Patent Application
20220284129
This invention relates to data security.
Data is everywhere nowadays, and few people in industrialized countries are not involved in some form of data transfer almost every day. Email, online purchases, bank transfers, online entertainment and news, requests for all manner of services, text messaging and even voice calls over digital networks, etc., are just a few of the seemingly countless instances of data transfer. In many cases, transfer involves some form of reproduction: Text written using one computing device (computer, smart phone, etc.) is passed in digitized form to another computing device for reproduction as an email or text message; data defining a copy of a document, or an address or reference such as a link to a web site, database entry, account, etc., is passed from one person to another; security keys are exchanged; etc. In these situations, more than one instance of some data set may, or even should, exist in more than one location at the same time, or, logically identically, more than one entity may have the ability to pass on to others the ability to hold, access, or otherwise control the data set.
In some other situations, however, only one entity at a time should be able to control further transfer of any instance of or reference to, or control over, a data set. Examples of such “single valid instance” data sets include some permissions, highly secret documents, digital cash, etc.
One problem when it comes to these situations is that there must exist some way to prove that an instance of digital information is indeed the only valid one, since, unlike physical objects, data is easy to perfectly copy. This applies as well to data sets that define permissions related to objects or digital information. One way to do this is to create a ledger that keeps track of the current authorized “owner” of the single-instance data set. This then moves the “problem” to being able to prove that the ledger itself is correct and has not been tampered with.
One form of ledger that has been proposed to solve this problem is a “blockchain”, which, in its simplest form, is sequence of data “blocks” such that each block is cryptographically linked to at least one previous block so as to make any change to previously entered blocks detectable. In some cases, the blockchain is in the form of a ledger that is distributed among several parties. Perhaps the most well-known current example of the such a distributed ledger based on a blockchain is Bitcoin.
One drawback of existing blockchain-backed ledger systems is lack of scalability. Distributed ledger technologies (DLT) such as Bitcoin by their nature require an entire blockchain to be held and managed by several disparate systems, which must then coordinate using some kind of consensus mechanism so that they all can agree on what the correct current state of the blockchain is. Moreover, permissionless DLT systems require some way to determine which entity is allowed to add to the blockchain. This leads to mechanisms such as proof-of-work, proof-of-stake, proof-of-space, etc., which introduce delay in addition to complexity. The Bitcoin system, for example, is designed not to update more frequently than about every ten minutes. Such a delay is unacceptable for many use cases where a potentially large number of data transfers need to be secured quickly. This drawback exists to varying degrees in most other blockchain solutions as well, such as Ethereum.
The problem of lack of scalability arises in almost every system that relies on a global ledger that tracks the status of every transferrable data set in the system. In systems that involve a large number of transactions per time unit, bandwidth alone is often a limiting factor, and if access to the single ledger is cut off, for example due to a simple server failure, then the whole system must often halt. One way to reduce the bandwidth demand on a single ledger host, and to increase guaranteed accessibility, is to distribute the ledger. This then causes a need to ensure that all copies of the ledger are synchronized and correct.
One consequence of this, which also reduces the ability to scale, is that each entity maintaining a copy of the ledger must be aware of the state of the ledger in many other or even all other entities that also maintain copies—if one entity changes the ledger, then other entities must be aware of this and either make the same change or agree to reject it. Distribution of a global ledger thus assumes statefulness, and also requires communication between all of either all or at least some minimum subset of the participating servers.
Bitcoin introduced a new monetary unit—Unspent Transaction Outputs (UTXOs)—now used by many cryptocurrencies. A UTXO represents a “piece” of a Bitcoin and, as its name implies, is an output of a Bitcoin transaction that represents the amount “left over” after a transaction has been completed. UTXOs are logically a set, with each transaction consuming elements from this set as transaction inputs, with the output remainder being created as a new element of the set. UTXOs can be spent only once (since they are destroyed after use), but in doing so, new UTXOs are created and can have different values and owners. This is unsuitable in many situations in which individual data units should be easily accounted for. The UTXO model also suffers from the same failure to scale as the general structure used in the Bitcoin system and similar schemes.
In the context of electronic financial transactions, a commonly used arrangement is for users to have accounts, in which transactions involve changes of balances. Such account-based transaction systems also suffer from a lack of scalability since almost every transaction must be processed through at least one centralized or even multi-party clearance system. Yet another drawback of existing account-based systems arises when a central authority such as a country's central bank wishes to emit new currency units. If the currency units are intended to be individualized, such as through serial numbers, then this individualization is usually lost when account balances are changed, since accounts generally are defined by single-value balances. This problem of course also arises in other contexts in which control rights are to be transferred for other data sets that are uniquely identifiable, whether they may be sub-divided or not.
Analogous problems arise in other contexts in which it is not the uniqueness of individual data sets that needs to be ensured upon transfer. One such situation is where the holder of a number of items wishes to transfer proof of ownership of some of these to a recipient in such a way that the recipient can be sure that the holder/sender didn't also commit those same items to yet another recipient.
Even where data sets are individualized, there is, moreover, in some situations a need to be able to transfer control of less than the whole of a data set.
Embodiments disclosed here are based on a novel decomposed (“sharded”), stateless blockchain arrangement, and have different aspects that may be used in different contexts. Some embodiments, for example, enable verifiable transfer of ownership of any type of data unit, such that only one entity at a time should be allowed to control of the data unit. In some implementations, data units themselves need not be transferred, but rather only information uniquely identifying the data unit involved in a transfer request, as well as information about the transferor and transferee. In some such embodiments, the data unit may be considered a kind of abstraction, in that no actual data set needs to pass uniquely from one entity to another. In other embodiments, the data unit itself may contain the data structures used to establish exclusive ownership of it.
A particular embodiment is disclosed below that enables “splitting” of an individualized data set so as to enable transfer of control of only a portion of if. In the case of data sets that have a nominal value, this may include the creation of additional data sets whose nominal values together equal that of the original, while still preserving individualization. In some other situations, data sets might represent other types of information that can be partitioned and transferred piece-wise in separate transactions. For example, a document might have portions that are to be routed to different recipients. As another example, items that are assigned a digital identity (a digital “twin”) may comprise multiple units that are to be shipped to different recipients. A shipment of 10000 packages of a medicine, for example, or of 10000 units of some machine part, may need to be transferred to different parties, with the corresponding original digital identities being “split” so that each portion is also individualized, with verifiably unique ownership.
Here, a transferrable set of data may be viewed as a data “unit”, even if it comprises more than one parameter, in that it is some body of information in digital form that is to be associated with only one entity at a time. Some examples of such data units are given above; others include digital titles to land or other property; digital versions of negotiable bearer instruments; access codes; copies of audio and/or video files that only one party at a time is authorized to control (such as in a Digital Rights Management framework), etc.
Ownership of a data unit may be purely logical, that it, without a requirement for any data file or the like to be moved from the system of a current owner to the system of a subsequent owner. Consequently, reference here to transferring a data unit, such as a bill, is to be interpreted not as a requirement to move a data set from one system to another (although this would be possible, for example, in conjunction with a verified change of ownership of the data set, or of portions of a document, etc.), but rather that ownership of the respective data unit is changed from a transferor to a transferee. This is common even in other contexts as well. For example, some exclusive ownership rights may be changed in some jurisdictions simply by having the current owner, or its properly authorized representative, after proper verification of identity, upload to a governmental authority a proper request to record the change. Some mechanism is then usually provided to allow the new owner to verify that the transfer went through.
