Data exchange system and method based on multi-party secure computation

Abstract

The application relates to the technical field of privacy calculation and information security, and discloses a data exchange system and method based on multi-party secure calculation, which comprises the following: a first institution node connected with a second institution node; a circuit topology template generation module sends a static topology template and a random code truth table containing a global circuit instance serial number, the random code truth table is received and maintained by a ring buffer module, and a physical slot pointer is maintained; an authorization token packaging module generates a single authorization token, a handshake connection request is packaged and sent by a non-intention transmission request module; a sliding window response module compares a request serial number with a consumption expectation pointer of a hash chain state machine module, and a post-input wire label set is used; a secret state fusion verification module takes the random code truth table according to the pointer, combines a template, performs secret logic calculation, and outputs a result. The application adopts the technical scheme of maintaining a physical slot pointer by a ring buffer and executing an asynchronous prefetch operation in combination with a hash chain state machine, and achieves the technical effect of high efficiency of random code truth table throughput in multi-party calculation.

Description

Technical Field

[0001] This invention relates to the fields of privacy computing and information security technology, specifically to a data exchange system and method based on multi-party secure computation. Background Technology

[0002] Cross-institutional data collaboration is common in financial risk control, medical data sharing, and government cooperation. While sharing data value, institutions must strictly conceal the underlying plaintext information. Multi-party secure computation technology has thus become a key means of breaking down data silos. It allows untrusted nodes to complete collaborative computation without exposing their original privacy. Modern business scenarios have extremely stringent compliance requirements for data flow; the underlying secure data exchange system needs to possess continuous operational stability and a very high level of protection to cope with complex interactive verification needs.

[0003] Existing encrypted data exchange systems typically employ standard scrambled circuits and unintended transmission protocols. These schemes provide rigorous mathematical cryptographic proofs at the theoretical level, guaranteeing the absolute confidentiality of the original data during transmission. Developers can quickly establish communication handshakes between nodes using static global session identifiers, resulting in a simple and direct protocol interaction process. In terms of code implementation, the system relies on plaintext conditional statements to intercept unauthorized access; this straightforward architecture significantly lowers the initial development barrier. Simultaneously, the system schedules underlying cryptographic functions for real-time encrypted computation as needed, ensuring that each interaction is completed within a defined, single-step security framework.

[0004] Existing on-demand real-time computing mechanisms tightly bind the generation of garbled truth tables to network requests. Under high concurrency, this consumes extreme computing power, and the serial superposition of encrypted computation and network transmission causes a sharp increase in interaction latency. Switching to a strategy of preloading the truth table in full would quickly fill physical storage with massive amounts of state data, making nodes highly susceptible to memory overflow and crashing. Conventional conditional execution modes expose significant physical side-channel vulnerabilities in encrypted environments. The number of underlying hardware instructions corresponding to the authorization and rejection / blocking code paths differs greatly. The response times of computing nodes when processing different branches are completely different. External attackers only need to collect interface latency parameters to bypass digital signatures and reverse-engineer the true privacy judgment state. Real public network communication inevitably involves packet loss and network jitter. Existing systems rely solely on static identifiers to confirm connection identity. Once a communication timeout triggers client retransmission, the receiving end cannot distinguish between legitimate retransmitted data packets and maliciously intercepted replay attacks. Forced blocking would cause the communication state machines at both ends to derail and freeze, while direct access would expose a significant vulnerability for data theft. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a data exchange system and method based on multi-party secure computation. This addresses the technical problems in existing multi-party computation-based data exchange schemes, such as excessively high communication latency due to the on-demand real-time generation of garbled truth tables or memory overflow caused by full preloading, the vulnerability of conventional conditional branches during encrypted verification to side-channel bypass attacks that leak privacy states, and the inability of static session identifiers to securely distinguish between normal network packet loss and retransmission and malicious replay attacks.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a data exchange system based on multi-party secure computation, comprising a first institutional node and a second institutional node connected via a data communication network; The first institutional node includes a ring cache module, an authorization token encapsulation module, an unintentional transmission request module, and a cryptographic fusion verification module; The second mechanism node includes a circuit topology template generation module, a hash chain state machine module, and a sliding window response module; The circuit topology template generation module is used to generate a static topology template and a scrambled truth table bound with a global circuit instance number, and sends the static topology template and the scrambled truth table to the first mechanism node. The ring cache module is used to receive and store the static topology template and the garbled truth table, and maintain the physical slot pointer used to locate the local cache address; The authorization token encapsulation module is used to obtain business rule parameters, the physical slot pointer and the global circuit instance number, and generate a single authorization token. The unintentional transmission request module is used to encapsulate the single authorization token and the global circuit instance number into a handshake connection request, and send the handshake connection request to the second organization node; The sliding window response module is used to receive the handshake connection request, compare the global circuit instance number in the handshake connection request with the consumption expectation pointer maintained by the hash chain state machine module, and send the input wire tag set to the first mechanism node after the comparison is successful. The dense-state fusion verification module is used to extract the garbled truth table from the ring cache module based on the physical slot pointer, perform dense-state logic calculation on the input wire label set in combination with the static topology template, and output the calculation result.

[0007] Preferably, the ring cache module is internally configured with a low water level threshold; The circular cache module counts the number of unconsumed garbled truth tables, and when the number of unconsumed garbled truth tables drops to the low watermark threshold, it sends an asynchronous prefetch request to the second mechanism node. The hash chain state machine module receives the asynchronous prefetch request, extracts the generation key for the new batch using a one-way hash function, generates a scrambled truth table for the new batch using the static topology template, and sends the scrambled truth table for the new batch to the ring cache module.

[0008] Preferably, the hash chain state machine module maintains the consumption expectation pointer and the generation advance pointer; The generation advance pointer is used to identify the largest circuit instance sequence number that has been generated and sent, and the consumption expectation pointer is used to identify the circuit instance sequence number that is expected to be consumed. When the hash chain state machine module receives the asynchronous prefetch request, it only performs a monotonically increasing operation on the generation advance pointer; after the sliding window response module finishes sending the input wire tag set, the hash chain state machine module performs a monotonically increasing operation on the consumption expectation pointer.

[0009] Preferably, the logic for the authorization token encapsulation module to generate the single authorization token includes: Extract the user query identifier from the business rule parameters, and obtain the system timestamp and auto-incrementing serial number; The user query identifier, the system timestamp, the auto-incrementing serial number, the physical slot pointer, and the global circuit instance number are concatenated into binary data to generate a data payload; The local private key of the first institution node is invoked, and a digital signature algorithm is used to perform a signature operation on the data payload to generate a signature result. The data payload and the signature result are then encapsulated together into the single authorization token.

[0010] Preferably, the comparison logic of the sliding window response module includes: When the global circuit instance number in the handshake connection request is equal to the consumption expectation pointer, the comparison is deemed successful. When the global circuit instance sequence number in the handshake connection request is less than the consumption expectation pointer, determine whether the global circuit instance sequence number in the handshake connection request falls within the range of the history tag cache window; If the global circuit instance number in the handshake connection request falls within the range of the historical tag cache window and the auto-incrementing serial number in the handshake connection request matches the historical auto-incrementing serial number within the range of the historical tag cache window, the handshake connection request is determined to be a retransmission request, and the historical input wire tag set is extracted from the range of the historical tag cache window and sent to the first mechanism node. The physical slot pointer is not involved in the comparison of the sliding window response module.

