From 996c7c8e882b3629d1a06a41536bdda6955c324b Mon Sep 17 00:00:00 2001 From: Jim Posen Date: Thu, 1 Jun 2017 11:59:19 -0700 Subject: BIP 157 & 158: client-side block filtering. --- README.mediawiki | 14 ++ bip-0157.mediawiki | 471 +++++++++++++++++++++++++++++++++++++++++++++++++++++ bip-0158.mediawiki | 431 ++++++++++++++++++++++++++++++++++++++++++++++++ 3 files changed, 916 insertions(+) create mode 100644 bip-0157.mediawiki create mode 100644 bip-0158.mediawiki diff --git a/README.mediawiki b/README.mediawiki index 723c65a..bcc1e3f 100644 --- a/README.mediawiki +++ b/README.mediawiki @@ -736,6 +736,20 @@ Those proposing changes should consider that ultimately consent may rest with th | Standard | Draft |- +| [[bip-0157.mediawiki|157]] +| Peer Services +| Client Side Block Filtering +| Olaoluwa Osuntokun, Alex Akselrod, Jim Posen +| Standard +| Draft +|- +| [[bip-0158.mediawiki|158]] +| Peer Services +| Compact Block Filters for Light Clients +| Olaoluwa Osuntokun, Alex Akselrod +| Standard +| Draft +|- | [[bip-0159.mediawiki|159]] | Peer Services | NODE_NETWORK_LIMITED service bit diff --git a/bip-0157.mediawiki b/bip-0157.mediawiki new file mode 100644 index 0000000..61ffc7e --- /dev/null +++ b/bip-0157.mediawiki @@ -0,0 +1,471 @@ +
+  BIP: 157
+  Layer: Peer Services
+  Title: Client Side Block Filtering
+  Author: Olaoluwa Osuntokun 
+          Alex Akselrod 
+          Jim Posen 
+  Comments-Summary: None yet
+  Comments-URI: https://github.com/bitcoin/bips/wiki/Comments:BIP-0157
+  Status: Draft
+  Type: Standards Track
+  Created: 2017-05-24
+  License: CC0-1.0
+
+ + +== Abstract == + +This BIP describes a new light client protocol in Bitcoin that improves upon +currently available options. The standard light client protocol in use today, +defined in BIP +37https://github.com/bitcoin/bips/blob/master/bip-0037.mediawiki, has +known flaws that weaken the security and privacy of clients and allow +denial-of-service attack vectors on full +nodeshttps://lists.linuxfoundation.org/pipermail/bitcoin-dev/2016-May/012636.html. +The new protocol overcomes these issues by allowing light clients to obtain +compact probabilistic filters of block content from full nodes and download full +blocks if the filter matches relevant data. + +New P2P messages empower light clients to securely sync the blockchain without +relying on a trusted source. This BIP also defines a filter header, which serves +as a commitment to all filters for previous blocks and provides the ability to +efficiently detect malicious or faulty peers serving invalid filters. The +resulting protocol guarantees that light clients with at least one honest peer +are able to identify the correct block filters. + +== Motivation == + +Bitcoin light clients allow applications to read relevant transactions from the +blockchain without incurring the full cost of downloading and validating all +data. Such applications seek to simultaneously minimize the trust in peers and +the amount of bandwidth, storage space, and computation required. They achieve +this by downloading all block headers, verifying the proofs of work, and +following the longest proof-of-work chain. Since block headers are a fixed +80-bytes and are generated every 10 minutes on average, the bandwidth required +to sync the block headers is minimal. Light clients then download only the +blockchain data relevant to them directly from peers and validate inclusion in +the header chain. Though clients do not check the validity of all blocks in the +longest proof-of-work chain, they rely on miner incentives for security. + +BIP 37 is currently the most widely used light client execution mode for +Bitcoin. With BIP 37, a client sends a Bloom filter it wants to watch to a full +node peer, then receives notifications for each new transaction or block that +matches the filter. The client then requests relevant transactions from the peer +along with Merkle proofs of inclusion in the blocks containing them, which are +verified against the block headers. The Bloom filters match data such as client +addresses and unspent outputs, and the filter size must be carefully tuned to +balance the false positive rate with the amount of information leaked to peer. It +has been shown, however, that most implementations available offer virtually +''zero privacy'' to wallets and other +applicationshttps://eprint.iacr.org/2014/763.pdfhttps://jonasnick.github.io/blog/2015/02/12/privacy-in-bitcoinj/. +Additionally, malicious full nodes serving light clients can omit critical data +with little risk of detection, which is unacceptable for some applications +(such as Lightning Network clients) that must respond to certain on-chain +events. Finally, honest nodes servicing BIP 37 light clients may incur +significant I/O and CPU resource usage due to maliciously crafted Bloom filters, +creating a denial-of-service (DoS) vector and disincentizing node operators from +supporting the +protocolhttps://github.com/bitcoin/bips/blob/master/bip-0111.mediawiki. + +The alternative detailed in this document can be seen as the opposite of BIP 37: +instead of the client sending a filter to a full node peer, full nodes generate +deterministic filters on block data that are served to the client. A light +client can then download an entire block if the filter matches the data it is +watching for. Since filters are deterministic, they only need to be constructed +once and stored on disk, whenever a new block is connected to the chain. This +keeps the computation required to serve filters minimal, and eliminates the I/O +asymmetry that makes BIP 37 enabled nodes vulnerable. Clients also get better +assurance of seeing all relevant transactions because they can check the +validity of filters received from peers more easily than they can check +completeness of filtered blocks. Finally, client privacy is improved because +blocks can be downloaded from ''any source'', so that no one peer gets complete +information on the data required by a client. Extremely privacy conscious light +clients may opt to anonymously fetch blocks using advanced techniques such a +Private Information +Retrievalhttps://en.wikipedia.org/wiki/Private_information_retrieval. + +== Definitions == + +[]byte represents a vector of bytes. + +[N]byte represents a fixed-size byte array with length N. + +''CompactSize'' is a compact encoding of unsigned integers used in the Bitcoin +P2P protocol. + +''double-SHA256'' is a hash algorithm defined by two invocations of SHA-256: +double-SHA256(x) = SHA256(SHA256(x)). + +The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", +"SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be +interpreted as described in RFC 2119. + +== Specification == + +=== Filter Types === + +For the sake of future extensibility and reducing filter sizes, there are +multiple ''filter types'' that determine which data is included in a block +filter as well as the method of filter construction/querying. In this model, +full nodes generate one filter per block per filter type supported. + +Each type is identified by a one byte code, and specifies the contents and +serialization format of the filter. A full node MAY signal support for +particular filter types using service bits. The initial filter types are defined +separately in [[bip-0158.mediawiki|BIP 158]], and one service bit is allocated +to signal support for them. + +=== Filter Headers === + +This proposal draws inspiration from the headers-first mechanism that Bitcoin +nodes use to sync the block +chainhttps://bitcoin.org/en/developer-guide#headers-first. Similar to +how block headers have a Merkle commitment to all transaction data in the block, +we define filter headers that have commitments to the block filters. Also like +block headers, filter headers each have a commitment to the preceding one. +Before downloading the block filters themselves, a light client can download all +filter headers for the current block chain and use them to verify the +authenticity of the filters. If the filter header chains differ between multiple +peers, the client can identify the point where they diverge, then download the +full block and compute the correct filter, thus identifying which peer is +faulty. + +The canonical hash of a block filter is the double-SHA256 of the serialized +filter. Filter headers are 32-byte hashes derived for each block filter. They +are computed as the double-SHA256 of the concatenation of the filter hash with +the previous filter header. The previous filter header used to calculate that of +the genesis block is defined to be the 32-byte array of 0's. + +=== New Messages === + +==== getcfilters ==== +getcfilters is used to request the compact filters of a particular +type for a particular range of blocks. The message contains the following +fields: + +{| class="wikitable" +! Field Name +! Data Type +! Byte Size +! Description +|- +| FilterType +| byte +| 1 +| Filter type for which headers are requested +|- +| StartHeight +| uint32 +| 4 +| The height of the first block in the requested range +|- +| StopHash +| [32]byte +| 32 +| The hash of the last block in the requested range +|} + +# Nodes SHOULD NOT send getcfilters unless the peer has signaled support for this filter type. Nodes receiving getcfilters with an unsupported filter type SHOULD NOT respond. +# StopHash MUST be known to belong to a block accepted by the receiving