The concept of “bill” should therefore be considered as also comprising some digital information or data structure state that identifies the entity that has the exclusive ability to request a change of control of ownership of the right represented by the bill. In the case of money, that “right” is the ability to control ownership of a concept of value that has been established by the emitting entity and is accepted by a transferee. In some other contexts, ownership may similarly be related to a more generalized concept of quantity such as the number of units defining some portion of a whole.
Merely by way of example, and because the terminology for payments is generally well-understood, embodiments are described below primarily in the context of digital cash. This example also has the advantage of including some concepts, both qualitative and quantitative, such as “denomination” or “value” that may not be present in other contexts. This is, however, just one possible use case and the concepts may be applied in other cases as well with modifications—if needed at all—that skilled programmers will be able to implement.
Assume, again by way of example only, that the type of data set one wishes to be able to transfer the ownership right to defines a unit of digital currency—a “digital bill”, or simply “bill”. Such a digital bill may be viewed as a data set or abstraction that has at least the attributes of 1) fixed nominal value and 2) a unique identifier such as a serial number.
As used here, the term “emitted money” (or, more generally, “emitted quantity”) is a number e that represents the total amount of money in current circulation, that is, the sum of the nominal values of all bills. In other contexts, e may represent some other total value or quantity. Bills are “emitted” by an “emission process”. In the physical world, bills (or other documents) are typically “emitted” by being printed or minted, preferably in some difficult-to-counterfeit way, and then put into circulation through some predefined process.
Different embodiments of the invention may operate with centrally emitted bills, in a permissioned arrangement, or with bills emitted by a non-central entity or cooperating group of entities in a permissionless arrangement. Below is described a method to emit digital bills, that is, data sets that represents some notion of value or quantity. Merely by way of example, the disclosure below will focus on the permissioned arrangement; changes to accommodate a permissionless system will be either described or will be within the skill of programmers familiar with blockchain technology.
After emission, some entity will be the authorized initial holder of a digital bill, that is, the “rights holder”, who/which may be termed an “owner” or “bearer” depending on the context and implementation. Without loss of generality, the terms “owner” and “bearer” are used here to designate the entity that currently holds the right to request a transfer or a change of that right, to another entity or to itself, such as from one corporate entity to another commonly owned corporate entity, from an individual to himself, such as between wallets or in a “split” as disclosed below, etc. The holding of rights to a digital bill may be represented or viewed as a data structure with at least the attributes: 1) an identifier IDU of the data unit, that is, the digital bill itself, and 2) an identifier IDOR of the transferor, that is, current owner. As for owner identity, note that some embodiments of this invention make it possible, although not necessary, to maintain the anonymity of actual parties to transactions even during the transfer process. This is consistent with the concept of “cash”: If Alice gives a $10 or €10 note to Bob, no other person or authority will typically know either's identity.
It is not required by any embodiment of this invention for a rights holder to be an individual human; rather, any identifiable entity may be an owner, etc., including individuals and groups, both physical and legal persons, and even software processes that are able to perform the actions described below to either initiate a transfer, receive ownership of a transferred data set, or both.
Assume by way of example that ownership rights to bills are to be transferred from one entity to another. In the context of the example of digital cash, this transfer might be a payment for something, a transfer for the purpose of deposit in some other data structure, etc. In general, all transfers of digital cash are a form of “payment”, which means simply the process of changing the designation of the bearer of a digital bill, or some other designation in the case full or partial self-to-self transactions such as a “split”. Payment is initiated by a payment order, which is represented by a digital data structure with at least the following attributes: 1) the digital bill to be transferred, 2) an identifier of the payer, who is the current bearer/transferor; and 3) an identifier of the payee/transfee, that is, the recipient, who is to be made the new bearer.
When designing a solution for transferring a data set, that is, a digital asset such as, for example, digital cash, some key considerations are:
The answers to these questions contain the following security concerns:
In case of physical cash, for example, money is defined by a central bank and is provided with security elements that enable high security. The central bank controls the emission process, so it may be made highly secure. The bearer is simply declared to be the physical bearer and the bearer is changed directly by the payer and payee: If Alice hands a €20 note to Bob, the transfer of ownership is complete and Bob is the new bearer.
Summary of Three Types of Prior Art Solutions
To illustrate some aspects and shortcomings (in particular, technical bottlenecks and lack of scalability) of existing methods for ensuring verifiable and unique transfer, consider some existing models of digital money: bank money in two different settings, and Bitcoin. These models are 1) a trusted server solution that assumes a trusted processor that has full control over the data and isn't audited/verified by external parties, 2) a modified solution in which data structures are augmented with cryptographic certificates that make the system externally auditable/verifiable, and 3) the Bitcoin system, which is similar to the second case, except that the emitting party is eliminated and replaced by a fixed emission rule, and the notion of ownership is slightly different.
In a trusted server solution:
The “bill” in this trusted server arrangement is a number in a bank account and may have any nominal value from 0 to e. Payment processing in this known solution means that the bills of the payer and payee are destroyed and new bills are emitted. The security of a bill and its ownership are based on full trust in the server, although the security of payment can be improved by server authentication, account holder authentication, and digitally signed payment orders.
One way to improve security is by replacing the trusted server with a certified ledger. This then creates a trust-free server solution, typically based on a blockchain. In this solution, the data that is processed by the bank (accounts and balances) is made public and secured by a certificate. Payment orders are recorded in a ledger and put into public domain. For privacy reasons, accounts may be anonymized. In order to verify a bill, one needs a full ledger in order to verify that e does in fact equal the sum of the values of all emitted bills. One problem with this solution is that the resources needed for verifying a bill do not scale.
Bitcoin-type solutions rely on a permissionless blockchain. In this solution:
In all three of the solutions just summarized, in case of fixed e, not only the number but also the nominal value of bills in use varies. In part, because of this, the verification of neither the bills nor their ownership is scalable.
Hash Functions
Hashing of data is a well-known procedure and is used often in embodiments of this invention. In general, a cryptographic hash function h converts binary data X of arbitrary size to a bitstring (called the “hash value” or just “hash”) x=h(X) of fixed size, typically 256 or 512 bits. Cryptographic hash functions are assumed to be “collision resistant”, which means it must be computationally infeasible to find a second, different binary input X′ that has the same hash value as the first, X. The SHA class of hash functions is just one common choice that may be used in embodiments here, but no embodiment depends on this choice. Another advantage of hash functions is that they are in general efficient to compute: even standard hash functions like SHA-2 or SHA-3, for example, enable about a million hash operations per second on an ordinary desktop computer with only one processor core.
Blockchain
Although the term “blockchain” itself, as well as related terms, do not yet have universally accepted definitions, typically a “blockchain” is understood as being a data structure comprising a series of usually cryptographically linked, where each block includes data corresponding to one or more transactions, hashed together with linking data, such as the hash of some data and/or metadata of at least one preceding block. The blockchain can then be used to create a ledger, which is typically an append-only database.
Some blockchain variants involve distribution and consensus, that is, copies of the entire blockchain are distributed to several entities, which then follow a procedure to come to some pre-defined notion of “consensus” as to what data is to be allowed to constitute the next block. Many of the blockchains used for cryptocurrencies follow this model, for example, since they, usually by design philosophy, wish to avoid any central permissioning authority.
In other “permissioned” configurations, at least one controlling entity may control access to a proprietary blockchain according to its own rules; governments, banks, enterprises, etc., will, for example, usually not want the operation of their blockchains to depend on consensus among distributed, often anonymous outside entities. In either case, once data is entered into a block of the blockchain, the entry is essentially irrefutable, that is, non-repudiable, since any tampering with the data would be reflected in the chained hash calculations and thus easily detected.