[0011] Preferably, after the comparison is successful, the sliding window response module uses the public key of the first agency node to perform plaintext digital signature verification on the single authorization token; After the plaintext digital signature is verified, the sliding window response module retrieves privacy data parameters based on the user query identifier; The sliding window response module generates an authorization status wire tag indicating that the signature verification has passed. Combined with the privacy data wire tag corresponding to the privacy data parameter and the virtual decoy tag, the input wire tag set is assembled.

[0012] Preferably, when performing dense-state logic calculations, the dense-state fusion verification module loads a pre-compiled unconditional jump execution path; The unconditional jump execution path includes an authorization status verification branch, a real business calculation branch, and a homogeneous decoy branch; The dense-state fusion verification module traverses the logic gates in the static topology template in a fixed order, so that the real business calculation branch and the isomorphic decoy branch have the same number of logic gate traversals during the calculation process.

[0013] Preferably, the static topology template is equipped with a multiplexer component at the topology aggregation end; The multiplexer component receives the Boolean variable output by the authorization status verification branch as a control signal; When the control signal indicates that the authorization is valid, the multiplexer component extracts the calculation result of the actual service calculation branch and exports it as an encrypted string; When the control signal indicates that the authorization is invalid, the multiplexer component extracts the calculation result of the isomorphic decoy branch and exports it as the ciphertext string; The first mechanism node uses the output decoding mapping table bound to the global circuit instance number to match and decode the ciphertext string to obtain the service determination status.

[0014] Preferably, after completing the dense state logic calculation, the dense state fusion verification module sends a state update instruction to the ring cache module; The first mechanism node performs arithmetic modulo operation oriented towards storage capacity boundary to achieve circular addressing on the physical slot pointer according to the state update instruction, and performs an auto-increment operation on the global circuit instance number; The first mechanism node sends a memory overwrite instruction to the operating system to set the physical memory slot position pointed to by the physical slot pointer before the arithmetic modulo operation to zero.

[0015] This invention provides a data exchange system and method based on multi-party secure computation. It has the following beneficial effects: 1. This invention employs a circular cache to maintain physical slot pointers, combined with a hash chain state machine to perform asynchronous prefetching operations, achieving high throughput of the truth table in multi-party computation. Compared to existing technologies that generate or preload the truth table on demand, this invention addresses the shortcomings of high network communication latency and memory overflow issues faced by computing nodes.

[0016] 2. This invention employs an unconditional jump execution path technique in the encrypted logic verification stage, which includes real business computation and isomorphic decoy branches. By limiting the number of logic gate traversals, it achieves the technical effect of masking the real authentication result. Compared to the existing technology that selectively triggers a single code execution branch based on authorization conditions, this invention solves the problem that the differences in computation time and power consumption make it susceptible to side-channel attacks, thus exposing the privacy state.

[0017] 3. This invention employs a sliding window mechanism to compare the global circuit instance sequence number with the consumption expectation pointer, and utilizes an auto-incrementing serial number to match historical caches. This achieves the technical effect of securely identifying retransmission requests and maintaining node state synchronization during unintentional transmission. Compared to existing technologies that rely on static session identifiers for connection confirmation, this invention addresses the shortcomings of state synchronization issues caused by packet loss and the system's susceptibility to malicious replay requests. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the system architecture according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the method flow according to an embodiment of the present invention; Figure 3 This is a flowchart of the offline preset and asynchronous construction phases in the data exchange method of this invention. Figure 4 This is a flowchart of the in-band authorization token generation stage in the data exchange method of this invention embodiment; Figure 5 This is a flowchart of the state tear prevention verification and transmission establishment stage in the data exchange method of this invention embodiment; Figure 6 This is a flowchart of the dense-state fusion calculation stage with a constant execution cycle in the data exchange method of this invention. Figure 7 This is a flowchart of the state decoding and asynchronous prefetch feedback stage in the data exchange method of this invention embodiment; Figure 8 The diagram shows the timing data and pointer state changes of the system under concurrent operation in this embodiment of the invention. (a) is the running trajectory diagram of the global circuit instance number and physical slot pointer, and (b) is the available capacity of the ring cache module and the prefetch trigger timing diagram.

[0019] Among them, 10 is the first institutional node; 11 is the ring cache module; 12 is the authorization token encapsulation module; 13 is the unintentional transmission request module; 14 is the encrypted state fusion verification module; 20 is the second institutional node; 21 is the circuit topology template generation module; 22 is the hash chain state machine module; and 23 is the sliding window response module. Detailed Implementation

[0020] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Reference Figure 1 This invention provides a data exchange system based on multi-party secure computation. The system includes a first institutional node 10 and a second institutional node 20, which are connected via a data communication network. The first institutional node 10 can be the party requesting data exchange or joint computation, and the second institutional node 20 can be the party providing privacy data or computational tags.

[0022] The first institutional node 10 includes a ring cache module 11, an authorization token encapsulation module 12, an unintentional transmission request module 13, and a cryptographic fusion verification module 14. The second institutional node 20 includes a circuit topology template generation module 21, a hash chain state machine module 22, and a sliding window response module 23.

[0023] The circuit topology template generation module 21 generates a static topology template and generates a truth table sequence corresponding to the static topology template based on a hash chain. Each truth table is bound to a monotonically increasing global circuit instance number and together with the corresponding output decoding mapping table, constitutes a circuit instance data unit. The circuit topology template generation module 21 sends the static topology template and truth table sequence to the first mechanism node 10.

[0024] The ring cache module 11 receives and stores the static topology template and truth table sequence. The ring cache module 11 also maintains two types of state parameters: one is a physical slot pointer used to locate the local cache address, and the other is a global circuit instance number used to identify the currently consumed circuit instance. The physical slot pointer cycles through the physical address range of the ring cache module 11, and the global circuit instance number monotonically increases according to the consumption order of the circuit instances.

[0025] The circular cache module 11 sets low-water thresholds and high-water thresholds. When the number of unconsumed truth tables falls below the low-water threshold, the circular cache module 11 sends an asynchronous prefetch request to the second mechanism node 20. The hash chain state machine module 22 maintains a generation advance pointer and a consumption expectation pointer. The generation advance pointer identifies the largest circuit instance sequence number that has been generated and issued, and the consumption expectation pointer identifies the circuit instance sequence number expected to be consumed during the online transmission phase. Asynchronous prefetching only advances the generation advance pointer and does not change the consumption expectation pointer. In response to the asynchronous prefetch request, the hash chain state machine module 22 generates a new batch of truth table sequences and sends them to the circular cache module 11.

[0026] The authorization token encapsulation module 12 receives business operation instructions, extracts the business rule parameters from the local environment of the first organization node 10, and obtains the current physical slot pointer, global circuit instance sequence number, and runtime timestamp of the ring cache module 11 to generate a single authorization token. This single authorization token binds the local cache call location, the current circuit instance status, and the business request time status to the same authorization payload, which is used to restrict the reuse of historical requests across rounds.

[0027] The unintentional transmission request module 13, combining the single-authorization token and the global circuit instance sequence number, initiates a query handshake connection to the second agency node 20. Upon receiving this query handshake connection, the sliding window response module 23 extracts the global circuit instance sequence number and compares it with the consumption expectation pointer stored in the hash chain state machine module 22. The physical slot pointer is only used for local location of the cache address on the first agency node 10 and does not participate in the hash chain consumption state judgment on the second agency node 20 side. When the comparison result meets the state synchronization condition or retransmission matching condition, the sliding window response module 23 establishes an unintentional transmission channel and sends the set of input wire tags bound to the current circuit instance to the first agency node 10.

