Method for providing related randomness for secure multi-party computing
By generating and distributing relevant randomness in a trusted execution environment, combined with confidential data processing technology, the problem of high offline stage in the existing secure multi-party computing protocol is solved, and low-cost and efficient secure multi-party computing is achieved, suitable for multi-party computing A kind of scenario.
Patent Information
- Application Number
- CN202410678824.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Priority Date
- 2023-05-30
- Filing Date
- 2024-05-29
- Publication Date
- 2024-12-03
Smart Images

Figure CN119070976A_ABST
Abstract
Description
Background Art
[0001] Modern protocols for secure multiparty computation (MPC) work according to the so-called preprocessing model. The result of the input-independent offline phase is a secret, shared correlated randomness (CR) in the form of tuples of subject elements, which are used in the input-dependent online phase to make operations such as multiplication more efficient. Existing offline phase protocols, also referred to below as Correlated Randomness Generators (CRGs), are based on heavyweight cryptographic designs such as homomorphic encryption or oblivious transfer. They suffer from high computational and communication complexity. Summary of the invention
[0002] A first general aspect of the present disclosure relates to a method for providing correlated randomness for secure multi-party computation using multiple computers.
[0003] The method includes
[0004] -Execute the Trusted Execution Environment system,
[0005] - Establish at least one relevant randomness generator in the trusted execution environment system,
[0006] - remote authentication of a unique associated randomness generator by all computers, or remote authentication of multiple associated randomness generators by one computer,
[0007] - at least one secret is injected into a uniquely associated randomness generator by all computers respectively, or at least one secret is injected into a respectively associated randomness generator by one computer,
[0008] - verifying the injected secret through one or more relevant randomness generators,
[0009] - Generates relevant randomness for the computer based on the injected secret,
[0010] -Distribute correlated randomness to computers, where each computer only gets the correlated randomness created for it.
[0011] A second general aspect of the present disclosure relates to a computer system designed to perform the method according to the first general aspect (or its embodiments).
[0012] A third general aspect of the present disclosure relates to a computer program designed to perform the method according to the first general aspect (or its embodiments).
[0013] A fourth general aspect of the present disclosure relates to a computer-readable medium or signal storing and / or containing a computer program according to the third general aspect (or its embodiments).
[0014] The techniques of the first, second, third and fourth general aspects may, in some cases, have one or more of the following advantages.
[0015] The present disclosure enables low-cost secure multi-party computation in a preprocessing model with active security and dishonest majority by replacing resource-intensive existing methods based on advanced cryptography with methods that use confidential data processing (Confidential Computing, CC) techniques at low cost, either independently or in combination with more efficient active security honest majority MPC protocols.
[0016] The total cost of secure multi-party computation (MPC) in the preprocessing model is largely dominated by the cost of the offline phase (this factor is up to an order of magnitude). The present disclosure uses confidential data processing (Confidential Computing, CC) techniques to significantly reduce the cost for the offline phase. These techniques rely on security mechanisms that can be supported by hardware devices to ensure the confidentiality and integrity of sensitive workloads.
[0017] The present disclosure enables the trade-off of performance for the strength of security guarantees and covers a range of usage scenarios with different usage complexity, security and performance characteristics. These can reduce the cost of MPC in the preprocessing model by at least an order of magnitude (for Figure 4 The per-party TEE-MPC CRG variant shown in is implemented in a way that the trust assumption is strengthened and security guarantees are generated either through formally provable cryptographic guarantees or through hardware and trust guarantees. In this case, this is to generate a secret shared correlated random value.
[0018] The present invention is superior to the following solutions:
[0019] - Direct use of confidential data processing CC, i.e., in the absence of secure multi-party computation MPC, for securing workloads: By integrating complex workloads into a trusted execution environment (TEE), a large trusted data processing base (TCB) emerges, which may provide a large attack surface for side-channel attacks. According to the present disclosure, the trusted execution environment TEE contains relatively small workloads with little complexity. That is, logic for locally generating (one or more) relevant randomness CRs, or a medium-complexity, medium-sized, special, actively secure, honest majority secure MPC protocol implementation for generating (one or more) relevant randomness CRs.
[0020] -Use TEE-protected, actively secure, general-purpose honest majority MPC protocol for securing workloads: The complexity and scope of the general-purpose honest majority MPC protocol are significantly larger than the protocol specifically tailored for generating correlated randomness CR according to the present disclosure.
[0021] The present disclosure enables implementations with different tradeoffs in computational / communication complexity and security.
[0022] Certain terms are used in this disclosure in the following manner.
[0023] "Secure multi-party computation" (MPC) is a cryptographic protocol that distributes computations across multiple parties, where each party cannot see the data of the other parties.