Witness
Below is described how cryptographic proofs are obtained to enable verification of the state of various data structures and operations. Such a proof is generally referred to in the field of cryptography as a “witness” or a “certificate”. One example of such a witness is a digital signature, which is the example used below in discussions of embodiments and aspects of the invention. Except where a specific signature mechanism is referenced, the term “digital signature” should therefore be read as including any form of witness or certificate that can perform the same function of cryptographically verifying a given data state.
One example of an alternative to digital signatures is the structure known as an “authenticated dictionary”. Using such a structure, one obtains a certificate for a state input by submitting a representation of the state input to the (potentially distributed) authenticated dictionary, in which the certificate comprises cryptographic information proving the participation of the state input, and returning an authenticator value as a trust reference. A later purportedly authentic representation of the state input may then be verified as being valid if cryptographic verification of the purported state input, the respective certificate, and the authenticator, succeeds.
Digital signatures are used in some steps of the embodiments described below as the witness. Embodiments of this invention do not require any specific form of signature service or witness, and the system designer may choose any system that satisfies the security requirements of the particular implementation. An advantageous service and digital signature, however, are provided by the data signature infrastructure developed and marketed under the name “KSI®” by Guardtime AS of Tallinn, Estonia. This system is described in general in U.S. Pat. No. 8,719,576 (also Buldas, et al., “Document verification with distributed calendar infrastructure”). In summary, for each of a sequence of accumulation rounds, also known as calendar periods (typically related one-to-one with physical time units, such as one second), the Guardtime infrastructure takes digital input records as inputs, that is, lowest-level tree “leaves”. These are then cryptographically hashed together in an iterative, preferably (but not necessarily) binary hash tree, ultimately yielding an uppermost root hash value (a “calendar value”) that encodes information in all the input records. This uppermost hash value is then entered into a “calendar”, which is structured as a form of a type of blockchain which, in some implementations, may involve aggregating calendar values into a progressive hash tree. The KSI system then returns a signature in the form of a vector, including, among other data, the values of sibling nodes in the hash tree that enable recomputation of the respective calendar value if a purported copy of the corresponding original input record is in fact identical to the original input record.
Note that no KSI signature is returned to any input entity until all inputs have been received for a given calendar period. This is because, until all inputs are received, it is not possible to compute the root value. One consequence of this is that, once a signature has been returned for an input, it is too late to attempt to get another signature for the same (or any other) input value in the same calendar period.
Although the KSI infrastructure can thus function as a timestamping/synchronization mechanism at the same time as providing digital signatures, it would also be possible to synchronize transaction commitments using any other chosen timing or time-stamping mechanism; moreover, other signature mechanisms may be chosen to form the basis of the various proofs described below. An accumulation cut-off and/or time synchronization mechanism such as the KSI infrastructure (as one example) provides is used in embodiments to prevent so-called “double spending”, that is, more than one transfer of the same data set to different recipients.
As long as it is formatted according to specification, almost any set of data, including concatenations or other combinations of multiple input parameters, may be submitted as the digital input records, which do not even have to comprise the same parameters. One advantage of the KSI system is that each calendar block, and thus each signature generated in the respective calendar time period, has an irrefutable relationship to the time when the block was created. In other words, a KSI signature also acts as an irrefutable timestamp, since the signature itself encodes time to within the precision of the calendar period.
Yet another advantage of the Guardtime infrastructure is that the digital input records that are submitted to the infrastructure for signature/timestamping do not need to be the “raw” data; rather, in most implementations, the raw data is optionally combined with any other desired input information (such as user ID, system information, various metadata, etc.) and then hashed. Given the nature of cryptographic hash functions, what gets input into the KSI system, and thus ultimately into the calendar blockchain, cannot be reconstructed from the hash, or from what is entered into the calendar blockchain.
Scalable Blockchains
An example of a trust-free solution is blockchain-based. One of the main concerns related to today's blockchain solutions is, however, their poor scalability. For convenience, and to help understand the novel embodiments of this invention, the general theory of scalable blockchain solutions is outlined; fault- and attack-tolerant implementations of blockchains are also discussed.
General Description of Main Components
A general view of the main components in some permissioned embodiments is depicted in
Although the components 100, 200, 500, and 600 are illustrated as being separate systems, which will be a common choice in practical implementations, in some cases it may be possible to implement two or more of these components on a common hardware and/or software platform. Conversely, as will be understood from the description below, some of the components depicted in
In
The routines, processes, storage functions, etc., described below must of course be performed by actual hardware and software platforms, even if any of these are done remotely, such as by using cloud computing, or in virtual machines, etc.
Each platform will include standard components such as system hardware 810 with at least one processor 811, volatile and/or non-volatile memory and/or storage, indicated “collectively” as component 812, and standard I/O access components 814 to enable communication with other entities and systems over any known type of network, wireless or wired. The processor-executable code organized as software modules used to carry out the various computations, routines, and functions described below may be stored and thus embodied in either or both types of memory/storage components 812. The software modules will thus comprise processor-executable code that, when run by the processor(s) 811, cause the processor(s) to carry out the corresponding functions. Some form of system software 820 will also be included, such as an operating system 821 and/or virtual machine hypervisor.
Platforms will also include an application layer 830, which comprises various software components/modules for performing the functions described below. Although some entities may run on a single hardware/software platform, such as is shown in
As is explained below, this is particularly the case with the blockchain machine 200, which will typically (but not necessarily) comprise a different virtual and/or physical computing platform for each node that maintains or inputs a bill ledger and even “a” node may be comprised of a cluster of separate platforms. Thus, reference to “the” blockchain, including in relation to “the” blockchain machine 200, is to be taken to mean the collection of separately operating blockchain shards/sub-ledgers, and the computing systems in which they are processed, that, together, encode the ownership history and status of all transferrable data units in the system by means of respective cryptographically linked data structures.
In the following description, nodes are interchangeably referred to as “gateways” to indicate that they are the computing systems via which users interact with the overall system.
The service infrastructure-process 10 will typically update the blockchain data structure based on some additional input, say, x. For example, as described below in the case of a digital cash implementation of embodiments of this invention, x may contain emission orders and transfer/payment orders. The service may also optionally implement a public rule-based process for computing a new version of the blockchain, for example, based on the previous version and the additional input x. An example of such a rule might that, for a given data set (such as cash “bill”), only one change may be made during some period, for example, the period during which signatures are formed; in implementations that use the Guardtime KSI system for signatures, this period may be typically be the “calendar period” or, more generally, the period during which inputs are aggregated to form a hash tree root value from which signatures are generated as hash chains. These concepts are described in more detail below.
The blockchain data structure used in some embodiments does not need to depend on the central controlling component 100 having a trusted long-term memory of all transactions; rather, in those embodiments, all necessary data for the service may be stored in the input and output components, with the blockchain mechanism itself being stored in the component 200. As is described below, “the” blockchain component 200 is implemented as separate portions, that is, shards, which are maintained and processed in respective “nodes”, comprising one or a cooperating “cluster” of computing systems preferably structured as in
Cryptographically Verifiable Data
All data in the blockchain should preferably be reliably verifiable without using any assumptions about the physical host machines. This may be achieved by using cryptographically verifiable data, that is, the blockchain itself and the additional input x contain cryptographic certificates that protect their integrity. The verification of input and output data may thus depend on the service rules and cryptographic certificates. There should therefore be an efficient verification process.