[0028] After obtaining the input wire tag set, the encrypted fusion verification module 14 calls the static topology template and the current truth table in the ring buffer module 11, which is located by the physical slot pointer and bound to the current global circuit instance number, to perform Boolean logic evaluation on the input wire tag set. The encrypted fusion verification module 14 outputs the ciphertext string corresponding to the service judgment result, and after the logic evaluation is completed, sends pointer stepping instructions and memory release instructions to the ring buffer module 11.

[0029] Reference Figure 2 This invention provides a data exchange method based on multi-party secure computation, comprising the following steps: S100, the second agency node 20 generates a static topology template and an initial truth table sequence, and sends them to the first agency node 10 for caching, while initializing the relevant state pointers of the first agency node 10 and the second agency node 20. S200, the first institutional node 10 obtains the business rule parameters and generates a single authorization token by combining the current physical slot pointer, global circuit instance number and system time status; S300, the first mechanism node 10 sends a transmission establishment request to the second mechanism node 20. The second mechanism node 20 performs state tear prevention verification based on the global circuit instance sequence number, and returns the set of input wire tags bound to the current circuit instance after the verification is passed. S400, the first mechanism node 10 performs parallel logic gate dense state evaluation based on the input wire label set, static topology template and current truth table to generate encrypted data stream; S500, the first institutional node 10 obtains the business output status according to the encrypted data stream, completes pointer advancement and clearing of consumed data, and triggers asynchronous data prefetching according to the number of unconsumed truth tables.

[0030] The execution process of each of the above steps will be explained in detail below, combining specific calculation logic and data processing mechanisms.

[0031] Reference Figure 3 The S100 stage mainly completes the distribution of static topology templates, hash chain key derivation, truth table preloading, and the establishment of subsequent asynchronous supplementation mechanisms, specifically including the following steps.

[0032] S110, the circuit topology template generation module 21 of the second institutional node 20 constructs a joint logic circuit for authorization verification and service comparison, and extracts the static connection relationship structure of the joint logic circuit as a topology template.

[0033] In this embodiment, the circuit topology template generation module 21 transforms the preset service comparison rules and authorized data verification logic into a network topology diagram composed of underlying logic gates. This network topology diagram retains the logic gate types and wire connection relationships, but does not include input data and key states. The resulting network topology diagram is the static topology template.

[0034] By separating dynamic data from static operators, the first mechanism node 10 does not need to repeatedly receive the complete circuit structure during subsequent calculations, thereby reducing the amount of data transmission during the online phase. The process of converting business rules into a logic gate network topology diagram can be completed by a secure multi-party computation circuit compiler, which is a conventional implementation method in this field and will not be elaborated here.

[0035] The circuit topology template generation module 21 sends the generated topology template to the first mechanism node 10. The ring cache module 11 of the first mechanism node 10 receives and statically saves the topology template. This topology template serves as the fixed execution path for the first mechanism node 10 to perform dense state calculations subsequently, and can remain unchanged during the runtime when the business rules do not change.

[0036] S120, the circuit topology template generation module 21 of the second mechanism node 20 derives an independent generation key sequence for each circuit instance based on a cryptographically secure hash function.

[0037] The same topology template needs to be matched with different encryption parameters in different computation rounds to avoid security risks caused by truth table reuse. The circuit topology template generation module 21 calculates and generates keys for each continuously generated dynamic circuit instance through a hash chain algorithm. The hash chain utilizes the irreversibility of one-way hash functions to bind each circuit instance in chronological order, thereby maintaining the consistency of the key states of both parties with low communication overhead.

[0038] For the A circuit instance to be executed, which generates a key. The hash recurrence relation is: ; in, A cryptographically secure hash function (e.g., the SHA-256 algorithm); Indicates the first The generation key for each circuit instance in the current round; This represents the generation key of the preceding node in the hash chain operation chain. hour, The initial master seed key preset by the system; This represents a memory concatenation operation for binary data; This indicates that the second institutional node 20 is the... The random salt value generated by the round circuit calculation is a preferred method. This random salt value can be generated by extracting physical hardware noise using a true random number generator. Indicates the first The physical identifier of each circuit instance is incremented.

[0039] Using the recursive method described above, the circuit topology template generation module 21 can continuously derive unpredictable and non-repeating key sequences starting from a single master seed. The key length can be determined based on the output bit width of the selected hash function.

[0040] In this embodiment, the hash chain state machine module 22 of the second mechanism node 20 simultaneously maintains the generation advance pointer. and consumption expectation pointer . Used to record the largest circuit instance sequence number that the second agency node 20 has generated and issued based on the hash chain; Used to record the circuit instance number that the second mechanism node 20 expects to be consumed by the first mechanism node 10 during the online inadvertent transmission phase.

[0041] Asynchronous prefetching process only makes Move backward without changing Only during the online transmission phase can the connection with the current... After interacting with the corresponding input wire label, Only then does the auto-incrementing process begin. By separating the generation and consumption states, the background prefetching does not prematurely change the online consumption judgment benchmark, thus preventing online requests from being misjudged as having incorrect states.

[0042] S130, the first institutional node 10 initializes the ring cache module 11 in local memory and sets the water level threshold parameter to manage the streaming loading of the truth table sequence.

[0043] The first institutional node 10 allocates a size of [size missing] in its local secure memory space. A ring cache module 11 is constructed using consecutive circular storage addresses. The first mechanism node 10 presets a low-water threshold within the ring cache module 11. High water level threshold Both satisfy .

[0044] In the project deployment, It can be estimated based on the average round-trip delay of network communication and the circuit consumption rate per unit time of nodes, so as to reduce the risk of buffer exhaustion before the prefetched data arrives; The amount of memory reserved for this business module by the first institutional node 10 can be determined.

[0045] During initialization, the first mechanism node 10 will be used to locate the physical slot pointer of the ring cache physical address. The value is assigned to 1, and the global circuit instance number used to identify the current consuming circuit instance is also assigned. The value is assigned to 1. Where, In 1 to Cyclic changes between them The number of calculations based on business operations increases monotonically. The second organizational node 20 will generate a progress indicator. Initialize to the largest circuit instance number in the initial truth table sequence, and consume the expected pointer. Initialize to 1.

[0046] The circuit topology template generation module 21 of the second mechanism node 20 uses the generation key calculated in step S120. The circuit topology template obtained in step S110 is used to generate a corresponding scrambled truth table. The scrambled truth table is formed by symmetrically encrypting the truth output results of each logic gate in the topology template using the generated key. The circuit topology template generation module 21 assembles an initial truth table sequence containing multiple sets of scrambled truth tables according to the hash chain derivation order.

[0047] The initial truth table sequence forms a sequence matrix. ,in Indicates the first The garbled truth table data bound to each circuit instance. The second mechanism node 20 generates each At the same time, a serial number corresponding to the circuit instance is generated. Bound output decoding map , It only includes the mapping relationship between the real business output label and the business enumeration value when the authorization verification is passed, and does not include the decoy branch output label. , The subsequent sets of input wire labels are all named after the same circuit instance number. Bind.

[0048] The circuit topology template generation module 21 sends the initial truth table sequence to the first mechanism node 10. The circular buffer module 11 of the first mechanism node 10 receives the sequence matrix and writes it into its internal storage space. During subsequent operation, the first mechanism node 10 uses physical slot pointers... Locate the storage slot in the ring cache module 11, and use the global circuit instance number. Identify the currently consumed garbled truth table instance, and then read the corresponding garbled truth table to participate in business calculations.

[0049] The ring cache module 11 is configured with capacity monitoring logic to count the number of unread and unconsumed garbled truth tables. When the number of unconsumed garbled truth tables drops to a low-water mark threshold... At that time, the ring cache module 11 generates a capacity warning instruction and triggers an asynchronous data prefetching action.