[0024] The term "MPC party" includes users of the secure multi-party computation MPC, such as computers, network users, etc.
[0025] A "Trusted Execution Environment" (TEE) is an environment for executing code where those executing the code can have a high degree of trust in the environment because they can ignore threats from the rest of the device. A TEE typically protects the confidentiality and integrity of the workload. The terms TEE and Confidential Computing (CC) are used synonymously in this document.
[0026] Remote Attestation (RA) is a CC technology that a system uses to authenticate its hardware and software configuration to a remote system. The purpose of remote attestation is to give the challenger the possibility to determine the degree of trust in the integrity of the prover's platform.
[0027] A “Correlated Randomness Generator” (CRG) is a program that implements an algorithm that generates correlated randomness for secret shares in the form of so-called tuples.
[0028] In the scope of this disclosure, the term "correlated randomness generator" is used to denote a specific implementation of a correlated randomness generator with the following parameters: the type of tuples to be generated, the number of tuples to be generated, the seed used to initialize the pseudo-random number generator (PRNG), and the components of the message authentication code (MAC) key of each MPC party of the MPC system (hereinafter referred to as m_i, where i is the index of the corresponding MPC party). The seed for the correlated randomness generator is created in a trusted execution environment that also executes the correlated randomness generator.
[0029] A MAC is needed in order to achieve active security in an environment with malicious users. It should be noted that the MAC key can either be generated by the CRG upon request or can be collected by the parties for whom the relevant randomness is generated. In the first case, an existing re-sharing protocol must be used, which adapts the tuple component of the relevant randomness to the MAC key of the receiving MPC system. In the following, the latter case is considered for readability reasons, although both cases are included by the present disclosure. The relevant randomness generator initializes the PRNG with the described seed in order to calculate the secret component of the requested CR tuple for all players. The relevant randomness generator is either instantiated as a regular program or as an MPC program in the variant described here. It should be noted that when the CRG is executed in a TEE, the secure PRNG functions implemented by the TEE are used (e.g. sgx_read_rand in the case of Intel SGX). BRIEF DESCRIPTION OF THE DRAWINGS
[0030] Figure 1 A system for providing correlated randomness for secure multi-party computation according to the present disclosure is schematically illustrated.
[0031] Figure 2 A system for providing correlated randomness for secure multi-party computation according to the present disclosure is schematically illustrated.
[0032] Figure 3 A system for providing correlated randomness for secure multi-party computation according to the present disclosure is schematically illustrated.
[0033] Figure 4 A system for providing correlated randomness for secure multi-party computation according to the present disclosure is schematically illustrated.
[0034] Figure 5is a flow chart illustrating a technique of the present disclosure for providing correlated randomness for secure multi-party computation. DETAILED DESCRIPTION
[0035] exist Figures 1 to 4 Various examples of systems 10 for providing correlated randomness for secure multi-party computation according to the present disclosure are shown in . All systems 10 include at least one trusted execution environment, in which a correlated randomness generator is set up.
[0036] exist Figure 1 A single trusted execution environment (TEE) with an associated randomness generator (CRG), also known as a single TEE-CRG, is shown in FIG.
[0037] exist Figure 2 , each party has a trusted execution environment (TEE) with a related randomness generator (CRG), that is, each party TEE-CRG.
[0038] exist Figure 3 , a single trusted execution environment (TEE) with a correlated randomness generator (CRG) is shown, which includes an honest-majority MPC protocol between multiple TEE-protected subsystems, i.e., a single TEE-MPC-CRG.
[0039] exist Figure 4 , a single trusted execution environment (TEE) with a correlated randomness generator (CRG) is shown, which includes an honest-majority MPC protocol between multiple TEE-protected subsystems on each side, i.e., TEE-MPC-CRG on each side.
[0040] The following compares the complexity of each field (Trusted Computing Base (TCB)).
[0041] The complexity of the communication is divided into the categories Local Area Network (LAN) and Wide Area Network (WAN). The first category is the low-cost communication with low latency and high bandwidth via the LAN within the same computing center (on-site or in the cloud), for example with μs latency and 100Gbit / s. The second category is the more expensive communication with high latency and low bandwidth via the WAN, for example with ms latency and 1Gbit / s, between different computing centers, such as multiple computing centers on-site, between multiple regions of the same public cloud, or between computing centers in different public clouds. Below, the LAN communication channel is shown as a solid line, while the WAN communication channel is shown as a dotted line. It should be noted that the WAN communication complexity of each variant is independent of the number of CR tuples generated.
[0042] The following symbols are used:
[0043] n refers to the number of MPC parties in the delegated MPC system,
[0044] t represents the number of tuples to be generated.