Mathematically, the blockchain may be defined by two functions:
Both V and U may depend on parameters, such as public and private cryptographic keys.
The system may set U(B′, x)=B′, if x is invalid. In other words, if the input value x is in any way improper, then the blockchain is not changed according to x.
System Scalability
As mentioned above, a common problem with existing solutions is a lack of scalability: As the number of transfers (“transactions”) increases, either the system cannot keep up in time, or the size of the ledger that many different entities must agree upon, update, and store becomes impractical. In embodiments of this invention, this problem is addressed in part by decomposing the ledger/blockchain, that is, it implements a partitioning rule (see sub-ledgers L1, L2, . . . ,
so that a transfer involves one particular part Li of the ledger, not always the whole blockchain B. Mathematically, one may represent this decomposition as a function D such that D(B)=(L1, L2, . . . ,
) and a composition function C such that C(L1, L2, . . . ,
)=B, that is, C(D(B))=B for any possible instance of the blockchain. In this case, one may use the notation:
B≅(L1,L2, . . . ,)
For full-fledged scalability of the system, both the verification of the ledger and the production of the ledger by the service should be scalable.
Verification Scalability
In a blockchain with scalable verification (see
V(B)=V1(L1)ΛV2(L2)Λ . . . Λ),
where V1, V2, and ), are predicates.
Service Scalability
Every physical server has limited processing speed, memory and network connectivity. In order to make the production of the blockchain scalable, embodiments of this invention decompose the service process into multiple processes S1, . . . , Sm (see such that if B′≅(L′1′, L′2′, . . . ,
), then for every input x:
U(B′,x)≅(U1(L′1,x), . . . ,,x)),
As mentioned above, in embodiments of this invention, the blockchain, which encodes the state of all emitted bills, is sharded, such that, instead of a single global ledger that all system actors must store, there are subledgers that, together, include the information for bills. Different computing systems—“nodes”—are responsible for storing and processing respective shards. Each node may itself optionally comprise a plurality of cooperating systems that may implement any consensus protocol to decide on any changes to the shard(s) they handle. Note that it would be possible for one node to be responsible for more than one shard, for example, until additional computing capacity is available or temporarily for purposes of system upgrade or maintenance.
Here, by “sharding” is meant a function λ that for every additional input x, returns an index λ(x)∈[1, . . . , ] such that for every index i, except i=λ(x), we have U(L′i, x))=L′i. In other words, each input x is directed to and influences only one subledger Lλ(x). In practice, this means that, given an additional input x, before sending it to the service, the system computes i←λ(x), and sends x to only the component Si of the service, because other components would have no reaction to x anyway. The practical implication of this is explained below but can be summarized here: each given bill that has been emitted is associated with one of the ledger shards, that is, sub-ledgers.
Such a decomposition, according to embodiments of this invention, can guarantee limited memory, processing, and communication requirements for the component servers, which, consequently, enables far superior scalability relative to prior art solutions that use a single, distributed global ledger. Increasing the supply of the number of data sets, such as through the issuance of new money or increased production of some goods (and their corresponding digital identities) or documents, would not necessarily increase the computational burden on any node, since it would be possible simply to add one or more additional nodes and assign the newly created money/identities/documents/etc. to subledgers in the added node(s).
The general logic of data processing in a scalable blockchain system according to embodiments of this invention is depicted in
L
λ(x)
=U
λ(x)(x,L′λ(x))
The newly computed subledger can be verified by applying the component Vλ(x) of the verification function.
Fault- and Attack-Tolerant Implementation of the Service
To eliminate or reduce trust requirements of the service, the implementation of the service is preferably made fault tolerant; for example, it should preferably guarantee that every correct and consistent transfer/payment order will eventually be processed by the service and ledger changed accordingly. To solve this problem, a redundant design may be implemented in which the service is provided in parallel with a cluster (900,
One multi-party communication protocol that may be used to provide fault tolerance may be any of the class consensus mechanisms that provide Byzantine fault tolerance (BFT) for the detection and correction of Byzantine faults. A “Byzantine fault” of a physical server means here any form of misbehavior of a server, including all kinds of sneaky adversarial behavior; this is a known definition. Byzantine fault tolerance thus implies attack tolerance.
There are known multi-party protocols developed for achieving fault tolerance for a limited set of Byzantine faults (limited number of faulty-nodes, etc.). These protocols are known to be very efficient for a relatively small numbers of physical servers. In the context of implementations of embodiments of this invention that use server clusters, the properties that the designer-chosen protocol should have include:
The additional input x is sent to all physical servers S1, . . . , Sm of the cluster. The servers may then use a gossip mechanism (a known concept, involving inter-server communication) such that if x reaches at least one non-faulty server, it reaches all non-faulty servers, or at least a predetermined minimum number of them.
The necessary number of redundant servers depends on the fault-tolerance requirements of the service. The system designer may choose any known fault tolerance solution based on the particular requirements of an implementation of the system for particular states.
Note that, in systems such as Bitcoin, fault tolerance is achieved only with massive, generally thousand-fold redundancy, with Byzantine faults being corrected by using a proof-of-work concept. Such massive redundancy will be impractical in many cases such as where a central bank wishes to implement digital cash. Moreover, proof-of-work schemes at that level typically introduce an uncertainty in the oversight and control of the system that will be unacceptable in many situations such as with central banks or governmental authorities.
Some embodiments implement a system that enables highly scalable and verifiable transfers of single-instance data sets that leverage the Guardtime KSI infrastructure summarized above. One illustrative embodiment is digital cash. This example has the advantage that it has certain features and requirements that are not present or may be relaxed in other scenarios. Transfer of cash, for example, involves the notion of “value” or “denomination, and often (but necessarily) a desire for transaction anonymity that, for example, transfer of a highly classified document file might not. The invention is not limited to such applications, however. Furthermore, embodiments are not limited to using the KSI infrastructure at all: other signature mechanisms may be used to generate proofs that can be used for verification; other timing mechanisms may be used in embodiments that use them; etc. Use of the KSI infrastructure is described by way of example only.
Assume by way of example that an embodiment of the invention is to be implemented for a digital cash system such that a central authority or administrator, such as a central bank, wishes to be able to control the emission of currency units (“bills”, for short), where the holder (“owner” or “bearer”) of a bill is able to request transfers, that is “payments”. To increase the trustworthiness of the system, it would preferably also be possible, even without using trusted services to perform transfers, for an external party such as an auditor to be able to audit both operations.
See
Three component processes, which may be implemented in any computing system or group of computing systems, are: 1) a Wallet process 910, capable of creating payment orders; 2) a Central Bank Wallet (CB Wallet) process 912 capable of creating emission orders and the payment orders; and 3) a Verify process 914 that carries out a verification procedure for checking cryptographic certificates of the blockchain.
Users of the system may have respective instances of the Wallet process 910, in their respective personal (or institutional) computing systems, and the Central Bank (or other central authority that issues whatever notion of “bills” is involved) should have an instance of the CB Wallet 912. The wallets, which may be configured as any known data structure, contain cryptographic keys for generating the cryptographic certificates (signatures) of Emission and Payment orders. (Note that embodiments of this system do *not* require keys to be generated according to the widely used Public Key Infrastructure—PKI—protocol, although this may be used in some embodiments depending on the preferences of the system designer, and, in some cases, of the users themselves.) Wallets may be implemented using any know data structures and coding methods.