[0050] The ring cache module 11 sends a background prefetch request to the hash chain state machine module 22 of the second mechanism node 20. This request carries the currently consumed maximum global circuit instance sequence number. and the current physical slot pointer The auxiliary verification information is used to reduce the risk of cursor displacement due to network out-of-order delivery. After receiving the request, the hash chain state machine module 22 continues to calculate the generation key sequence for subsequent batches according to the same hash recursion formula based on the current local hash chain execution state, and generates a new batch of scrambled truth table data in combination with the topology template.

[0051] The hash chain state machine module 22 sends the new batch of garbled truth table data back to the ring cache module 11 of the first institutional node 10, whereby the ring cache module 11 writes it into an available storage slot. In this way, static topology parsing and dynamic truth table generation are separated, and the truth table supplementation process can be performed in the background without occupying the main business communication link.

[0052] To ensure consistency between the offline pre-set garbled truth table, the online transmission input wire labels, and the final output decoding mapping table, each circuit instance is assigned a monotonically increasing global circuit instance number. For any circuit instance The garbled truth table generated by the second node 20 Output decoding mapping table Input wire label set and hash chain generation key All with the same Binding.

[0053] Before performing the secure state assessment, the first institutional node 10 reads locally... Received as well as of Perform a consistency check; if the three If there is a discrepancy, the current round of calculation is terminated and a status anomaly alarm is output. This instance binding mechanism can reduce circuit instance mismatch caused by ring buffer address wraparound, network retransmission, or asynchronous prefetch out-of-order issues.

[0054] Reference Figure 4 The S200 stage is used to bind the user query identifier, business rule parameters, cache instance status, and time status into a single authorization token after an online business is triggered. Specifically, it includes the following steps.

[0055] S210, the authorization token encapsulation module 12 of the first institutional node 10 captures the business operation instruction issued by the terminal device, and extracts the corresponding plaintext business rule parameters locally based on the business operation instruction.

[0056] Online computation is triggered by terminal business interaction. After receiving the business request message, the authorization token encapsulation module 12 parses the business context information and extracts the plaintext business rule parameters required for this data exchange from the rule configuration library or business engine of the first institutional node 10 according to the business context. These plaintext business rule parameters are denoted as... This is used to compare the data with the data provider's data in subsequent dense state calculations.

[0057] The data structure can be adjusted according to the business scenario. Taking joint assessment as an example, A pre-configured numerical comparison threshold can be set, with its data bit width matching the input pin width of the static topology template, so that subsequent Boolean operations can be performed normally.

[0058] In a preferred embodiment, to reduce the direct exposure of the original user unified identifier during cross-organizational data exchange, the first organization node 10 and the second organization node 20 pre-agree on user identifier mapping rules. Before initiating a request, the first organization node 10 stores the original user unified identifier... Convert to user query identifier The second institutional node 20 completes the process based on the same mapping rules or a pre-synchronized mapping table. Matching with local user records. It is used for querying and locating information, but does not directly reveal the user's original identity information.

[0059] S220, the authorization token encapsulation module 12 reads the current physical slot pointer from the ring cache module 11. and the current global circuit instance number The upper-layer business state parameters are concatenated with the above pointer states in the underlying memory to achieve atomic binding of multi-dimensional states.

[0060] After completing the extraction of plaintext business rule parameters, the authorization token encapsulation module 12 sends a status query command to the ring cache module 11 to obtain the physical slot pointer where the current pending garbled code truth table is located. And the global circuit instance number bound to the truth table of this garbled text. Simultaneously, the authorization token encapsulation module 12 extracts the unified user identifier involved in this interaction. It generates user query identifiers according to pre-agreed blinding, hashing, or pseudo-anonymity mapping rules. As an optional approach, Depend on The salt value agreed upon by both parties is obtained through a hash operation.

[0061] The authorization token encapsulation module 12 also calls the system's underlying clock to obtain the system timestamp. It also generates an auto-incrementing serial number to identify the current communication session. . Used to limit the validity period of authorization tokens Used to distinguish different business requests initiated by the same user within a similar time period.

[0062] In concurrent service request scenarios, the first institutional node 10 uses mutex locks, atomic counters, or transactional queues to monitor physical slot pointers. and global circuit instance number Distribute the same Only one business request can occupy the table at any given time, thus avoiding multiple business requests from repeatedly consuming the same garbled truth table instance.

[0063] The authorization token encapsulation module 12 encapsulates the user query identifier according to a preset serialization specification. System timestamp Self-incrementing serial number Global circuit example number and physical slot pointer The data is concatenated into a binary data payload. This serialization specification can use fixed-length field alignment or add fixed delimiters between adjacent fields to avoid parsing ambiguity caused by directly concatenating fields of different lengths.

[0064] Using the above concatenation method, the user query status, time-based replay prevention status, session uniqueness status, and circuit cache status are all written into the same data payload. This data payload is then used in the signature operation, thereby preventing attackers from using historically valid fields to recombine and generate new, valid requests.

[0065] S230, the authorization token encapsulation module 12 uses the local private key to perform a signature operation on the concatenated data payload to generate a single authorization token, and uses the single authorization token as the prerequisite input for subsequent cryptographic verification.

[0066] To ensure the bound data payload is tamper-resistant, the authorization token encapsulation module 12 calls the local private key deployed in the secure execution environment of the first institution node 10 to perform a digital signature operation on the binary data payload, generating a single-use authorization token. During system initialization, the public key corresponding to the first institution node 10 has been synchronized to the second institution node 20, or configured by a trusted third-party certificate authority for subsequent signature verification.

[0067] The data structure of this single-authorization token can be represented as follows: ; ; ; in, The plaintext field payload representing a single authorization token. This indicates that the first institutional node 10 is based on a local private key pair. The generated digital signature value; It is not merely an unresolvable signature result, but rather a combination of field payload and signature value, so that the second agency node 20 can read it. , and And perform signature verification and status comparison. Indicates by and A single authorization token composed of all of these; This represents a digital signature function based on the local private key of the first institution node 10; This represents the user query identifier formed after blinding, hashing, or pseudo-anonymization mapping; Represents the system timestamp; Indicates an auto-incrementing serial number; Indicates the current global circuit instance number being consumed; This indicates the physical slot pointer currently being used by the ring cache module 11; This indicates a memory concatenation operation between data.

[0068] A combined structure of "field payload + signature value" is used, instead of a single unresolvable signature result. The second agency node 20 receives... Afterwards, able to from Read from , , and and combined Complete signature verification and status comparison. The digital signature function can be implemented using the RSA algorithm, SM2 algorithm, or elliptic curve digital signature algorithm.

[0069] Once generated, the single-authorization token serves as both an identity credential in network requests and a prerequisite input for establishing unintended transmission channels and evaluating logic gates in encrypted circuits. If If the field data is inconsistent with the actual running status, subsequent signature verification, status comparison, or encrypted binding verification will fail.

[0070] Reference Figure 5 The S300 stage is used to verify the circuit instance status, retransmission request status, and authorization token validity before the input wire tag is sent, and specifically includes the following steps.

[0071] S310, the unintentional transmission request module 13 of the first institutional node 10 will generate the single-use authorization token and user query identifier. System timestamp Self-incrementing serial number Global circuit example number and physical slot pointer Encapsulate it into a plaintext metadata set, and use it to initiate a handshake connection request for parameter query to the second organization node 20.