[0045] Note that n is allowed to be in the low single-digit range, typically between 2 and 5, while t is in the order of billions. For the case where the CRG itself is implemented via MPC (using an honest majority protocol):
[0046] m refers to the number of MPC parties in the CRG MPC system,
[0047] P(n,m,t) is the runtime complexity of the relevant MPC protocol,
[0048] C(n,m,t) is the corresponding communication complexity.
[0049] For a single TEE-CRG, the runtime complexity is in the order of O(nt), the complexity of LAN communication is in the order of O(t), and the complexity of WAN communication is in the order of O(nt). The security mechanism is implemented based on a single TEE. Since only a simple implementation of the CRG logic is performed, the size of the trusted data processing base is small.
[0050] For each TEE-CRG, the runtime complexity is in the order of O(n2 t), the complexity of LAN communication is in the order of O(nt), and the complexity of WAN communication is in the order of O(n2). The security mechanism is based on a single TEE implementation per computer. Since only a simple implementation of the CRG logic is performed, the size of the trusted data processing base is small.
[0051] For a single TEE-MPC-CRG, the runtime complexity is of the order of P(n,m,t), the complexity of LAN communication is of the order of O(m t+C(n,m,t)) and the complexity of WAN communication is of the order of O(mnt). The security mechanism is implemented based on multiple TEEs. Since a specialized MPC protocol implementation is used for CR generation, the size of the trusted data processing base is moderate.
[0052] For each TEE-MPC-CRG, the runtime complexity is in the order of O(n P(n,m,t)), the complexity of LAN communication is in the order of O(nm t+n C(n,m,t)), and the complexity of WAN communication is in the order of O(m 2 n2). The security mechanism is implemented based on multiple TEEs. Since a professional MPC protocol is used for CR generation, the size of the trusted data processing base is medium.
[0053] Figure 1 A system 10 for providing correlated randomness for secure multi-party computation according to the present disclosure is schematically shown, wherein the system includes a trusted execution environment with a correlated randomness generator, namely, a single TEE-CRG.
[0054] This can be considered as the simplest variant of the proposed system 10. Here, the trusted execution environment system 20 exists in a unique trusted execution environment, in which a unique associated randomness generator 21 is executed. Two computers 12 and 14 are shown as examples. These computers are parties of the system 10 for secure multi-party computation. There can be multiple parties or computers in the system 10, for example two to five. The system 10 can be a component of a cloud system or a cloud system itself. In particular, the system 10 can be a cloud stack or a part thereof for building a scalable application for secure multi-party computation.
[0055] The process of system 10 can be represented as follows. One of computers 12 and 14 or an external unit starts a process or VM-based trusted execution environment 20. Trusted execution environment 20 can be physically located at one of computers 12 and 14 or at another location. Figure 1 In the case shown in FIG. 1 , computer 12 will start and possibly execute trusted execution environment 20 because the two entities are connected via one communication connection 13 in the form of a LAN connection with low latency and high bandwidth. Another computer 14 is connected to trusted execution environment 20 via another communication connection 15 in the form of a WAN connection with high latency and low bandwidth. Thus, for example, in Figure 1 In the figure, the part on the left side of the vertical dashed line can be in a computer, a rack, a server, a computing center or a local area. The right part is far away from this and can be reached via the Internet, for example.
[0056] Within the trusted execution environment 20, a relevant randomness generator 21 is used. This can be done as a compiled binary or interpreted code. Public parameters, such as arithmetic parameters such as bases, primes, type and number of query tuples, can be hard-coded into the logic or provided as arguments to the hosting environment. For example, this can be done via environment variables.
[0057] Both MPC computers 12 and 14 perform remote attestation to establish trust in the associated randomness generator 21 protected in the trusted execution environment 20. If the remote attestation is successful, then the computers 12 and 14 "know" that the hardware and / or software state of the associated randomness generator 21 protected in the trusted execution environment 20 is as expected. In this regard, the computers 12 and 14 may optionally explicitly verify the common configuration provided in the previous step and abort the process if the configuration does not correspond to expectations.
[0058] Computers 12 and 14 inject relevant secret parameters, for example secrets in the form of their MAC key components, into relevant randomness generators 21 protected in trusted execution environment 20. This can be performed using mechanisms for injecting secrets available in remote attestation systems.
[0059] The protected associated randomness generator 21 in the trusted execution environment 20 checks whether the parameters are normal and compatible. This may also include checking the correctness of the MAC key. This check may be, for example, a byte check or a rationality check.
[0060] The logic of the relevant randomness generator 21 starts and starts to generate relevant randomness according to the input secret and optionally additionally according to the configuration. The components of the relevant randomness for all computers 12 and 14 are generated by the relevant randomness generator 21 protected in the trusted execution environment 20. If a PRNG is used for the generator, a seed must still be generated in the trusted execution environment (for example, sgx_read_rand() in the case of Intel SGX). The generator can generate a seed or agree on a seed with other generators. Here, the difference between a random material generator and a random number generator is that a random number generator requires a seed in any case, while a random material generator does not require a seed.