The Verify procedure may be implemented in any known, chosen manner. How a KSI signature is verified is described above. The implementations of Wallet 910 and CB Wallet 912 preferably themselves contain respective instances of the Verify procedure since this will allow them to verify signatures without needing to query an external trusted system to do so (at least if KSI signatures are used), but such reliance may be an acceptable option in some implementations.
The illustrated system has three main processes: Creation, Transfer, and Verification. In the context of digital cash, these may be termed Emission, Payment, and Verification. In other words, the central authority creates unique data units (such as a bill); these data units are transferred from one entity to another (such as a payer to a payee); and the parties involved or a third-party auditor should preferably be able to verify that a transfer is valid and correct. It would be possible to dispense with the Verification process if all users and other entities trust (or are required to trust) the system, but in practical implementations this will generally not be acceptable.
Emission
The Emission process changes the amount e of money in the system. New data units/bills are thereby inserted into the system, the initial owner of which will typically be the central bank (or other initial owner/holder of the data units). Note that the central bank will in many cases provide the emission service for digital cash itself, although this is not required. In some jurisdictions, for example, some commercial banks are authorized by the central authority to issue new bills. A unique identifier, such as a serial number, is associated with each data unit. In implementations in which the data units being transferred represent such items as units of digital currency, checks, instances of stock certificates or other rights, other negotiable bearer instruments, etc., all of which are examples of other types of data units/“bills”, emitted bills will also have an associated nominal value. In implementations in which the data units represent digital identities of quantifiable items, the “nominal value” may instead represent quantity.
Emission may proceed from the central bank wallet 912, which sends an emission order to the service infrastructure 10, which then updates the blockchain to include newly emitted bills. This may be carried out by the input component 500 or in the central controlling system 100, depending on where the associated processing functions have been chosen to be carried out in particular implementations. Note that updating “the” blockchain means here a change (or, in one embodiment, creation) of the sub-ledger associated with the newly emitted bills.
Transfer/Payment
The Transfer/Payment process changes the designation of the owner of a bill, that is, of the entity that hold the rights to it. It is therefore necessary to be able to identify both the payer/transferor and payee/transferee. In embodiments of this invention, “identify” does not necessarily imply knowledge of the actual identities of the parties involved, although this is an option. Rather, the identifiers used by the service for the parties in a transfer may maintain the parties' anonymity, such as using a party-generated public key. In other embodiments, however, anonymity may not be required, in which case any other chosen identifier may be used, such as a national ID number. In some implementations, the service may be provided not by a central governmental authority, but rather by a private entity that issues and controls other value or quantity units (whether or not convertible to legal tender of any kind) or types of data units; in such cases, the controlling private entity may also assign user/owner identifiers, which may be associated or associatable with actual identity, or not. A transfer involves changing a bearer identifier associated with that bill. To carry out a payment order (transfer request), the current bearer's Wallet sends the payment order to the Service, which then updates the blockchain (in particular the sub-ledger(s) associated with the bill(s) involved) to indicate the designated payee as the new owner. Users 400, which will include both transferors and transferees of bills, may access the service via the API 300, which communicates transfer requests and completions to the components 500 or 600. Note that different users may, and typically will, have different instances of the API 300; only one instance is depicted in
Verification
The information in payment orders, emission orders, and the different parts of the blockchain, may be made cryptographically verifiable, for example, by obtaining KSI (or other) signatures for each. The blockchain may provide the following verifiable proofs, which may be checked via the Verify process, which in turn may be carried out by a verification component located within whichever entity wishes to verify information. In
Scalability
One of the advantages of embodiments is that they are highly scalable. This results from novel decomposition of both the verification process, and the Service process. See
Decomposition of the Verification
The blockchain B in embodiments here is decomposed into the following independently verifiable parts:
To verify the status of one bill, a Wallet needs only the identity of that bill's bill ledger, and preferably the Emission ledger EL (to verify that the bill was validly issued in the first place).
Decomposition of the Service
The input component 500 will include a respective gateway front/interface component 510 that receives information such as payment orders, confirmations, etc., from user systems 400, and determines, based on the identifier of the bill, which of the sub-ledgers tracks the bill. Gateway front input components/channels IC1, . . . , I
receive payment orders from the respective front input component 510, as well as any input queries from the respective gateways. It then routes, for example, the payment order information for the bill to the correct, corresponding gateway GW1, . . . , G
, (determined based on the data unit identifier) via the respective gateway front input components/channels IC1, . . . , I
.
To complete a transfer/transaction, the gateways GW1, G will route transfer information via respective gateway output sub-components/channels OC1, . . . , O
so that the front output component 610 may communicate with the recipient/payee user, for example, via the API 300. The front output component 610 is preferably also configured to perform such tasks as caching payment information, access control, filtering, and serve as a routing layer between the client API and gateway output component OC1, . . . , O
instances.
Emission of new bills may (in a permissioned arrangement), as mentioned, under the control of the central system 100, such as a server or server group under the control of a central bank. The central system 100 thus includes a controller 130 configured to communicate with a core 210 controlling component of the blockchain machine and with the emitter output component EOC. An emission component 110 is also provided, which communicates with a dedicated emitter gateway GW0, which may then incorporate any newly emitted bills into the corresponding emission ledger EL (
Sharding Function λ and Service Configuration
In one embodiment, predetermined bits of the serial number (or of the identifier of whatever other data sets the system has been implemented for) of a bill determine with which gateway ledger it is associated. Association may be of two types: 1) in some embodiments, the gateway itself stores and maintains the data structure that comprises the sub-ledger and bill ledger (described below) that an identifier directs to; and 2) in some embodiments, the sub-ledger is included in a data structure that defines the respective data unit itself, and is passed by a transferor to a gateway upon a transfer request. For now, the description will focus on case 1).
As one example of a method for associating data units with gateways (that, is sub-ledgers), assume there are gateways in the system, where an identifier of each gateway is an m-bit number, where m=log2
. The m highest order (or other) bits of a bill's serial number may then be used to “point to”, that is, determine, in which gateway the bill's ledger is maintained and/or processed.
=4 and m=2. The identifiers of the 16 bills in this example are thus 0000, 0001, . . . , 1110, 1111, and the gateways are numbered 00, 01, 10, and 11. Assume that the identifier of a bill is 1001. Since the two most significant bits (MSB) are 10, any transactions relating to the bill will be directed to Gateway 10, which holds the bill ledger 1001 for the bill. Note that one advantage of this arrangement is that bill identifiers may be chosen to help balance the expected load on the different gateways. If the maximum number of bills is k=2n, they have at least n-bit serial numbers. This means that, on average, in an equally load-distributed configuration, every gateway will maintain 2n-m bill ledgers.
The sharding function λ(x) may, for example, be defined to be 0 if x is an Emission order, since this will not correspond to any particular gateway. If x is a Payment order, however, then λ(x) may return them highest bits of the bill's serial number. The identifier i of the gateway that operates a payment order x (called the associated gateway) may thus be computed from the bill's serial number via the sharding function: i=λ(x). In short, a function is applied to the identifier of each bill to determine with which one of the gateways (and thus ledgers) it is to be “assigned” to.
Data Structures
Embodiments rely on several data structures and the processes by which they are established, changed, and maintained. These include different notions of “blockchain”, data signatures, etc.