[0072] After completing local parameter processing, the first agency node 10 enters the network interaction phase. Handshake messages can be encrypted and protected at the transport layer using a secure transmission protocol; at the application layer, the plaintext metadata set is used by the second agency node 20 for parsing and performing status verification. The plaintext metadata set sent by the inadvertent transmission request module 13 to the second agency node 20 can be represented as follows: .

[0073] Among them, the global circuit instance number The core beacon used by the second institutional node 20 for state synchronization judgment; physical slot pointer It is only used to assist in identifying the local cache location of the first institutional node 10, and does not participate in the hash chain consumption state advancement on the side of the second institutional node 20.

[0074] S320, after receiving the handshake request, the sliding window response module 23 of the second mechanism node 20 extracts the global circuit instance number from the request message. Self-incrementing serial number and timestamp And initiate the anti-tear verification logic.

[0075] In a distributed network environment, communication interruptions, timeout retransmissions, or link delays can lead to inconsistencies in the hash chain progression states of both parties. The second institutional node 20 maintains a consumption expectation pointer for this purpose. and size The history tab cache window. The maximum number of retransmissions can be determined based on the maximum number of retransmissions in the underlying network protocol stack and the average system timeout period, in order to cover duplicate requests caused by normal network jitter.

[0076] Sliding window response module 23 will and Perform a comparison and process according to the following logic: when At this point, both state machines are in a synchronized state, allowing the continuation of the new tag extraction and transmission process.

[0077] when At that time, the request may originate from a network retransmission. Because... This only represents the local cache slot of the first institutional node 10, which may repeat as the circular cache wraps around. It does not participate in the state synchronization judgment. At this time, the sliding window response module 23 makes further judgments. Check whether it falls within the historical sliding window range and verify the information in the request message. Does it match the serial number in the local history? This judgment can be expressed as: ; in, This indicates the retry decision result as a Boolean value. Represents the membership relation of a set; This represents the currently valid historical sliding window range; This represents the logical AND operation; The auto-incrementing serial number carried in the current request; For system historical cache records and The corresponding historical auto-incrementing serial number; This represents the numerical equality matching operation.

[0078] when When true, the sliding window response module 23 determines that the request is a valid idempotent retry. At this time, there is no need to consume a new garbled truth table instance. The sliding window response module 23 directly extracts the input wire tag set of the corresponding session from the historical tag cache window and resends it to the first mechanism node 10, and then ends the current processing flow.

[0079] when Not equal to And when the conditions for a valid retry are not met, for example Exceeding the current consumption expectation pointer ,or Falling into the window of history but If the comparison fails, the sliding window response module 23 determines that there is a risk of state machine disorder or replay tampering, and actively disconnects the current communication connection, while outputting an out-of-bounds alarm message to the system log center.

[0080] S330, under the condition of state machine synchronization, the second mechanism node 20 is based on the user query identifier. Retrieve local privacy data parameters, establish a communication channel with the first agency node 10, and issue the corresponding wire tag.

[0081] After the tear-proof verification, the sliding window response module 23 uses the pre-set public key of the first institution node 10 to process the single-authorization token. In and Perform plaintext signature verification and validation. Whether it is within the preset valid time range. Meanwhile, the sliding window response module 23 will... The parsed , , , and The handshake request is compared with the corresponding field in the handshake request; if any field is inconsistent, the current handshake request is deemed invalid.

[0082] After the signature verification is successful and the fields match, the sliding window response module 23 responds according to... Retrieving local privacy data parameters If the signature verification fails, the timestamp expires, or the field comparison fails, or Unable to match a valid user record, the sliding window response module 23 refuses to establish an unintended transmission channel. In the data provider's business scenario... It can assess user scores, risk levels, or characteristic asset status values.

[0083] After the retrieval is completed, the unintentional transmission request module 13 and the sliding window response module 23 establish an unintentional transmission channel. Using the unintentional transmission protocol, the first agency node 10 can obtain the underlying circuit tag corresponding to its service input, while the second agency node 20 cannot know the actual service input value of the first agency node 10. The wire tag is essentially a random key string representing logic 0 and logic 1 in the obfuscated circuit.

[0084] In this embodiment, data from different stakeholders uses different tagging interaction methods. Plaintext rule parameters. This is a private input belonging to the first institutional node 10. The first institutional node 10 obtains the corresponding wire tag set through an unintentional transmission channel, denoted as... .

[0085] For single-authorization tokens After the plaintext verification is successful, the second agency node 20 will send the verification result, the timestamp validity result, and... Binding relationships are mapped to authorized status wire tags And generate the current global circuit instance number. Corresponding pointer label Privacy data parameters For local data belonging to the second agency node 20, the second agency node 20 directly selects its corresponding wire label set. And send it to the first agency node 10.

[0086] In addition, the second agency node 20 generates decoy request tags with the same bit width as the actual business input based on independent random numbers. and decoy privacy data tags . , , , , as well as Together, they constitute the input wire label set for this round, and each is associated with the current circuit instance number. Binding.

[0087] Before sending the input wire tag set, the second mechanism node 20 first sends the corresponding data for this interaction. , The input wire tag set and session state are encapsulated into a history object and written to the history tag cache window in an atomic transaction manner; then the consumption expectation pointer is updated to... Finally, the set of input wire tags for this round is sent to the first institutional node 10.

[0088] Due to historical writing and The advancement occurs before tag transmission. Even if a network interruption occurs during tag transmission, the second agency node 20 can retrieve the same set of input wire tags from the historical tag cache window and retransmit them in subsequent retransmission requests. When the number of records in the historical tag cache window reaches... When the limit is reached, the second agency node 20 uses a first-in-first-out (FIFO) approach to remove the oldest historical session record in order to free up cache space.

[0089] The underlying encryption negotiation and data blinding process of the unintentional transmission channel can be implemented using the basic unintentional transmission protocol or an extended protocol based on a pseudo-random number generator. These are conventional implementation methods in this field and will not be elaborated further here.

[0090] Reference Figure 6 In the S400 stage, after the input wire tag is acquired, the first mechanism node 10 evaluates the confusion circuit locally and reduces the risk of side channel leakage caused by the difference in authorized state by using an equal-length execution path. Specifically, it includes the following steps.

[0091] S410, the dense state fusion verification module 14 of the first institutional node 10 completes the initialization of the dense state evaluation engine, and sets the static topology template and physical slot pointers. Locate and match the global circuit instance number The bound garbled truth table and the received wire tags are loaded into system memory.

[0092] After the data exchange is completed unintentionally via the transmission channel, the encrypted fusion verification module 14 calls its internal encrypted evaluation engine as the underlying Boolean operation unit. The encrypted fusion verification module 14 sends a data extraction command to the ring buffer module 11, based on the current physical slot pointer. Extract the garbled truth table from the corresponding cache slot, and then determine the current global circuit instance number. Verify the instance binding relationship of the garbled character truth table. If the instance sequence number of the garbled character truth table matches... If consistent, record the garbled truth table as the current garbled truth table. .

[0093] Meanwhile, the dense-state fusion verification module 14 loads a pre-set static topology template from the local persistent storage medium. It also reads the set of wire tags received in the previous stage into the register.

[0094] In this embodiment, the set of wire tags to be loaded includes tags representing authorization status, plaintext rule parameters, and privacy data parameters. It also includes pointer tags corresponding to global circuit instance numbers. and virtual request tags With virtual privacy data tags .

[0095] in, Generated by the second agency node 20 after the plaintext signature is verified, it is used to represent the comprehensive verification status of the binding relationship between the authorization token, timestamp, and instance sequence number in the encrypted circuit; Used to verify whether the current set of input labels corresponds to the same circuit instance as the current garbled truth table. and The bit width is consistent with the actual business input label, which is used to support the calculation of subsequent isomorphic decoy branches.