[0061] The generated secret components of the relevant randomness are either transmitted in batches to the respective computers 12 and 14 at the end of the generation process, or transmitted via a streaming mechanism in parallel with the generation process. Each party only obtains the relevant randomness components created for it.
[0062] Figure 2 A system 10 for providing relevant randomness for secure multi-party computation according to the present disclosure is schematically shown, wherein the system comprises a trusted execution environment for each party, and the trusted execution environments respectively have a relevant randomness generator, namely, each party TEE-CRG.
[0063] Here, the trusted execution environment system 20 consists of two trusted execution environments 22 and 24, in which the relevant randomness generator 21 is executed respectively. Two computers 12 and 14 are shown as examples. The provision of two trusted execution environments here is only exemplary; in the case of TEE-CRG per party, there are exactly as many trusted execution environment systems as parties in the system 10. These computers are the parties of the system 10 for secure multi-party computing. There can be multiple parties or computers in the system 10, for example two to five. The system 10 can be a component of a cloud system or a cloud system itself. In particular, the system 10 can be a cloud stack or a part thereof for building a scalable application for secure multi-party computing.
[0064] The basic idea of this variant is to set up a separate collocated trusted execution environment 22 or 24 for each of the MPC parties, in which the logic of the respective randomness generator 21 is respectively located.
[0065] In this case, the process can be as follows. Each MPC party or computer 12, 14 starts a trusted execution environment 22, 24 based on a process or VM. Alternatively, this can also be done by an external device. The trusted execution environments 22, 24 are physically located in a common location or at least near the corresponding MPC parties in terms of network delays, and should therefore provide low costs and high transmission bandwidth. According to the terminology introduced above, the MPC parties use LAN channels 13, 15 to communicate with "their" associated randomness generators 21 protected in the corresponding trusted execution environments 22, 24. The associated randomness generators 21 can use the WAN channel 23 to communicate with each other. The logic of the associated randomness generators 21 is used in the trusted execution environments 22, 24, respectively. This may be the case, for example, as compiled binary code or as interpreted code.
[0066] Each MPC user performs remote attestation to establish trust in the associated randomness generator 21 protected in the trusted execution environment 22, 24. If the remote attestation is successful, the computers 12 and 14 "know" that the hardware and / or software state of the associated randomness generator 21 protected in the trusted execution environment 22, 24 is as expected. Public parameters, such as arithmetic parameters such as bases, primes, type and number of query tuples, etc., can be hard-coded into the logic or provided as arguments for the managed environment, for example, via environment variables.
[0067] Computers 12 and 14 inject relevant secret parameters, such as secrets in the form of their MAC key components, into the corresponding relevant randomness generators 21 protected in the corresponding trusted execution environments 22, 24. This can be performed using the mechanisms for injecting secrets available in the remote attestation system. If the remote attestation fails, the secret parameters are not injected and the process is aborted.
[0068] Each relevant randomness generator 21 protected in a trusted execution environment 22, 24 verifies that the parameters are correct and compatible. For example, this may include identical MAC keys, prime numbers, etc. The trusted execution environment 22, 24 verifies that its configuration is compatible and correct by means of remote attestation.
[0069] The trusted execution environments 22, 24 securely exchange MAC key components and optionally verify their non-disputability using secret injection provided by remote attestation. Alternatively, the MPC party can inject its MAC key share into all trusted execution environments.
[0070] The associated protected randomness generators 21 in the trusted execution environments 22, 24 agree on a common value, which is later used as a seed by the random number generator, and a unique user number for each of them. As explained previously, this is again done using remote attestation and secure injection of secrets.
[0071] The specific solution for ensuring that the same seed and clear user number are used is to let the relevant randomness generator 21 decide the allocation of seeds and user numbers. Either the generator can make the seeds itself for this purpose, or there is another auxiliary system of the trusted execution environment. Since mutual trust in the host trusted execution environment 22 or 24 has been established previously by remote attestation, the relevant randomness generator 21 can be sure that the allocation list is valid, that is, each user has a clear user number and each relevant randomness generator 21 uses the allocated user number and the specified seed.
[0072] The logic of the relevant randomness generator 21 starts and starts to generate relevant random numbers according to the parameters. In particular, the same seed is used for the random number generator used to create the relevant randomness. Therefore, all relevant randomness generators 21 protected in the trusted execution environment 22, 24 locally generate exactly the same relevant randomness for all parties. The agreed allocation of user numbers ensures that each of the relevant randomness generators 21 protected in the trusted execution environment 22, 24 only provides the corresponding secret component of the relevant randomness to the corresponding party. The corresponding party is the party that has started or hosted the trusted execution environment.