Blockchain
As used herein, a Block is a cryptographically verifiable data structure that consists of Data D and a Signature S:
See
Block Signing
In one implementation, Data D in a block is a sequence (d0, d1, dk) of hash values, which typically have a fixed size, and the sequence (some of whose values may be missing) is signed. Any known signature method may be used, but the KSI signature has the advantage that there are in general no more syntax or semantics rules that a KSI blockchain has to fulfill. Signature S may, for example, be a KSI tree signature KSITreeSig(d0, d1, . . . , dk) on D.
For every item di∈D one can compute a KSI signature, in the form of a chain ci=KSISig(i, D) for di, which forms a cryptographic proof that di is the i-th component of D. In other words, if the value i is included as a parameter along with associated data in the input to the KSI signature system, the signature vector returned will enable recomputation through the KSI hash tree up to an irrefutable, known value, that is, the corresponding calendar value, but only if the correct value i is included as a parameter in the input submitted for verification.
KSI Signatures
U.S. Pat. No. 8,719,576, mentioned above, gives a more detailed explanation, but the use of KSI-generated signature is summarized here for completeness. One feature of the KSI signature system is that it operates in time periods, which may be referred to as “calendar periods”, “aggregation rounds”, etc. For every calendar period t, the KSI system inputs values as “leaves” (lowest-level values), combines these in a Merkle tree, computes a current root value (the “calendar value”) rt and then may return to each “leaf” entity the signature vector allowing recomputation of rt given the same leaf value as was input.
KSI Signature KSITreeSig
More formally, a KSI tree signature s←KSITreeSig(x0, x1, . . . , xk) for a sequence x0, x1, . . . , xk of hash values is computed via the following steps.
One way to render the value r irrefutable with respect to a particular entity is to sign it using any known public key signature algorithm (depicted as “Signature”); this may then tie the irrefutability to the holder of the public key. Another way would be to include the entity's private key as part of the lowest-level input, that is, as either a tree “leaf” itself or as a parameter included in the input set.
If s is a KSI tree signature for a sequence x0, x1, . . . , xk, then for every i=0, 1, . . . , k, the KSI signature KSISig(i; x0, x1, . . . , xk) is a pair (s, ci), where:
c
0=(00;x1,x23),c1=(01;x0,x23),c2=(10;x3,x01),c3=(11;x2,x01)
Thus, consider the hash chain for the second hash tree leaf from the left in
Note that the order of the signature elements may be chosen to be different than that shown, as long as the chosen order is known and maintained as a convention by the signature-generating entity, and all parties that need to verify data given its signature.
“Cash” Blockchain
In the context of money, “cash” has the property that each unit (“bill”) is uniquely identified, for example, by its serial number, has a set value (denomination), and has a requirement for well-controlled emission (no counterfeiting). These properties may also be found in other unique-instance, and uniquely identifiable, data units that embodiments may be used to enable provably unique transfer of. In implementations involving digital cash (or the like), for the blockchain viewed as a whole, the data part D of a block Bt=(D, S) may include (in some embodiments):
In embodiments in which the KSI infrastructure is used to generate signatures, the signature part S may be a chained KSI tree signature (ct, xt), which may be defined recursively:
x
0=(x00,x10, . . . ,xk0)=(h(E0),h(P10), . . . ,h(Pk0))
c
0=KSITreeSig(x00,x10, . . . ,xk0)
x
t
=h((x0t−1,Et),h(x1t−1,P1t) . . . ,h(xkt−1,Pkt))
c
t=KSITreeSig(h(xt−1,Dt))
where h is a cryptographic hash function. Thus, for each payment order, a signature (KSI or otherwise) is generated, which preferably encodes the signatures of previous payment orders.
In other embodiments, instead of including the entire past signature chain, only the immediately previous signature may be included. In the KSI structure, for example, the calendar encodes all previous signatures as well, and also is synchronized with time, such that the previous signature will also be irrefutably time-stamped.
It is not necessary for data unit transfer requests to be grouped into “rounds” or to be otherwise time-limited—a transfer request may be kept “pending” until all information exchange has completed and ownership has been changed to the transferee. One advantage of doing so, however, is that it makes it possible to ensure a maximum settlement time (assuming no system failures), or at least a time by which a transfer request must either have completed or been denied. Grouping transfer requests also makes generation of proofs more efficient. In some embodiments, therefore, data unit transfer requests are received in input/aggregation periods, such that there is a specific cut-off time. Use of the KSI signature infrastructure then has the advantage that it is already configured to generate signatures in calendar periods, which are typically of a fixed length, such as one second. A KSI signature thus “automatically” provides not only proof of correctness of an input, but also a timestamp.
If such time-based grouping is not implemented, other synchronization mechanisms may, however, be used instead. It would also be possible to “batch” input requests until some minimum number of requests have been received, before completing the transfer requests, or to implement any other such rules. Such “batching” may, however, lead to uncertain settlement times, but that may be acceptable in some contexts. It would also be possible for nodes/gateways to include routines that implement rules allowing for other options such as a selectable maximum transfer settlement time, or other conditions.
In many cases, there may be a large number of input periods in a row during which a bill, or a particular bill, is not transferred at all. Rather than actually iteratively hashing even the “null” Pit values, an incrementing index nullinarow may be included instead. During verification, this index may indicate how many consecutive null values occurred, such that the verifier will know to hash the non-null P value just before the no-transfer periods nullinarow to get the non-null P value just after those periods end. In other words, the index can be used to reduce the number of hash computations needed up-front to only those relating to actual payment orders, with remaining hashing computations being done only as needed later for verification. Note that, if the KSI signature infrastructure is used, the index nullinarow itself may be derived from the time indications of the signatures of the non-null payment orders at either “end” of the null-periods, such that it would not be necessary to explicitly include nullinarow at all.
Ledger Decomposition
Continue to assume by way of example that KSI signatures are used. (Suitable similar operations may be used for other signature schemes, as skilled system designers will appreciate.) The blockchain can be decomposed into:
A novel structure of the KSI signature infrastructure that is adapted to help implement this is depicted in
Emission Ledger
Various data structures may provide proof that a bill was validly issued and is still validly in circulation. These include the Emission ledger itself, which includes or communicates with data structures relating to Emission order and Bill series. See
Emission Ledger includes:
Emission order preferably includes:
Bill series includes:
Bill Ledger
A Bill ledger is created for each bill and forms a “sub-ledger”, in that it tracks only a subset of the bills in circulation, namely, the bill to which is it assigned. The Bill ledger may be a list of blocks, the first of which may be called the Emission block. See
Each block contains:
The Block hash may be computed as the hash of a concatenation of the data part and the previous block hash. In case of the first block, it may be the hash of the data part.
The Proof part may be either empty (null) or contain a signature of the block hash.
The Data part may be either empty or contain a Payment order with, for example, the following fields:
The Emission block will always have a non-null data part, since it represents the issuance of a valid bill. A bill ledger is full if all of its blocks have non-empty proofs (
Proofs
Embodiments of the invention provide several proofs that enable verification of the status of a bill and of a transfer, and thus allow for easy auditing of the system as a whole and its various functional parts. These proofs include Proof of Money (POM), Proof of Emission (POE), Proof of Ownership (POO), Proof of Transfer (POT). These proofs may be KSI signatures, that is, hash chains leading to an irrefutable root, which may be recomputed from a given “candidate” input value—if, recomputing the KSI hash chain upwards with the sibling values in the signature, the same root value is reached as when the respective structure was signed, then the candidate input value must be the same as the original value for which the signature was generated. Any other known signature method may be used instead, however, depending on the level of security and auditability desired in each given implementation of the invention. Each proof has an Input, and Output, and Semantics, such as:
Proof of Money (POM)
Proof of Emission (POE)
Proof of Ownership (POO)
Proof of Transfer (POT)
Every digital bill in embodiments of this invention may thus be provided with a cryptographic proof that can be verified without relying on the trustworthiness of the operator of the service (for example, the central bank) or intermediaries. The correct operation of the system as a whole is also provable in real-time, which makes it secure against both inside and outside attacks on the integrity of the system and allows continuous mathematical verification of the total money supply, greatly reducing the cost of operations. In implementations in which KSI signatures are used as proofs, the only cryptographic primitive used in verification (of the money supply as a whole or individual bills) is a hash function (for example, SHA256 or SHA512), which means that the proofs are designed to withstand potential attacks by quantum computers.