[0096] To reduce the impact of operating system scheduling on execution time, the dense-state fusion verification module 14 locks the aforementioned core data in contiguous physical memory pages through the operating system kernel interface. This memory-resident approach reduces the additional time differences caused by page faults or cache paging, providing the necessary runtime conditions for subsequent equal-length execution.

[0097] In this embodiment, authorization verification is divided into a plaintext signature verification layer and a cryptographic binding layer. The plaintext signature verification layer is completed by the second agency node 20 before the inadvertent transmission channel is established, and is used for verification. The signature value in timestamp and field payload Whether it is valid; the encrypted binding layer is completed by the encrypted fusion verification module 14 within the obfuscation circuit, and is used to verify the authorized state wire tag. With the current global circuit instance number The binding relationship between them. This layered processing method avoids fully implementing complex digital signature algorithms in Boolean circuits, while preserving the verification relationship between the authorization state and the current circuit instance.

[0098] S420, the dense state fusion verification module 14 is based on homogeneous redundant logic, fully traverses the authorized state verification branch, the real business calculation branch and the invalid random number decoy branch, and performs parallel Boolean evaluation.

[0099] In conventional software control flow, computation is typically interrupted and an exception is returned immediately upon authorization verification failure, resulting in a significant difference between this execution path and the successful authorization path. To reduce timing leaks caused by this difference, the static topology template Γ is pre-compiled into three unconditional execution paths at the underlying Boolean logic level. During the evaluation process, the dense-state fusion verification module 14 traverses all logic gates in the topology in a predetermined order, without performing conditional jumps based on input data.

[0100] The first path is the authorization verification branch. The encrypted fusion verification module 14 uses the authorization status label verification logic and instance sequence number matching logic to verify the labels. as well as Perform an evaluation. This branch outputs a encrypted Boolean variable representing the authorization verification result. The operational relationship can be expressed as: ; in, This represents the authorization status tag verification function, used to determine whether the authorization status tag generated by the second agency node 20 during the plaintext signature verification stage is valid; Represents the function for matching the ordinal numbers of dense-state instances; This indicates the wire label after the current global circuit instance number is mapped to the dense state space; This represents the Boolean logical AND operation.

[0101] By moving the full digital signature verification to the plaintext verification stage of the second agency node 20, and performing only the authorization status label and instance sequence number binding verification in the encrypted circuit, the size of the Boolean circuit can be reduced and the efficiency of online evaluation can be improved.

[0102] The second path is the real business branch. The encrypted fusion verification module 14, based on the business rule comparison logic, compares the tags representing the rule parameters of the first institutional node 10. Labels representing the privacy data of the second agency node 20 Perform a ciphertext comparison. This branch outputs a ciphertext boolean variable representing whether the business rule has been met. The operational relationship can be expressed as: ; in, This represents a dense comparison function that can perform numerical comparison or character matching.

[0103] The third path is the isomorphic decoy branch. The dense-state fusion verification module 14 performs a virtual comparison operation on the loaded meaningless noise labels, with the same logical depth as the real business branch, and outputs a ciphertext Boolean variable. The operational relationship can be expressed as: ; In this embodiment, the three computation branches are configured to have a predetermined, consistent gate traversal length. For branches with fewer gates than the target gate depth, the dense-state fusion verification module 14 inserts data-independent padding logic gates or no-operation logic gates during the circuit pre-compilation stage, ensuring that the authorized verification branch, the real business branch, and the homogeneous decoy branch have the same or substantially consistent number of logic gate traversals and memory access patterns under a preset execution path. This processing method can reduce the differences in CPU clock cycles and memory access behavior among different branches, thereby reducing the risk of timing-side channel leakage.

[0104] S430, the encrypted fusion verification module 14 calls the multiplexer logic unit at the end of the circuit to mask and integrate the results of each branch, and generate the final output encrypted string.

[0105] Once the data streams from all three branch paths reach the logical topology convergence point, the dense-state fusion verification module 14 uses multiplexing logic gates to route the actual business results and the decoy results. This multiplexing logic uses the authorization verification result as a control signal and completes the result selection through bitwise operations, avoiding the use of software-level conditional jumps.

[0106] The encrypted fusion verification module 14 calculates the final output ciphertext string. The control algebraic relation can be expressed as: ; Or it can be expressed as: ; in, This represents the final ciphertext string output from the circuit's output pin; This represents the Boolean logic XOR operation; Indicates the Boolean NOT operation; Represents the actual business branch result The output label obtained after valid business encoding; This indicates a random invalid output label that does not belong to a valid decoding map.

[0107] when When the authorization is valid, the multiplexer outputs the actual service branch result after valid service encoding. ;when When the corresponding authorization is invalid, the multiplexer outputs an invalid output domain result generated by an independent random seed. . Belongs to the output decoding map table Registered valid output fields Not belonging to Any valid enumerated value registered, therefore the first agency node 10 cannot... Decode the valid business status as pass, rejection, etc.

[0108] With the above structure, both valid and invalid authorization states require traversing the preset logic gate set, and the number of gate evaluations and memory access patterns remain consistent or nearly consistent. The specific obfuscated gate evaluation, dense state comparison, and multiplexing logic can be implemented using Yao's obfuscated circuit evaluation protocol or a half-gate optimized operation protocol, which are conventional implementation methods in this field and will not be elaborated further here.

[0109] It should be noted that for plaintext verification failures, expired timestamps, or... If a request cannot be matched with a valid user record, the second agency node 20 directly refuses to establish an unintentional transmission channel during the communication establishment phase. For cases where communication has been established but an authorized status tag expires, instance number mismatches, or the input wire tag set is replaced during the secure state evaluation phase, the secure state fusion verification module 14 outputs an invalid output domain result through a multiplexer. Therefore, illegal requests before communication is established are intercepted by the plaintext signature verification layer, while tag mismatches or abnormal encrypted states after communication is established are intercepted by the encrypted binding layer.

[0110] Reference Figure 7 After the dense state evaluation is completed, the output result is decoded in the S500 stage, and the background prefetch is triggered according to the remaining truth table state in the ring buffer module 11. Specifically, the following steps are included.

[0111] S510, the first agency node 10 obtains the output string after the encrypted state evaluation is completed, and uses the preset decoding mapping table to restore the encrypted string into the business judgment result.

[0112] After the encrypted fusion verification module 14 completes the aggregation of the results from the multiplexer, it submits the final exported encrypted output result to the upper-layer business environment. The first institutional node 10 has already obtained the plaintext decoding table corresponding to the business output status through the secure channel during the offline preset stage or the unintentional transmission stage.

[0113] This plaintext decoding table is related to the current global circuit instance number. Bound output decoding map The second node 20 generates the corresponding garbled truth table. Generate synchronously. Only the mapping relationship between the real business output tags and the actual business status under the successful authorization state is registered; invalid output tags generated by isomorphic decoy branches are not registered. Specifically, Only valid output fields are covered; invalid outputs resulting from authorization failures or tag mismatches will not be parsed as normal business results.

[0114] The first node 10 will output the final ciphertext result and... The valid entries in the ciphertext are matched. If the ciphertext output matches a preset enumeration value, the first institutional node 10 confirms the validity of the authorization status in this round and obtains the corresponding real business comparison result. In a joint risk control scenario, this business comparison result can be parsed as a risk assessment status, admission result, or marketing screening result.