peer. This is the case if the peer had previously sent a headers or inv message with that block or any descendents. A node that receives getcfilters with an unknown StopHash SHOULD NOT respond. +# The height of the block with hash StopHash MUST be greater than or equal to StartHeight, and the difference MUST be strictly less than 1,000. +# The receiving node MUST respond to valid requests by sending one cfilter message for each block in the requested range, sequentially in order by block height. + +==== cfilter ==== +cfilter is sent in response to getcfilters, one for +each block in the requested range. The message contains the following fields: + +{| class="wikitable" +! Field Name +! Data Type +! Byte Size +! Description +|- +| FilterType +| byte +| 1 +| Byte identifying the type of filter being returned +|- +| BlockHash +| [32]byte +| 32 +| Block hash of the Bitcoin block for which the filter is being returned +|- +| NumFilterBytes +| CompactSize +| 1-5 +| A variable length integer representing the size of the filter in the following field +|- +| FilterBytes +| []byte +| NumFilterBytes +| The serialized compact filter for this block +|} + +# The FilterType SHOULD match the field in the getcfilters request, and BlockHash must correspond to a block that is an ancestor of StopHash with height greater than or equal to StartHeight. + +==== getcfheaders ==== +getcfheaders is used to request verifiable filter headers for a +range of blocks. The message contains the following fields: + +{| class="wikitable" +! Field Name +! Data Type +! Byte Size +! Description +|- +| FilterType +| byte +| 1 +| Filter type for which headers are requested +|- +| StartHeight +| uint32 +| 4 +| The height of the first block in the requested range +|- +| StopHash +| [32]byte +| 32 +| The hash of the last block in the requested range +|} + +# Nodes SHOULD NOT send getcfheaders unless the peer has signaled support for this filter type. Nodes receiving getcfheaders with an unsupported filter type SHOULD NOT respond. +# StopHash MUST be known to belong to a block accepted by the receiving peer. This is the case if the peer had previously sent a headers or inv message with that block or any descendents. A node that receives getcfheaders with an unknown StopHash SHOULD NOT respond. +# The height of the block with hash StopHash MUST be greater than or equal to StartHeight, and the difference MUST be strictly less than 2,000. + +==== cfheaders ==== +cfheaders is sent in response to getcfheaders. Instead +of including the filter headers themselves, the response includes one filter +header and a sequence of filter hashes, from which the headers can be derived. +This has the benefit that the client can verify the binding links between the +headers. The message contains the following fields: + +{| class="wikitable" +! Field Name +! Data Type +! Byte Size +! Description +|- +| FilterType +| byte +| 1 +| Filter type for which hashes are requested +|- +| StopHash +| [32]byte +| 32 +| The hash of the last block in the requested range +|- +| PreviousFilterHeader +| [32]byte +| 32 +| The filter header preceding the first block in the requested range +|- +| FilterHashesLength +| CompactSize +| 1-3 +| The length of the following vector of filter hashes +|- +| FilterHashes +| [][32]byte +| FilterHashesLength * 32 +| The filter hashes for each block in the requested range +|} + +# The FilterType and StopHash SHOULD match the fields in the getcfheaders request. +# FilterHashesLength MUST NOT be greater than 2,000. +# FilterHashes MUST have one entry for each block on the chain terminating with tip StopHash, starting with the block at height StartHeight. The entries MUST be the filter hashes of the given type for each block in that range, in ascending order by height. +# PreviousFilterHeader MUST be set to the previous filter header of first block in the requested range. + +==== getcfcheckpt ==== +getcfcheckpt is used to request filter headers at evenly spaced +intervals over a range of blocks. Clients may use filter hashes from +getcfheaders to connect these checkpoints, as is described in the +[[#client-operation|Client Operation]] section below. The +getcfcheckpt message contains the following fields: + +{| class="wikitable" +! Field Name +! Data Type +! Byte Size +! Description +|- +| FilterType +| byte +| 1 +| Filter type for which headers are requested +|- +| StopHash +| [32]byte +| 32 +| The hash of the last block in the chain that headers are requested for +|} + +# Nodes SHOULD NOT send getcfcheckpt unless the peer has signaled support for this filter type. Nodes receiving getcfcheckpt with an unsupported filter type SHOULD NOT respond. +# StopHash MUST be known to