Protocols
The main steps (“protocols”) of Printing (creating new digital bills), Emission, Payment, and Bill Ledger Adjustment carried out by the different entities (“parties”) in the system are summarized here.
Printing
Emission
Payment
Bill Ledger Adjustment Procedure (
Let L′ be the reduced bill ledger in Payer's wallet with blocks B0, B1, . . . , Bt′; let Bt be the new block of the bill ledger created by combining the payment order and the KSI signature obtained from the gateway that controls the respective bill; and let t>t′ be the time value in KSI signature.
The adjustment procedure involves the following steps:
Wallet Query
Users may wish or need to know what bills they control, that is, what bills are in their respective Wallets. This may be accomplished, via the API, by issuing a query to the output component 600. The query should then include both the owner's identifier, and the owner's signature, such as public key. The output component may then return a list of the bills associated with that public key. It would also be possible to specify time ranges for the list to be retrieved, or to request, for example, the list of only the n most recent changes in the user's wallet; this would also enable confirmation of a most recent transfer.
Prevention of Double Spending
A major concern in all systems that involve transfer of a single valid instance of a data set is “double spending”. In other words, a recipient (such as a payee) should be able to know that the sender (such as a payer) did not also transfer the same data set to some other recipient as well. In the context of digital cash, for example, a payee needs to be sure that the payer did not also give the same bill to another payee.
Embodiments of this invention have several mechanisms that can not only detect attempted double spending, but can prevent it from happening at all. According to one optional feature, transfer requests are received per input period, that is, with a cut-off time. As one example, transfer requests may be synchronized to some time base, such as KSI calendar periods. A ledger rule is then implemented such that no more than one signature may be requested per bill per calendar period. Alternatively, the ledger for each bill may be configured so as to accept no more than one request for update per settlement period, which may be the same as a KSI calendar period, or may be synchronized (to set a “cut-off”) to any other time system.
As described above, when a gateway has received a request to transfer ownership rights of a data unit, it updates the respective sub-ledger and (preferably) obtains (and may return to the payer) a digital signature confirming the transfer order. Already at that point, if the requester were to attempt to request yet another transfer of the same data unit, the request will be rejected, since the respective sub-ledger will already have been adjusted to indicate that the requester is no longer the owner of the data unit. Moreover, since, in this case, transfers “settle” at the end of a synchronization or period or programmed settlement period, the only way a payer could attempt to double-spend a bill would be to request yet another signature in another period. By that time, however, the earlier, valid transfer will already have been either completed by the gateway, or has enqueued it for transfer, in that the earlier signature will already have been entered into the bill ledger, along with the key of the new owner— by the time the payer attempts to double-transfer the bill, he will no longer be the owner and thus will be unable to do so. In practice, this means that, once a payer has issued a valid payment order and this has been acknowledged by the service, the payer no longer “owns” the bill and thus cannot spend it again.
Implementing settlement/update “periods” has the advantage of enabling aggregation of transfer requests, which then leads to an efficient use of signature mechanisms (such as KSI) that create signatures for groups of inputs during such periods. It would be possible, however, not to implement such periods at all. Gateways could, for example, process transaction order immediately, as fast as possible, and use other non-aggregating signature schemes.
No Inter Sub-Ledger Communication Necessary
One advantage of embodiments of this invention is that decomposition of the blockchain into sub-ledgers as disclosed here does not require inter-shard, that is, inter sub-ledger communication and coordination. Because the ledgers' state is divided into sub-units that don't have cross-dependency, there is no need to know state of any other data unit, or consult the state of any other sub-ledger, in order to process and complete a transfer of a given data unit. This contrasts with prior art systems such as Ethereum, in which there must be inter-ledger coordination, because its data “units” are transferred from one ledger to another. Embodiments of this invention do not require such cross-chain interaction, that is, inter-node communication, and thus, in the described cases, greatly increase system efficiency, speed, and, especially, scalability.
Owner-Held Bill Ledgers
In most of the description above, and in
Ledger Uniqueness and Auditability
There should preferably be a guarantee that only one instance of the ledger data structure exists. One way to ensure this is to allow only one transaction of an asset (or part of asset) per round, with the ledger being signed (and optionally also timestamped, such as with a KSI signature) upon either each change, or at the end of each round, thereby “sealing” its state at that time.
The output component associated with each respective gateway may be made accessible to an external auditing entity. Consider the threat that a sender collaborates with its gateway to create multiple versions of a ledger, or even “lend out” some of the sender's data units “off the books”. Each recipient is preferably associated with a specific output component, and each such component is accessible, so that the auditing entity may query all or some sub-set of the output components, for example at random times. Each output component may then respond with the state of its associated account, along with the corresponding gateway proofs, at those times. Over time, as the number of “query points” increases, the probability of successfully concealing “lending” out gets smaller.
As for transparency, outside parties should preferably not be enabled to see entire ledgers, but at most only individual blocks within it. On the other hand, a central authority or auditor may be enabled to inspect all ledgers. Note that this represents a difference relative to most existing blockchain systems, in which the trustworthiness of the system depends on all blocks being made public.
The embodiments described above provide a “permissioned” system, in which the central emitting entity ultimately controls not only emission of data units, but also forms the ultimate trust anchor of the system in that its public key is a trusted common reference string. As such, the central entity is the ultimate authority to operate gateways and issue bills, and, therefore, the ultimate authority to determine if a transfer order accepted or not. Using the permissioned embodiments, a central bank, for example, would be able to block even valid transfer requests. This is referred to as lacking “censorship resistance”.
An alternative “permissionless” aspect of the invention provides censorship resistance. According to this permissionless aspect, there is no need for the input component 500 to receive transfer requests, examine the identifier of the data unit to be transferred, and then route the request to the associated responsible gateway. Instead, the transferor itself has the information about the correct routing; this information may be encoded in the data unit itself, and will as before be a function of the data unit's identifier.
Rather than “a” gateway (which may be a cooperating cluster of platforms) associated with the sub-ledger corresponding to the data unit, there may instead be a group of validator “nodes”, and the transferor may submit the transfer request to one, several, or all of the associated nodes, which thus operate as a “logical gateway”.
The node(s) to which the request has been sent may circulate it to the other nodes, which may then choose to accept or reject it, and each node that has accepted it may include it in a respective proposed block. Following any known consensus mechanism, the nodes may then come to consensus about which block is to be created, that is, which transfer orders to include in the current round. A node may choose to reject a proposed block, for example, if the transferor's signature is invalid, or it would lead to too small of a “reward” (if included). A node that has proposed an invalid block may be “punished”, for example, being denied the opportunity to propose further blocks for some time, having some reward reduced or eliminated, etc.
As for “rewards”, one option would be to impose a transaction fee on transfers, which is given to nodes that successfully propose the corresponding block. A node may therefore, for example, reject a block because it views the transaction fee reward (or fees, from the accumulated transactions in the block) as being too small. In this aspect, no single party can unilaterally prevent a requested transaction.