[0115] If the match fails, that is, the ciphertext output result is not found. Any valid enumerated value in the current round of output is considered invalid by the first mechanism node 10. This situation can be caused by invalid authorization status labels, mismatched instance numbers, or inconsistencies between the input label set and the current garbled truth table. This is because the output labels of the isomorphic decoy branch are not written to... Exported when authorization fails It cannot be restored to a valid business state such as pass or reject. At this time, the business control engine of the first agency node 10 intercepts the current business process, records the exception log, and returns an unauthorized access or status exception prompt to the upper layer application.

[0116] S520, First Organization Node 10, in local memory, pointers to physical slots. Perform a circular step and check the global circuit instance number. Perform monotonically increasing operations while simultaneously clearing the consumed garbled truth table in the circular cache module 11.

[0117] After a single circuit instance completes decoding and evaluation, the first mechanism node 10 updates its local pointer state to maintain the hash chain consumption order and cache cyclic reuse. The update process includes updating the physical slot pointer and updating the global instance sequence number, and its algebraic expression logic is as follows: ; ; in, This indicates the physical slot pointer for the next round after the update assignment is performed. This indicates the global circuit instance number in the next round after the update assignment; ← indicates state progression and variable assignment operations; mod indicates arithmetic modulo operation for memory boundaries; This indicates the maximum truth table capacity allocated to the ring cache module during the initialization phase; 1 indicates a single step increment.

[0118] Perform a modulo operation to ensure that the local cache address is between 1 and... Reuse them in a loop; Instead of performing modulo operations, the hash chain increments monotonically with the number of circuit instance consumptions to maintain the uniqueness and continuity of the hash chain state verification. Therefore, cache slots can be reused cyclically, and circuit instance numbers will not be duplicated due to address wraparound.

[0119] While updating the pointer state, the first mechanism node 10 adjusts the pointer based on the previous physical slot pointer. and the corresponding pre-prepared global circuit instance number The consumed data in the circular cache module 11 is cleaned up. The cleanup targets include data related to... Bound garbled truth table Output decoding mapping table And local cached temporary tag reference information.

[0120] Specifically, the first institutional node 10 initiates a continuous memory overwrite instruction to the underlying operating system, changing the pointer of the previous physical slot. Pointing to, and related to, the instance number of the preceding global circuit. The bound garbled truth table occupies zero physical memory slots or is filled with random noise data. This process reduces the risk of a malicious process extracting the consumed garbled truth table from memory and frees up storage space for subsequent batches of data to be written. Memory overwriting and physical erasure can be accomplished by calling the standard memory cleanup interface provided by the operating system kernel, which is a common implementation method in this field.

[0121] S530, the first mechanism node 10 monitors the number of unconsumed truth tables in the ring cache module 11, and triggers an asynchronous data prefetch callback for the second mechanism node 20 when the number of unconsumed truth tables reaches the low watermark threshold.

[0122] In scenarios with continuous service requests, the pre-loaded obfuscated circuit computing resources will be continuously consumed. To avoid waiting for the second mechanism node 20 to generate the truth table in real time during online computation, the ring cache module 11 counts the number of truth tables that have not been read and consumed after each memory cleanup. The first mechanism node 10 compares this number with the low watermark threshold. A comparison is made to determine whether background prefetching needs to be triggered. This determination can be expressed as: ; in, This indicates the number of unconsumed garbled truth tables currently counted in the ring cache module 11; This represents the low water level threshold parameter.

[0123] In this embodiment, This can be determined based on the system's average circuit consumption rate per unit time and the network round-trip communication cycle. As an optional configuration, It should be no less than 1.5 times the product of the two mentioned above, in order to reserve buffer space for network jitter and response latency.

[0124] When the capacity condition is met, the ring cache module 11 wakes up the background network daemon thread and initiates an asynchronous prefetch request for the truth table of the new batch to the hash chain state machine module 22 of the second institutional node 20 through an independent bypass communication connection. To avoid repeatedly initiating prefetch requests in the low-water state, the first institutional node 10 sets a prefetch lock flag after triggering asynchronous pull; before the new batch of data is returned and the memory address is filled, this prefetch lock flag prevents the addition of new prefetch callbacks of the same type.

[0125] The second mechanism node 20 responds to the asynchronous prefetch request, based on the generation advance pointer maintained in the hash chain state machine. The scrambled truth table data for the subsequent sequence is generated and pushed to the circular cache module 11 of the first mechanism node 10. During this asynchronous prefetching process, the second mechanism node 20 only advances... Do not advance the consumption expectation pointer ; The process proceeds only after the input wire tag interaction for the current circuit instance is completed during the online unintentional transmission phase.

[0126] Through this asynchronous prefetch closed loop based on the low watermark, the generation of the garbled truth table and network transmission are transferred to the background for execution. The online main process mainly performs local memory reading and dense state evaluation of the logic gates corresponding to the garbled truth table, thereby reducing the waiting time for a single data exchange.

[0127] To further clarify the collaborative working process of the technical solution described in this invention, a specific working scenario example will be used for illustration below.

[0128] Reference Figure 8 The first institutional node (10) is designated as a financial institution, and the second institutional node (20) as a credit reporting agency. Financial institutions need to determine whether a user meets loan eligibility criteria; the local business rule parameter is a credit score threshold. The credit reporting agency records the user's private data parameters, namely their true credit score, in its local database. Both sides need to avoid direct exposure. and Under the premise of [the above conditions], the final access determination results will be output to financial institutions.

[0129] During the system initialization phase, the credit reporting agency's circuit topology template generation module 21 generates a static topology template for numerical comparison operations. The data is then sent to the financial institution. The financial institution configures a local ring cache module 11 and sets a maximum capacity limit. Low water level threshold The circuit topology template generation module 21 generates 100 batches of scrambled truth table sequences based on the initial master seed key and sends them to the financial institution's circular cache module 11. At this time, the local physical slot pointer... Global circuit instance number The expected consumption index maintained by the credit reporting agency. .

[0130] When a user initiates a loan application with a financial institution, it triggers an online calculation process. The financial institution's authorization token encapsulation module 12 extracts local business rule parameters. And obtain the pointer to the physical slots currently to be consumed. With global circuit instance number Combined with user query identifiers System timestamp and auto-incrementing serial number The authorization token encapsulation module 12 uses its local private key to sign the data payload after concatenating the above parameters, generating a single authorization token. Subsequently, the financial institution's unintentional transmission request module 13 encapsulates the plaintext parameter set into a handshake request and sends it to the credit reporting agency.

[0131] After receiving the request, the sliding window response module 23 of the credit reporting agency extracts the information from the message. and local The state machine is synchronized by comparison. The sliding window response module 23 verifies the synchronization via public key pair. Perform signature verification, checking the consistency of the timestamp and fields. After successful verification, the sliding window response module 23... Retrieve the user's real credit score .

[0132] During the tag interaction phase, both parties establish an unintentional transmission channel. The financial institution's unintentional transmission request module 13, based on private parameters... The corresponding input wire tag is obtained through blind selection of the transmission protocol without being noticed. Credit reporting agencies cannot know exactly which plaintext values ​​corresponding to which tags financial institutions extracted. Meanwhile, the credit reporting agency's sliding window response module 23 will use local data... Corresponding wire label Authorized status wire label pointer label The generated virtual decoy tags are sent directly to the financial institution. After the interaction is complete, the sliding window response module 23 updates the consumption expectation pointer to... .