[0073] The generated secret components of the associated randomness are either transmitted to the corresponding parties in batches at the end of the generation process or transmitted via a streaming mechanism in parallel with the generation process.
[0074] Figure 3 A system 10 for providing correlated randomness for secure multi-party computation according to the present disclosure is schematically shown, wherein the system includes a trusted execution environment system 20 with a correlated randomness generator, wherein the trusted execution environment system 20 includes an honest-majority MPC protocol (honest-majority) between multiple TEE-protected subsystems, that is, a single TEE-MPC-CRG.
[0075] Here, the trusted execution environment system 20 is composed of, for example, three trusted subsystems 26, 28, and 30, each of which has a trusted execution environment, and in which a related randomness generator 21 is executed. The related randomness generator 21 shown can be regarded as an example. In fact, the trusted subsystems 26, 28, and 30 of the trusted execution environment system 20 can be configured as a generator of an honest majority protocol.
[0076] More than three subsystems, for example up to ten or five subsystems are possible. The number of subsystems does not have to be related to the number of computers. One or more computers themselves can be part of the generator.
[0077] Two computers 12 and 14 are shown as examples. These computers are parties of the system 10 for secure multi-party computation. There may be multiple parties or computers in the system 10, for example two to five. The system 10 may be a component of a cloud system or a cloud system itself. In particular, the system 10 may be a cloud stack or a part thereof for building a scalable application for secure multi-party computation.
[0078] The basic idea of the scheme is to use more efficient active-safe honest majority MPC protocols for generating correlated randomness between multiple protected subsystems 26, 28, 30 of a trusted execution environment. These protocols do not require an offline phase and have up to two orders of magnitude smaller computational complexity and up to three orders of magnitude smaller communication complexity than active-safe dishonest majority MPC protocols.
[0079] MPC users are used in close proximity to each other using low latency and high bandwidth LAN channels 13 and 23. The relatively small secret component of the generated correlation randomness is transmitted via a potentially high latency and low bandwidth WAN communication channel 15 to all users except the only computer working with the subsystem 26, 28, 30 protected by the trusted execution environment.
[0080] In this variant and the following variants, it is provided that a different implementation of a trusted execution environment is used for each of the relevant randomness generators 21 protected by subsystems 26, 28, 30. This means that an attacker must simultaneously compromise the security of more than half of the participating trusted execution environments or subsystems 26, 28, 30 in order to overcome the security of system 10.
[0081] In this case, the flow may be as follows. An MPC user or computer 12, 14 or external entity starts the required number of process or VM-based subsystems 26, 28, 30, each with an associated randomness generator 21 for protection in a trusted execution environment 20. These subsystems 26, 28, 30 are themselves MPC users. All trusted execution environments 26, 28, 30 are physically located together, for example in the same computing center, and are located near one of the computers (here computer 12). Being near involves having to do with low network latency. There should be a transmission via LAN channels 23 and 13 that utilizes high bandwidth.
[0082] The protocol logic of the relevant randomness generator 21 is used within the trusted execution environment 26, 28, 30 respectively. This can be done as compiled binary code or as interpreted code. The logic consists of a special, actively secure, honest-majority MPC protocol implementation for generating relevant randomness. Public parameters, such as arithmetic parameters such as base, prime numbers, type and number of tuples requested, can be hard-coded in the logic or provided as arguments to the hosting environment, such as via environment variables.
[0083] Each MPC party or computer 12, 14 performs remote attestation to establish trust in all relevant randomness generators 21 protected in the corresponding trusted execution environment 26, 28, 30. Figure 3 In the situation shown in , the computer 12 performs remote authentication for the relevant randomness generator 21 protected in the corresponding trusted execution environment 26, 28, 30, that is, the corresponding MPC subsystem. If the remote authentication is successful, the computer 12, 14 "knows" that the hardware and / or software status of the relevant randomness generator 21 protected in the trusted execution environment 26, 28, 30 is as expected. If the remote authentication fails, the process is terminated.
[0084] Using secret sharing, computers 12 and 14 inject relevant secret parameters, for example secrets in the form of their MAC key components, into relevant randomness generators 21 protected in trusted execution environments 26 , 28 , 30 (MPC subsystem).
[0085] For the MAC key component and assuming an associated randomness generator 21 protected in a trusted execution environment 26, 28, 30 with m MPC subsystems, each computer 12, 14 divides its MAC key component into m components and sends one of the components to each MPC subsystem in the form of an associated randomness generator 21 protected in the trusted execution environment. This means that none of the MPC subsystems in the form of an associated randomness generator 21 protected in the trusted execution environment possesses the secret full MAC key of the MPC party or can reconstruct the secret full MAC key. The security guarantee continues to apply even if m / 2-1 parties are compromised due to a compromised implementation of the trusted execution environment.