belong to a block accepted by the receiving peer. This is the case if the peer had previously sent a headers or inv message with any descendent blocks. A node that receives getcfcheckpt with an unknown StopHash SHOULD NOT respond. + +==== cfcheckpt ==== +cfcheckpt is sent in response to getcfcheckpt. The +filter headers included are the set of all filter headers on the requested chain +where the height is a positive multiple of 1,000. The message contains the +following fields: + +{| class="wikitable" +! Field Name +! Data Type +! Byte Size +! Description +|- +| FilterType +| byte +| 1 +| Filter type for which headers are requested +|- +| StopHash +| [32]byte +| 32 +| The hash of the last block in the chain that headers are requested for +|- +| FilterHeadersLength +| CompactSize +| 1-3 +| The length of the following vector of filter headers +|- +| FilterHeaders +| [][32]byte +| FilterHeadersLength * 32 +| The filter headers at intervals of 1,000 +|} + +# The FilterType and StopHash SHOULD match the fields in the getcfcheckpt request. +# FilterHeaders MUST have exactly one entry for each block on the chain terminating in StopHash, where the block height is a multiple of 1,000 greater than 0. The entries MUST be the filter headers of the given type for each such block, in ascending order by height. + +=== Node Operation === + +Full nodes MAY opt to support this BIP and generate filters for any of the +specified filter types. Such nodes SHOULD treat the filters as an additional +index of the blockchain. For each new block that is connected to the main chain, +nodes SHOULD generate filters for all supported types and persist them. Nodes +that are missing filters and are already synced with the blockchain SHOULD +reindex the chain upon start-up, constructing filters for each block from +genesis to the current tip. They also SHOULD keep every checkpoint header in +memory, so that getcfcheckpt requests do not result in many +random-access disk reads. + +Nodes SHOULD NOT generate filters dynamically on request, as malicious peers may +be able to perform DoS attacks by requesting small filters derived from large +blocks. This would require an asymmetical amount of I/O on the node to compute +and serve, similar to attacks against BIP 37 enabled nodes noted in BIP 111. + +Nodes MAY prune block data after generating and storing all filters for a block. + +=== Client Operation === + +This section provides recommendations for light clients to download filters with +maximal security. + +Clients SHOULD first sync the entire block header chain from peers using the +standard headers-first syncing mechanism before downloading any block filters or +filter headers. Clients configured with trusted checkpoints MAY only sync +headers started from the last checkpoint. Clients SHOULD disconnect any outbound +peers whose best chain has significantly less work than the known longest +proof-of-work chain. + +Once a client's block headers are in sync, it SHOULD download and verify filter +headers for all blocks and filter types that it might later download. The client +SHOULD send getcfheaders messages to peers and derive and store the +filter headers for each block. The client MAY first fetch headers at evenly +spaced intervals of 1,000 by sending getcfcheckpt. The header +checkpoints allow the client to download filter headers for different intervals +from multiple peers in parallel, verifying each range of 1,000 headers against +the checkpoints. + +Unless securely connected to a trusted peer that is serving filter headers, the +client SHOULD connect to multiple outbound peers that support each filter type +to mitigate the risk of downloading incorrect headers. If the client receives +conflicting filter headers from different peers for any block and filter type, +it SHOULD interrogate them to determine which is faulty. The client SHOULD use +getcfheaders and/or getcfcheckpt to first identify +the first filter headers that the peers disagree on. The client then SHOULD +download the full block from any peer and derive the correct filter and filter +header. The client SHOULD ban any peers that sent a filter header that does not +match the computed one. + +Once the client has downloaded and verified all filter headers needed, ''and'' +no outbound peers have sent conflicting headers, the client can download the +actual block filters it needs. The client MAY backfill filter headers before the +first verified one at this point if it only downloaded them starting at a later +point. Clients SHOULD persist the verified filter headers for last 100 blocks in +the chain (or whatever finality depth is desired), to compare against headers +received from new peers after