPublicKey, AkdError> {
self.0.get_public_key().await
}
}
/// The parameters that dictate how much of the history proof to return to the consumer
/// (either a complete history, or some limited form).
#[derive(Copy, Clone)]
pub enum HistoryParams {
/// Returns a complete history for a label
Complete,
/// Returns up to the most recent N updates for a label
MostRecent(usize),
/// Returns all updates since a specified epoch (inclusive)
SinceEpoch(u64),
}
impl Default for HistoryParams {
/// By default, we return a complete history
fn default() -> Self {
Self::Complete
}
}
/// Helpers
pub(crate) fn get_marker_version(version: u64) -> u64 {
(64 - version.leading_zeros() - 1).into()
}
/// Helpers for testing
/// This enum is meant to insert corruptions into a malicious publish function.
#[derive(Debug, Clone)]
pub enum PublishCorruption {
/// Indicates to the malicious publish function to not mark a stale version
UnmarkedStaleVersion(AkdLabel),
/// Indicates to the malicious publish to mark a certain version for a username as stale.
MarkVersionStale(AkdLabel, u64),
}
#[cfg(test)]
impl<TC: Configuration, S: Database + 'static, V: VRFKeyStorage> Directory<TC, S, V> {
/// Updates the directory to include the updated key-value pairs with possible issues.
pub(crate) async fn publish_malicious_update(
&self,
updates: Vec<(AkdLabel, AkdValue)>,
corruption: PublishCorruption,
) -> Result<EpochHash, AkdError> {
// The guard will be dropped at the end of the publish
let _guard = self.cache_lock.read().await;
let mut update_set = Vec::<AzksElement>::new();
if let PublishCorruption::MarkVersionStale(ref akd_label, version_number) = corruption {
// In the malicious case, sometimes the server may not mark the old version stale immediately.
// If this is the case, it may want to do this marking at a later time.
let stale_label = self
.vrf
.get_node_label::<TC>(akd_label, VersionFreshness::Stale, version_number)
.await?;
let stale_value_to_add = TC::stale_azks_value();
update_set.push(AzksElement {
label: stale_label,
value: stale_value_to_add,
})
};
let mut user_data_update_set = Vec::<ValueState>::new();
let mut current_azks = self.retrieve_azks().await?;
let current_epoch = current_azks.get_latest_epoch();
let next_epoch = current_epoch + 1;
let mut keys: Vec<AkdLabel> = updates
.iter()
.map(|(akd_label, _val)| akd_label.clone())
.collect();
// sort the keys, as inserting in primary-key order is more efficient for MySQL
keys.sort();
// we're only using the maximum "version" of the user's state at the last epoch
// they were seen in the directory. Therefore we've minimized the call to only
// return a hashmap of AkdLabel => u64 and not retrieving the other data which is not
// read (i.e. the actual _data_ payload).
let all_user_versions_retrieved = self
.storage
.get_user_state_versions(&keys, ValueStateRetrievalFlag::LeqEpoch(current_epoch))
.await?;
info!(
"Retrieved {} previous user versions of {} requested",
all_user_versions_retrieved.len(),
keys.len()
);
let commitment_key = self.derive_commitment_key().await?;
for (akd_label, val) in updates {
match all_user_versions_retrieved.get(&akd_label) {
None => {
// no data found for the user
let latest_version = 1;
let label = self
.vrf
.get_node_label::<TC>(&akd_label, VersionFreshness::Fresh, latest_version)
.await?;
let value_to_add =
TC::compute_fresh_azks_value(&commitment_key, &label, latest_version, &val);
update_set.push(AzksElement {
label,
value: value_to_add,
});
let latest_state =
ValueState::new(akd_label, val, latest_version, label, next_epoch);
user_data_update_set.push(latest_state);
}
Some((_, previous_value)) if val == *previous_value => {
// skip this version because the user is trying to re-publish the already most recent value
// Issue #197: https://github.com/facebook/akd/issues/197
}
Some((previous_version, _)) => {
// Data found for the given user
let latest_version = *previous_version + 1;
let stale_label = self
.vrf
.get_node_label::<TC>(
&akd_label,
VersionFreshness::Stale,
*previous_version,
)
.await?;
let fresh_label = self
.vrf
.get_node_label::<TC>(&akd_label, VersionFreshness::Fresh, latest_version)
.await?;
let stale_value_to_add = TC::stale_azks_value();
let fresh_value_to_add = TC::compute_fresh_azks_value(
&commitment_key,
&fresh_label,
latest_version,
&val,
);
match &corruption {
// Some malicious server might not want to mark an old and compromised key as stale.
// Thus, you only push the key if either the corruption is for some other username,
// or the corruption is not of the type that asks you to delay marking a stale value correctly.
PublishCorruption::UnmarkedStaleVersion(target_akd_label) => {
if *target_akd_label != akd_label {
update_set.push(AzksElement {
label: stale_label,
value: stale_value_to_add,
})
}
}
_ => update_set.push(AzksElement {
label: stale_label,
value: stale_value_to_add,
}),
};
update_set.push(AzksElement {
label: fresh_label,
value: fresh_value_to_add,
});
let new_state =
ValueState::new(akd_label, val, latest_version, fresh_label, next_epoch);
user_data_update_set.push(new_state);
}
}
}
let azks_element_set: Vec<AzksElement> = update_set.to_vec();
if azks_element_set.is_empty() {
info!("After filtering for duplicated user information, there is no publish which is necessary (0 updates)");
// The AZKS has not been updated/mutated at this point, so we can just return the root hash from before
let root_hash = current_azks.get_root_hash::<TC, _>(&self.storage).await?;
return Ok(EpochHash(current_epoch, root_hash));
}
if let false = self.storage.begin_transaction() {
error!("Transaction is already active");
return Err(AkdError::Storage(StorageError::Transaction(
"Transaction is already active".to_string(),
)));
}
info!("Starting database insertion");
current_azks
.batch_insert_nodes::<TC, _>(&self.storage, azks_element_set, InsertMode::Directory)
.await?;
// batch all the inserts into a single transactional write to storage
let mut updates = vec![DbRecord::Azks(current_azks.clone())];
for update in user_data_update_set.into_iter() {
updates.push(DbRecord::ValueState(update));
}
self.storage.batch_set(updates).await?;
// now commit the transaction
if let Err(err) = self.storage.commit_transaction().await {
// ignore any rollback error(s)
let _ = self.storage.rollback_transaction();
return Err(AkdError::Storage(err));
}
let root_hash = current_azks
.get_root_hash_safe::<TC, _>(&self.storage, next_epoch)
.await?;
Ok(EpochHash(next_epoch, root_hash))
// At the moment the tree root is not being written anywhere. Eventually we
// want to change this to call a write operation to post to a blockchain or some such thing
}
}