[0086] The MPC subsystems or their trusted execution environments use remote attestation with each other to verify that their configurations are compatible and functioning properly. This verification is performed over the LAN channel 23.
[0087] In the MPC subsystem in the form of a trusted execution environment, the separately protected correlation randomness generator 21 starts generating correlation randomness using the previously described configuration parameters as public input and (at least) the secret common MAC key as private input. Here, the generation of correlation randomness is implemented as an MPC program working with secret common data.
[0088] Therefore, if at least m / 2+1 parties are honest, i.e. the underlying implementation of their trusted execution environment has not been compromised, none of the MPC subsystems in the form of the associated randomness generator 21 protected in the trusted execution environment can read the generated associated randomness. The result of the generation process is the secret component of the MPC subsystem in the form of the associated randomness generator 21 protected in the trusted execution environment in the secret component of the associated randomness for the interrogating computer.
[0089] The secret components of the relevant randomness for the querying computer are transmitted to the corresponding computers 12 and 14 either in batches at the end of the generation process or in parallel with the generation process via a streaming mechanism. To this end, each relevant randomness generator 21 protected in the trusted execution environment sends all components of the specific MPC party for each tuple component of the relevant randomness to the corresponding computers 12 and 14.
[0090] Figure 4 A system 10 for providing correlated randomness for secure multi-party computation according to the present disclosure is schematically shown, wherein the system includes a trusted execution environment system 20 having a correlated randomness generator, wherein the correlated randomness generator includes an honest-majority MPC protocol between multiple TEE-protected subsystems of each party, i.e., a TEE-MPC-CRG for each party.
[0091] Here, the trusted execution environment system 20 consists of two trusted execution environments 32 and 34, each of which consists of a plurality of trusted subsystems 26, 28, 30, each of which has a trusted execution environment in which an associated randomness generator 21 is executed. More than three subsystems, for example up to ten or five subsystems, are possible. The number of subsystems does not have to be related to the number of computers.
[0092] Here, the trusted execution environments 32 and 34 are composed of, for example, three trusted subsystems, each of which has a trusted execution environment, in which the relevant random generator 21 (such as Figure 3 ). Therefore, Figure 4 The system 10 shown in FIG. 1 can be viewed as Figure 2 and 3 A combination of the systems shown in .
[0093] Two computers 12 and 14 are shown as examples. These computers are parties of the system 10 for secure multi-party computation. There may be multiple parties or computers in the system 10, for example two to five. The system 10 may be a component of a cloud system or may itself be a cloud system. In particular, the system 10 may be a cloud stack or a part thereof for building a scalable application for secure multi-party computation.
[0094] The basic idea of the solution is to use at each computer 12 , 14 an instance of a trusted execution environment 32 , 34 with an MPC subsystem in the form of an associated randomness generator protected in the trusted execution environment.
[0095] Here, the MPC subsystem in the form of a protected correlated randomness generator in the trusted execution environment also agrees on a common seed for the PRG used to generate correlated randomness. In this way, all correlated randomness generators 21 of the MPC subsystem generate the same correlated randomness. Figure 3 Compared to the system shown in , here the correlated randomness generated is not transmitted via an expensive WAN communication channel with high latency and low bandwidth.
[0096] Each MPC party or each computer 12, 14 or external unit starts a number m of process-based or VM-based trusted execution environments. These trusted execution environments constitute MPC subsystems, respectively, and the MPC subsystems respectively have associated randomness generators 21 protected in the trusted execution environment. The trusted execution environments for the same MPC subsystems respectively having associated randomness generators 21 protected in the trusted execution environment are physically located near the corresponding computers 12, 14 or at least near the corresponding computers 12, 14 with respect to network delays, and should be able to achieve transmission with high bandwidth.
[0097] The protocol logic of the relevant randomness generator 21 is used separately within the trusted execution environment. This can be done as compiled binary code or interpreted code. The logic consists of a special active security honest majority (honest-majority) MPC protocol implementation for generating relevant randomness. Public parameters, such as arithmetic parameters such as base, prime numbers, type and number of tuples requested, can be hard-coded in the logic or provided as arguments to the hosting environment, such as via environment variables.
[0098] Each MPC party or computer 12, 14 performs remote authentication to establish trust in all relevant randomness generators 21 protected in the MPC subsystem in the corresponding trusted execution environment 32, 34. Figure 4In the scenario shown in , the computer 12 performs remote authentication for the associated randomness generator 21 protected in the trusted MPC subsystem. If the remote authentication succeeds, the computer 12, 14 "knows" that the hardware and / or software state of the associated randomness generator 21 protected in the trusted execution environment 32, 34 is as expected. If the remote authentication fails, the process is aborted.