restart. They MAY store more filter headers to +avoid redownloading them if a rescan is later necessary. + +Starting from the first block in the desired range, the client now MAY download +the filters. The client SHOULD test that each filter links to its corresponding +filter header and ban peers that send incorrect filters. The client MAY download +multiple filters at once to increase throughput, though it SHOULD test the +filters sequentially. The client MAY check if a filter is empty before +requesting it by checking if the filter header commits to the hash of the empty +filter, saving a round trip if that is the case. + +Each time a new valid block header is received, the client SHOULD request the +corresponding filter headers from all eligible peers. If two peers send +conflicting filter headers, the client should interrogate them as described +above and ban any peers that send an invalid header. + +If a client is fetching full blocks from the P2P network, they SHOULD be downloaded +from outbound peers at random to mitigate privacy loss due to transaction +intersection analysis. Note that blocks may be downloaded from peers that do not +support this BIP. + +== Rationale == + +The filter headers and checkpoints messages are defined to help clients identify +the correct filter for a block when connected to peers sending conflicting +information. An alternative solution is to require Bitcoin blocks to include +commitments to derived block filters, so light clients can verify authenticity +given block headers and some additional witness data. This would require a +network-wide change to the Bitcoin consensus rules, however, whereas this +document proposes a solution purely at the P2P layer. + +The constant interval of 1,000 blocks between checkpoints was chosen so that, +given the current chain height and rate of growth, the size of a +cfcheckpt message is not drastically from a +cfheaders between two checkpoints. Also, 1,000 is a nice round +number, at least to those of us who think in decimal. + +== Compatibility == + +This light client mode is not compatible with current node deployments and +requires support for the new P2P messages. The node implementation of this +proposal is not incompatible with the current P2P network rules (ie. doesn't +affect network topology of full nodes). Light clients may adopt protocols based +on this as an alternative to the existing BIP 37. Adoption of this BIP may +result in reduced network support for BIP 37. + +== Acknowledgments == + +We would like to thank bfd (from the bitcoin-dev mailing list) for bringing the +basis of this BIP to our attention, Joseph Poon for suggesting the filter header +chain scheme, and Pedro Martelletto for writing the initial indexing code for +btcd. + +We would also like to thank Dave Collins, JJ Jeffrey, Eric Lombrozo, and Matt +Corallo for useful discussions. + +== Reference Implementation == + +Light client: [https://github.com/lightninglabs/neutrino] + +Full-node indexing: https://github.com/Roasbeef/btcd/tree/segwit-cbf + +Golomb-Rice Coded sets: https://github.com/Roasbeef/btcutil/tree/gcs/gcs + +== References == + + + +== Copyright == + +This document is licensed under the Creative Commons CC0 1.0 Universal lisence. diff --git a/bip-0158.mediawiki b/bip-0158.mediawiki new file mode 100644 index 0000000..dc28154 --- /dev/null +++ b/bip-0158.mediawiki @@ -0,0 +1,431 @@ +
+  BIP: 158
+  Layer: Peer Services
+  Title: Compact Block Filters for Light Clients
+  Author: Olaoluwa Osuntokun 
+          Alex Akselrod 
+  Comments-Summary: None yet
+  Comments-URI: https://github.com/bitcoin/bips/wiki/Comments:BIP-0158
+  Status: Draft
+  Type: Standards Track
+  Created: 2017-05-24
+  License: CC0-1.0
+
+ + +== Abstract == + +This BIP describes a structure for compact filters on block data, for use in the +BIP 157 light client protocolbip-0157.mediawiki. The filter +construction proposed is an alternative to Bloom filters, as used in BIP 37, +that minimizes filter size by using Golomb-Rice coding for compression. This +document specifies two initial types of filters based on this construction that +enables basic wallets and applications with more advanced smart contracts. + +== Motivation == + +[[bip-0157.mediawiki|BIP 157]] defines a light client protocol based on +deterministic filters of block content. The filters are designed to +minimize the expected bandwidth consumed by light clients, downloading filters +and full blocks. This document defines two initial filter types, ''basic'' and +''extended'', to provide support for advanced applications while reducing the +filter size for regular wallets. + +== Definitions == + +[]byte represents a vector of