Once the transfer request block is agreed upon, the requests may be processed according to the respective sub-ledgers and block ledgers as in the permissioned aspect. Copies of these may stored in corresponding validator nodes; alternatively, there could still be a single gateway that maintains the respective sub-ledgers and receives the agreed-upon transfer orders from the nodes after they have reached consensus. As such, certain features of the permissioned and permissionless aspects are the same, but the way in which transfer requests are input and routed is different.
The nodes may optionally operate in rounds, optionally synchronized to some time base, or defined by a minimum number of received transaction.
Transactions that have not yet been not processed/included in block may be stored in a common pool “mempool”, and a proposer node may choose transactions from that mempool if they are not sent directly to that node already. According to this aspect, decentralized nodes therefore replace “the” gateway server with a cluster of fault-tolerant validator nodes performing same the logic in parallel, with each of the validator nodes checking signatures independently and voting on whether a proposed block is acceptable or not.
Note important differences between this permissionless aspect and how a typical prior art blockchain forms blocks and uses consensus: No node is required to have a copy of the entire global blockchain, but rather creates only blocks relating to respective ones of the sub-ledgers, as indicated in the transfer requests. This means there is also no need for coordination and inter-communication among all nodes. Transactions associated with different blockchain shards may thus still be processed in parallel, which maintains the far superior ability of the different aspects of this invention to scale.
“Splitting”
There are situations in which a user may wish or need to transfer less than the entire value/quantity of a data unit the user owns. Just a couple examples of such situations include: a user owns a data unit with a monetary value of €50, but wishes to use it to buy something that costs only €30; and a manufacturer who has a container of 1000 units of a product, and thus is the current owner of the right to the whole 1000, wishes to split this into separate shipments of 600, 300, and 100. In these situations, the request that a transferor user submits to a node is not directed to a particular bill/data unit, but rather to an amount.
In the example illustrated in
In the illustration, the ownership of Bb0, which has a denomination of 20, is set to Ky, whereas the ownership of Ba0, with a denominated value of 30, remains set to Kx. As before, each block preferably also includes a digital signature (not shown, for simplicity) which, depending on the chosen method, may also function as a timestamp), as well as a current owner identifier (such as keys Kx, Ky, Kz).
This process may be carried out more than once. In the illustrated example, user Kx holds the ownership of the “sub-unit” valued at 30 for blocks Ba0 and Ba1, but then wishes to transfer only a value/amount v=25 to Kz. The node therefore “splits” Ba1 in the next round into Bx0, now owned by Kz and having a value of 25, and By0, whose owner remains Kx and whose value is the remainder 5. In
Although
Note that the blocks derived from B0 and the action of splitting may be maintained and carried out within the same node, operating on the same sub-ledger, as B0 itself; no communication with any other node is necessary, which has the added advantage that splitting operations for bills assigned to different shards may be carried out in parallel. In normal cases this may be done in a single transfer period just as any other transfer. Thanks to the sharding of the ledger, the splitting operation may thus be atomic. Moreover, note that the splitting of a bill does not change the total emitted quantity e.
By sacrificing this atomicity, it would, however, be possible to locate split sub-units in different shards than the current one, although this will require communication and appropriate security protocols between the current shard and the recipient shard, or with some higher-level agent/node/administrator/core entity.
Even when a data unit is split, it is preferably to maintain the relationship between its ID and the shard in which its sub-ledger is located.
See
If one cannot assume a low enough bound on the number of anticipated splits, another scheme would be to have multi-byte (or multi-word, etc.) IDs. A bit, for example, the least significant bit (LSB) could be used as a flag to indicate if there is additional appended ID information. Still another scheme would be for the node to maintain, for each bill, a linked list, table, tree, etc. storing the ID history, from emission onward, or at least of the current IDs used for the most recent generation of all sub-units of the original bill.
An optional design choice is to requires binary, that is, two-way, splits. One advantage of this is that the sub-ledger will then always be structured as a binary tree, such that appended bits could be used to indicate “left” or “right” (or, as shown in
Regardless of the method chosen for creating IDs of data units, because of the essential tree structure of the sharded bill ledger, an auditing entity may additionally easily check that no value has been created or destroyed by adding the current values indicated in all the most recent blocks; the sum should then be equal to value indicated in the initial block of the sub-ledger.
The system may optionally also “count-certify” the sub-ledger at any point, by iteratively pairwise hashing the most recent block values, that is, forming a hash tree of the indicated values, concatenated with the values themselves, to form a root value that represents the “total bill”. This option is illustrated in =h(x01,x23)∥16 is reached and can be used to certify the current value state of the blocks, that is, of all fractional bills created, of a sub-ledger.
Splitting may be initiated “as needed”. A transferor user requests a transfer to a transferee user for an amount specified in the request. In a basic splitting embodiment, only one bill owned by the transferor is involved, such that the bill's identity is specified in the request. Executable code within the corresponding node then determines if the current value of the bill is the same as the amount to be transferred. If it is, then the transfer can proceed as described above. If not, however, and the requested transfer amount is less than the value of the bill, then the node may initiate the required split. This would correspond approximately to the physical situation of presenting a specific paper note to a sell and then getting “change” for the excess value of the note.
In some other situations, the transferor may own more than one bill associated with the same node, and if the identity of the transferor is encoded identically for all bills the transferor owns (such as the same public key), then the node may be configured to be able to detect this through a simple scan, for example, of a table or list maintained of current owners and which bills they currently control. In such cases, in an alternate embodiment, the user submits a request for a transfer of a value, but the node may be configured to determine which bill(s) to transfer and/or split. Code may then be included in the node to which the value that identifies the bill(s) currently owned by the transferor user and may proceed as follows:
1) If there is a bill whose current value is exactly the requested amount then handle the single-bill transfer as described above. Else:
2) Applying any known algorithm, determine if there is a combination of transferor-owned bills in the node whose values add up to the requested amount. If so, then transfer those bills to the transferee as described above, that is, carry out the necessary set of single-bill transfers to satisfy the request and change ownership of those bills to the transferee. Else:
3) Select one or more transferor-owned bill(s) associated with the node whose value (or combined value) is greater than or equal to the requested amount. The selection criterion may be chosen by the system designer in different ways. Some examples include:
In general, the node may in this “splitting” embodiment use any known bin-packing or other selection algorithm to choose which bill(s) to select to satisfy a request. Note that this will be with variations, however, since the embodiment allows for splitting of a bill. Some of the selection criteria may employ any known “minimum coin” algorithm, again with the variation that bills may be split, but in some situations the system may seek the maximum number of bills to satisfy a transfer request. If not, however, the node applies any known routine to determine if there is a combination of transferor-owned bills whose values add up to the requested amount.
The different “splitting” embodiments may be used in both a permissioned and a permissionless system. In the case of a permissioned system, whatever central system is included may require at least notification that a node has carried out a data unit split so as to be able to keep track of at least the number of bills/data units in circulation, and possibly also to update any database it may maintain of the IDs and current denominations of bills in circulation.
This application is a continuation-in-part of and claims priority of U.S. patent application Ser. No. 17/194,303, filed 7 Mar. 2021, which claims priority of U.S. Provisional Patent Application No. 62/986,530, filed 6 Mar. 2020.
| Number | Date | Country | |
|---|---|---|---|
| Parent | 17194303 | Mar 2021 | US |
| Child | 17546905 | US |