[0099] Using secret sharing, computers 12 and 14 inject relevant secret parameters, for example secrets in the form of their MAC key components, into respectively relevant randomness generators 21 protected in trusted execution environments 32 , 34 (MPC subsystems).
[0100] For the MAC key components and assuming an associated randomness generator 21 protected in a trusted execution environment 32, 34 (MPC subsystem) with m MPC subsystems, each computer 12, 14 divides its MAC key component into m components and sends one of the components to each MPC subsystem in the form of an associated randomness generator 21 protected in the trusted execution environment. This means that none of the MPC subsystems in the form of an associated randomness generator 21 protected in the trusted execution environment possesses the secret full MAC key of the MPC party or can reconstruct the secret full MAC key. The security guarantee continues to apply even if m / 2-1 parties are compromised due to a compromised implementation of the trusted execution environment.
[0101] All parties protected in a trusted execution environment with the associated randomness generator 21 of MPC subsystem #i (e.g., execution environment 32) perform remote attestation for all parties protected in a trusted execution environment with the associated randomness generator 21 of MPC subsystem #j (e.g., execution environment 34), i.e., for all combinations i, j = 1, ..., n, where i! = j. If the remote attestation fails, the process is aborted.
[0102] A common secret seed value for secret pseudo-random number generation is generated for all relevant randomness generators 21 protected in a trusted execution environment 32, 34 (MPC subsystem) as follows: Each MPC subsystem #i has m protected generators G_i,k. At the end, for each k, the protected generators G_1,k, .... G_n,k must have a common seed (and only these generators). Similar to the MAC part, the exchange occurs at the generator level, not at the MPC subsystem level.
[0103] Since assigning user numbers to MPC subsystems protected in a trusted execution environment is not sensitive, the following mechanism can be used (e.g., according to Figure 2) for allocation. To ensure that the same seed and clear user number are used, the relevant randomness generators will determine the allocation of seeds and user numbers. Since mutual trust in the host trusted execution environment has been previously established through remote attestation, the relevant randomness generators can be confident that the allocation list is valid, that is, each user has a clear user number, and each relevant randomness generator uses the allocated user number and the specified seed.
[0104] To generate relevant randomness, each MPC subsystem protected by a trusted execution environment executes Figure 3 The MPC protocol described by the system shown in . A common secret shared seed for the MPC-based random number generator is used to generate correlated randomness. The seed is common between the MPC subsystems and secret between the generators. Therefore, all MPC subsystems protected by the trusted execution environment produce exactly the same correlated randomness. The agreed-upon assignment of user numbers ensures that each MPC subsystem protected by the trusted execution environment only provides the corresponding secret component of the correlated randomness to the corresponding party.
[0105] The secret components of the generated associated randomness are either transmitted to the corresponding parties in a loose sequence at the end of the generation process or transmitted via a streaming mechanism simultaneously with the generation process. The only interaction occurs between the MPC parties or computers and the MPC subsystems protected by the corresponding associated trusted execution environment.
[0106] Figure 5 is a flow chart illustrating the disclosed technique for providing correlated randomness for secure multi-party computation. In particular, the flow chart shows a computer-implemented method 50 for providing correlated randomness for secure multi-party computation using multiple computers.
[0107] According to the method 50, firstly, a trusted execution environment system 20 is executed 51. The execution environment system 20 may correspond to, for example, Figures 1 to 4 One or a combination of the execution environment systems 20 shown in .
[0108] Then, at least one relevant randomness generator 21 is set up 52 in the trusted execution environment system. In this case, at most one relevant randomness generator 21 is provided in a trusted execution environment.
[0109] Remote authentication of uniquely correlated random generators by all computers 53 (e.g. Figure 1 and 3 ), or remote authentication of multiple related random generators by a computer (such as in Figure 2 and 4 in the system).
[0110] Furthermore, at least one secret is respectively injected 54 into a uniquely associated randomness generator by all computers or at least one secret is respectively injected 54 into one associated randomness generator by one computer.
[0111] The injected secret is then checked 55 by one or more associated randomness generators. This may, for example, include a byte check or a plausibility check.
[0112] Furthermore, the associated randomness is generated 56 for the computer based on the injected secret. Configuration parameters such as arithmetic parameters like base, prime numbers, type and number of query tuples may also be taken into account during the generation.
[0113] Finally, the relevant randomness is distributed 57 to the computers, where each computer only gets the relevant randomness created for it.
[0114] According to one specific embodiment, the method further comprises: executing the trusted execution environment system as a single trusted execution environment and establishing a corresponding communication connection between the computer and the trusted execution environment.