bytes. + +[N]byte represents a fixed-size byte array with length N. + +''CompactSize'' is a compact encoding of unsigned integers used in the Bitcoin +P2P protocol. + +''Data pushes'' are byte vectors pushed to the stack according to the rules of +Bitcoin script. + +''Bit streams'' are readable and writable streams of individual bits. The +following functions are used in the pseudocode in this document: +* new_bit_stream instantiates a new writable bit stream +* new_bit_stream(vector) instantiates a new bit stream reading data from vector +* write_bit(stream, b) appends the bit b to the end of the stream +* read_bit(stream) reads the next available bit from the stream +* write_bits_big_endian(stream, n, k) appends the k least significant bits of integer n to the end of the stream in big-endian bit order +* read_bits_big_endian(stream, k) reads the next available +* k bits from the stream and interprets them as the least significant bits of a big-endian integer + +The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", +"SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be +interpreted as described in RFC 2119. + +== Specification == + +=== Golomb-Coded Sets === + +For each block, compact filters are derived containing sets of items associated +with the block (eg. addresses sent to, outpoints spent, etc.). A set of such +data objects is compressed into a probabilistic structure called a +''Golomb-coded set'' (GCS), which matches all items in the set with probability +1, and matches other items with probability 2^(-P) for some integer +parameter P. + +At a high level, a GCS is constructed from a set of N items by: +# hashing all items to 64-bit integers in the range [0, N * 2^P) +# sorting the hashed values in ascending order +# computing the differences between each value and the previous one +# writing the differences sequentially, compressed with Golomb-Rice coding + +The following sections describe each step in greater detail. + +==== Hashing Data Objects ==== + +The first step in the filter construction is hashing the variable-sized raw +items in the set to the range [0, F), where F = N * +2^P. Set membership queries against the hash outputs will have a false +positive rate of 2^(-P). To avoid integer overflow, the number of +items N MUST be <2^32 and P MUST be <=32. + +The items are first passed through the pseudorandom function ''SipHash'', which +takes a 128-bit key k and a variable-sized byte vector and produces +a uniformly random 64-bit output. Implementations of this BIP MUST use the +SipHash parameters c = 2 and d = 4. + +The 64-bit SipHash outputs are then mapped uniformly over the desired range by +multiplying with F and taking the top 64 bits of the 128-bit result. This +algorithm is a faster alternative to modulo reduction, as it avoids the +expensive division +operationhttps://lemire.me/blog/2016/06/27/a-fast-alternative-to-the-modulo-reduction/. +Note that care must be taken when implementing this reduction to ensure the +upper 64 bits of the integer multiplication are not truncated; certain +architectures and high level languages may require code that decomposes the +64-bit multiplication into four 32-bit multiplications and recombines into the +result. + +
+hash_to_range(item: []byte, F: uint64, k: [16]byte) -> uint64:
+    return (siphash(k, item) * F) >> 64
+
+hashed_set_construct(raw_items: [][]byte, P: uint, k: [16]byte) -> []uint64:
+    let N = len(raw_items)
+    let F = N << P
+
+    let set_items = []
+
+    for item in raw_items:
+        let set_value = hash_to_range(item, F, k)
+        set_items.append(set_value)
+
+    return set_items
+
+ +==== Golomb-Rice Coding ==== + +Instead of writing the items in the hashed set directly to the filter, greater +compression is achieved by only writing the differences between successive +items in sorted order. Since the items are distributed uniformly, it can be +shown that the differences resemble a geometric +distributionhttps://en.wikipedia.org/wiki/Geometric_distribution. +''Golomb-Rice'' +''coding''https://en.wikipedia.org/wiki/Golomb_coding#Rice_coding +is a technique that optimally compresses geometrically distributed values. + +With Golomb-Rice, a value is split into a quotient and remainder modulo +2^P, which are encoded separately. The quotient q is +encoded as ''unary'', with a string of q 1's followed by one 0. The +remainder r is represented in big-endian by P bits. For example, +this is a table of Golomb-Rice coded values using P=2: + +{| class="wikitable" +! n !! (q, r) !! c +|- +| 0 || (0, 0) || 0 00 +|- +| 1 || (0, 1) || 0 01 +|- +| 2 || (0, 2) || 0 10 +|- +| 3 || (0, 3) || 0 11 +|- +| 4 || (1, 0) || 10 00 +|- +| 5 || (1, 1) || 10 01 +|- +| 6 || (1, 2) || 10 10 +|- +| 7 || (1, 3) || 10 11 +|- +| 8 || (2, 0) || 110 00 +|- +| 9 || (2, 1) || 110 01 +|} + +