[0115] According to one specific embodiment, the method further comprises: executing the trusted execution environment system as a trusted execution environment for each computer, establishing a corresponding communication connection between the computer and the trusted execution environment, and establishing a communication connection between the trusted execution environments.
[0116] According to one specific embodiment, the method further comprises: the trusted execution environments exchanging secrets and / or information about remote attestation, and the trusted execution environments agreeing on a seed for an associated randomness generator based on the secrets and / or the information about remote attestation.
[0117] According to one embodiment, the method further comprises executing each trusted execution environment into a plurality of subsystems, each of which has a trusted execution environment with an associated randomness generator, wherein corresponding communication connections are established between the computer and all subsystems, and wherein communication connections are established between the subsystems.
[0118] According to one specific embodiment, the method further comprises that the plurality of subsystems communicate using a secure MPC protocol.
[0119] According to one specific embodiment, the method furthermore comprises that the plurality of subsystems communicate via a LAN connection.
[0120] According to one embodiment, the method further comprises remote mutual authentication of the plurality of subsystems.
[0121] According to one specific embodiment, the method further comprises at least one computer starting a trusted execution environment.
[0122] According to one specific embodiment, the method further comprises that the booting computer has a LAN connection to the trusted execution environment.
[0123] According to one specific embodiment, the method further comprises: the secret comprises a key component of a message authentication code MAC.
[0124] According to one embodiment, the method further comprises: the secret and / or the seed being kept confidential in the trusted execution environment system.
Claims
1. A computer-implemented method (50) for providing correlated randomness for secure multi-party computation using a plurality of computers (12, 14), wherein the method (50) comprises the following steps: - executing (51) a trusted execution environment system (20), - establishing (52) at least one relevant randomness generator (21) in the trusted execution environment system (20), - remote authentication (53) of a single associated randomness generator (21) by all computers (12, 14), or remote authentication (53) of a plurality of associated randomness generators (21) by one computer (12), respectively, - injecting (54) at least one secret into a uniquely associated randomness generator (21) via all computers (12, 14) or injecting (54) at least one secret into a respective associated randomness generator (21) via one computer (12), - verifying (55) the injected secret by one or more associated randomness generators (21), - generating (56) relevant randomness for said computers (12, 14) according to the injected secret, - distributing (57) the relevant randomness to the computers (12, 14), wherein each computer (12) only receives the relevant randomness created for it.
2. The method (50) according to claim 1, wherein the trusted execution environment system (20) is implemented as a single trusted execution environment (20) and wherein a corresponding party communication connection (13, 15) is established between the computer (12, 14) and the trusted execution environment (20).
3. The method (50) according to claim 1, wherein for each computer (12), the trusted execution environment system (20) is executed as a trusted execution environment (22, 24), wherein a corresponding party communication connection (13, 15) is established between the computer (12, 14) and the trusted execution environment (22, 24), and wherein a communication connection is established between the trusted execution environments (22, 24).
4. A method (50) according to claim 2 or 3, wherein the trusted execution environments (22, 24) exchange secrets and / or information about remote authentication, and wherein the trusted execution environments (22, 24) agree on a seed for an associated randomness generator (21) based on the secrets and / or information about remote authentication.
5. A method (50) according to any one of claims 2 to 4, wherein each trusted execution environment (20) is implemented as a plurality of subsystems (26, 28, 30), each of which has a trusted execution environment with an associated randomness generator (21), wherein a corresponding party communication connection (13) is established between the computer (12) and all subsystems (26, 28, 30), and wherein a communication connection is established between the subsystems (26, 28, 30).
6. The method (50) of claim 5, wherein the plurality of subsystems (26, 28, 30) communicate using a secure MPC protocol.
7. The method (50) of claim 5 or 6, wherein the plurality of subsystems (26, 28, 30) communicate via a LAN connection.
8. The method (50) according to any one of claims 5 to 7, wherein the plurality of subsystems (26, 28, 30) remotely authenticate (53) each other.
9. The method (50) according to any one of claims 2 to 8, wherein at least one computer (12) starts a trusted execution environment (20).
10. The method (50) of claim 9, wherein a booting computer (12) has a LAN connection to the trusted execution environment (20).
11. The method (50) according to any one of claims 1 to 10, wherein the secret comprises a key component of a message authentication code (MAC).
12. The method (50) according to any one of claims 1 to 11, wherein the secret and / or seed is kept confidential in the trusted execution environment system (20).
13. A computer system designed to carry out the method (50) according to any one of claims 1 to 12.
14. A computer program designed to carry out the method (50) according to any one of claims 1 to 12.
15. A computer readable medium or signal storing and / or containing a computer program according to claim 14.