+golomb_encode(stream, x: uint64, P: uint):
+    let q = x >> P
+
+    while q > 0:
+        write_bit(stream, 1)
+        q--
+    write_bit(stream, 0)
+
+    write_bits_big_endian(stream, x, P)
+
+golomb_decode(stream, P: uint) -> uint64:
+    let q = 0
+    while read_bit(stream) == 1:
+        q++
+
+    let r = read_bits_big_endian(stream, P)
+
+    let x = (q << P) + r
+    return x
+
+ +==== Set Construction ==== + +A GCS is constructed from three parameters: +* L, a vector of N raw items +* P, which determines the false positive rate +* k, the 128-bit key used to randomize the SipHash outputs + +The result is a byte vector with a minimum size of N * (P + 1) +bits. + +The raw items in L are first hashed to 64-bit unsigned integers as +specified above and sorted. The differences between consecutive values, +hereafter referred to as ''deltas'', are encoded sequentially to a bit stream +with Golomb-Rice coding. Finally, the bit stream is padded with 0's to the +nearest byte boundary and serialized to the output byte vector. + +
+construct_gcs(L: [][]byte, P: uint, k: [16]byte) -> []byte:
+    let set_items = hashed_set_construct(L, P, k)
+
+    set_items.sort()
+
+    let output_stream = new_bit_stream()
+
+    let last_value = 0
+    for item in set_items:
+        let delta = item - last_value
+        golomb_encode(output_stream, delta, P)
+        last_value = item
+
+    return output_stream.bytes()
+
+ +==== Set Querying/Decompression ==== + +To check membership of an item in a compressed GCS, one must reconstruct the +hashed set members from the encoded deltas. The procedure to do so is the +reverse of the compression: deltas are decoded one by one and added to a +cumulative sum. Each intermediate sum represents a hashed value in the original +set. The queried item is hashed in the same way as the set members and compared +against the reconstructed values. Note that querying does not require the entire +decompressed set be held in memory at once. + +
+gcs_match(key: [16]byte, compressed_set: []byte, target: []byte, P: uint, N: uint) -> bool:
+    let F = N << P
+    let target_hash = hash_to_range(target, F, k)
+
+    stream = new_bit_stream(compressed_set)
+
+    let last_value = 0
+
+    loop N times:
+        let delta = golomb_decode(stream, P)
+        let set_item = last_value + delta
+
+        if set_item == target_hash:
+            return true
+
+        // Since the values in the set are sorted, terminate the search once
+        // the decoded value exceeds the target.
+        if set_item > target_hash:
+            break
+
+        last_value = set_item
+
+    return false
+
+ +Some applications may need to check for set intersection instead of membership +of a single item. This can be performed far more efficiently than checking each +item individually by leveraging the sorted structure of the compressed GCS. +First the query elements are all hashed and sorted, then compared in order +against the decompressed GCS contents. See +[[#golomb-coded-set-multi-match|Appendix B]] for pseudocode. + +=== Block Filters === + +This BIP defines two initial filter types: +* Basic (0x00) +* Extended (0x01) + +==== Contents ==== + +The basic filter is designed to contain everything that a light client needs to +sync a regular Bitcoin wallet. A basic filter MUST contain exactly the following +items for each transaction in a block: +* The outpoint of each input, except for the coinbase transaction +* Each data push in the scriptPubKey of each output, ''only if'' the scriptPubKey is parseable +* The txid of the transaction itself + +The extended filter contains extra data that is meant to enable applications +with more advanced smart contracts. An extended filter MUST contain exactly the +following items for each transaction in a block ''except the coinbase'': +* Each item within the witness stack of each input (if the input has a witness) +* Each data push in the scriptSig of each input + +Note that neither filter type interprets P2SH scripts or witness scripts to +extract data pushes from them. If necessary, future filter types may be designed +to do so. + +==== Construction ==== + +Both the basic and extended filter types are constructed as Golomb-coded sets +with the following parameters. + +The parameter P MUST be set to 20. This value was +chosen as simulations show that it minimizes the bandwidth utilized, considering +both the expected number of blocks downloaded due to false positives and the +size of the filters themselves. The code along with a demo used for the +parameter tuning can be found +[https://github.com/Roasbeef/bips/blob/83b83c78e189be898573e0bfe936dd0c9b99ecb9/gcs_light_client/gentestvectors.go here]. + +The parameter k MUST be set to the first 16 bytes of the hash of +the block for which the filter is constructed. This ensures the key is +deterministic while still varying from block to block. + +Since the value N is required to decode a GCS, a serialized GCS +includes it as a prefix, written as a CompactSize. Thus, the complete +serialization of a filter is: +* N, encoded as a CompactSize +* The bytes of the compressed filter itself + +==== Signaling ==== + +This BIP allocates a new service bit: + +{| class="wikitable" +|- +| NODE_COMPACT_FILTERS +| style="white-space: nowrap;" | 1 << 6 +| If enabled, the node MUST respond to all BIP 157 messages for filter types 0x00 and 0x01 +|} + +== Compatibility == + +This block filter construction is not incompatible with existing software, +though it requires implementation of the new filters. + +== Acknowledgments == + +We would like to thank bfd (from the bitcoin-dev mailing list) for bringing the +basis of this BIP to our attention, Greg Maxwell for pointing us in the +direction of Golomb-Rice coding and fast range optimization, and Pedro +Martelletto for writing the initial indexing code for btcd. + +We would also like to thank Dave Collins, JJ Jeffrey, and Eric Lombrozo for +useful discussions. + +== Reference Implementation == + +Light client: [https://github.com/lightninglabs/neutrino] + +Full-node indexing: https://github.com/Roasbeef/btcd/tree/segwit-cbf + +Golomb-Rice Coded sets: https://github.com/Roasbeef/btcutil/tree/gcs/gcs + +== Appendix A: Alternatives == + +A number of alternative set encodings were considered before Golomb-coded +sets were settled upon. In this appendix section, we'll list a few of the +alternatives along with our rationale for not pursuing them. + +==== Bloom Filters ==== + +Bloom Filters are perhaps the best known probabilistic data structure for +testing set membership, and were introduced into the Bitcoin protocol with BIP +37. The size of a Bloom filter is larger than the expected size of a GCS with +the same false positive rate, which is the main reason the option was rejected. + +==== Cryptographic Accumulators ==== + +Cryptographic +accumulatorshttps://en.wikipedia.org/wiki/Accumulator_(cryptography) +are a cryptographic data structures that enable (amongst other operations) a one +way membership test. One advantage of accumulators are that they are constant +size, independent of the number of elements inserted into the accumulator. +However, current constructions of cryptographic accumulators require an initial +trusted set up. Additionally, accumulators based on the Strong-RSA Assumption +require mapping set items to prime representatives in the associated group which +can be preemptively expensive. + +==== Matrix Based Probabilistic Set Data Structures ==== + +There exist data structures based on matrix solving which are even more space +efficient compared to Bloom +filtershttps://arxiv.org/pdf/0804.1845.pdf. We instead opted for our +GCS-based filters as they have a much lower implementation complexity and are +easier to understand. + +== Appendix B: Pseudocode == + +=== Golomb-Coded Set Multi-Match === + +
+gcs_match_any(key: [16]byte, compressed_set: []byte, targets: [][]byte, P: uint, N: uint) -> bool:
+    let F = N << P
+
+    // Map targets to the same range as the set hashes.
+    let target_hashes = []
+    for target in targets:
+        let target_hash = hash_to_range(target, F, k)
+        target_hashes.append(target_hash)
+
+    // Sort targets so matching can be checked in linear time.
+    target_hashes.sort()
+
+    stream = new_bit_stream(compressed_set)
+
+    let value = 0
+    let target_idx = 0
+    let target_val = target_hashes[target_idx]
+
+    loop N times:
+        let delta = golomb_decode(stream, P)
+        value += delta
+
+        inner loop:
+            if target_val == value:
+                return true
+
+            // Move on to the next set value.
+            else if target_val > value:
+                break inner loop
+
+            // Move on to the next target value.
+            else if target_val < value:
+                target_idx++
+
+                // If there are no targets left, then there are no matches.
+                if target_idx == len(targets):
+                    break outer loop
+
+                target_val = target_hashes[target_idx]
+
+    return false
+
+ +== Appendix C: Test Vectors == + +TODO: To be generated. + +== References == + + + +== Copyright == + +This document is licensed under the Creative Commons CC0 1.0 Universal lisence. -- cgit v1.2.3