[0133] The dense-state fusion verification module 14 of financial institutions extracts physical slots. Truth table of garbled characters Verify that its instance number matches the current one. Once consistent, load the topology template. The input wire tags are used to perform Boolean logic evaluation. During the calculation, the dense-state fusion verification module 14 does not rely on conditional jumps, but instead completely traverses all logic gates according to the static topology template. The calculation branch representing the real business comparison (i.e., judging 650>600) and the decoy branch using virtual decoy tags are executed isomorphically, producing consistent computation time and memory access trajectories. The multiplexer at the end of the circuit receives the output of each branch. Since the pre-authorization state verification is valid, the multiplexer extracts the result of the real business comparison branch as the final ciphertext string. Export and discard the results of the decoy branch. Financial institutions then use a pre-defined output decoding mapping table. ,Will Decode the status as a valid business state and determine that the user meets the admission criteria.

[0134] After a single circuit instance calculation is completed, the financial institution updates the pointer state locally and simultaneously cleans up memory. Combined with... Figure 8 The trajectory diagram of part (a) shows the global circuit instance number. It exhibits a monotonically increasing step-like upward trajectory, ensuring that each business request is bound to a unique instance; physical slot pointer When the capacity limit is reached The model is extracted in real time, resulting in a sawtooth-shaped cyclic reuse trajectory.

[0135] In concurrent request scenarios, the local garbled character truth table is continuously consumed. Combined with... Figure 8 As can be seen from the available capacity and prefetch triggering sequence diagram in part (b), the number of unconsumed truth tables in the ring cache module 11 continuously decreases as business progresses. When the number of unconsumed truth tables drops to the low watermark threshold... At that time, the ring cache module 11 generates an early warning command, triggering an asynchronous prefetch request to the credit reporting agency. The credit reporting agency's hash chain state machine module 22, in the background, generates a progress pointer... A new batch truth table is generated and sent back to populate it. In the graph, this is represented by the cache capacity curve rebounding after hitting the low watermark, ensuring that the main online computing process is not blocked.

[0136] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A data exchange system based on secure multi-party computation, characterized in that, This includes a first and a second agency node connected via a data communication network; The first institutional node includes a ring cache module, an authorization token encapsulation module, an unintentional transmission request module, and a cryptographic fusion verification module; The second mechanism node includes a circuit topology template generation module, a hash chain state machine module, and a sliding window response module; The circuit topology template generation module is used to generate a static topology template and a scrambled truth table bound with a global circuit instance number, and sends the static topology template and the scrambled truth table to the first mechanism node. The ring cache module is used to receive and store the static topology template and the garbled truth table, and maintain the physical slot pointer used to locate the local cache address; The authorization token encapsulation module is used to obtain business rule parameters, the physical slot pointer and the global circuit instance number, and generate a single authorization token. The unintentional transmission request module is used to encapsulate the single authorization token and the global circuit instance number into a handshake connection request, and send the handshake connection request to the second organization node; The sliding window response module is used to receive the handshake connection request, compare the global circuit instance number in the handshake connection request with the consumption expectation pointer maintained by the hash chain state machine module, and send the input wire tag set to the first mechanism node after the comparison is successful. The dense-state fusion verification module is used to extract the garbled truth table from the ring cache module based on the physical slot pointer, perform dense-state logic calculation on the input wire label set in combination with the static topology template, and output the calculation result.

2. The data exchange system based on multi-party secure computation according to claim 1, characterized in that, The ring cache module is internally configured with a low water level threshold. The circular cache module counts the number of unconsumed garbled truth tables, and when the number of unconsumed garbled truth tables drops to the low watermark threshold, it sends an asynchronous prefetch request to the second mechanism node. The hash chain state machine module receives the asynchronous prefetch request, extracts the generation key for the new batch using a one-way hash function, generates a scrambled truth table for the new batch using the static topology template, and sends the scrambled truth table for the new batch to the ring cache module.

3. The data exchange system based on multi-party secure computation according to claim 2, characterized in that, The hash chain state machine module maintains the consumption expectation pointer and the generation advance pointer; The generation advance pointer is used to identify the largest circuit instance sequence number that has been generated and sent, and the consumption expectation pointer is used to identify the circuit instance sequence number that is expected to be consumed. When the hash chain state machine module receives the asynchronous prefetch request, it only performs a monotonically increasing operation on the generation advance pointer; after the sliding window response module finishes sending the input wire tag set, the hash chain state machine module performs a monotonically increasing operation on the consumption expectation pointer.

4. The data exchange system based on multi-party secure computation according to claim 1, characterized in that, The logic for the authorization token encapsulation module to generate the single authorization token includes: Extract the user query identifier from the business rule parameters, and obtain the system timestamp and auto-incrementing serial number; The user query identifier, the system timestamp, the auto-incrementing serial number, the physical slot pointer, and the global circuit instance number are concatenated into binary data to generate a data payload; The local private key of the first institution node is invoked, and a digital signature algorithm is used to perform a signature operation on the data payload to generate a signature result. The data payload and the signature result are then encapsulated together into the single authorization token.

5. The data exchange system based on multi-party secure computation according to claim 4, characterized in that, The logic for comparison performed by the sliding window response module includes: When the global circuit instance number in the handshake connection request is equal to the consumption expectation pointer, the comparison is deemed successful. When the global circuit instance sequence number in the handshake connection request is less than the consumption expectation pointer, determine whether the global circuit instance sequence number in the handshake connection request falls within the range of the history tag cache window; If the global circuit instance number in the handshake connection request falls within the range of the historical tag cache window and the auto-incrementing serial number in the handshake connection request matches the historical auto-incrementing serial number within the range of the historical tag cache window, the handshake connection request is determined to be a retransmission request, and the historical input wire tag set is extracted from the range of the historical tag cache window and sent to the first mechanism node. The physical slot pointer is not involved in the comparison of the sliding window response module.

6. The data exchange system based on multi-party secure computation according to claim 5, characterized in that, After the comparison is successful, the sliding window response module uses the public key of the first agency node to perform plaintext digital signature verification on the single authorization token; After the plaintext digital signature is verified, the sliding window response module retrieves privacy data parameters based on the user query identifier; The sliding window response module generates an authorization status wire tag indicating that the signature verification has passed. Combined with the privacy data wire tag corresponding to the privacy data parameter and the virtual decoy tag, the input wire tag set is assembled.

7. The data exchange system based on multi-party secure computation according to claim 6, characterized in that, When performing dense-state logic calculations, the dense-state fusion verification module loads a pre-compiled unconditional jump execution path. The unconditional jump execution path includes an authorization status verification branch, a real business calculation branch, and a homogeneous decoy branch; The dense-state fusion verification module traverses the logic gates in the static topology template in a fixed order, so that the real business calculation branch and the isomorphic decoy branch have the same number of logic gate traversals during the calculation process.

8. The data exchange system based on multi-party secure computation according to claim 7, characterized in that, The static topology template is equipped with a multiplexer component at the topology aggregation end; The multiplexer component receives the Boolean variable output by the authorization status verification branch as a control signal; When the control signal indicates that the authorization is valid, the multiplexer component extracts the calculation result of the actual service calculation branch and exports it as an encrypted string; When the control signal indicates that the authorization is invalid, the multiplexer component extracts the calculation result of the isomorphic decoy branch and exports it as the ciphertext string; The first mechanism node uses the output decoding mapping table bound to the global circuit instance number to match and decode the ciphertext string to obtain the service determination status.

9. The data exchange system based on multi-party secure computation according to claim 1, characterized in that, After completing the dense state logic calculation, the dense state fusion verification module sends a state update instruction to the ring cache module. The first mechanism node performs arithmetic modulo operation oriented towards storage capacity boundary to achieve circular addressing on the physical slot pointer according to the state update instruction, and performs an auto-increment operation on the global circuit instance number; The first mechanism node sends a memory overwrite instruction to the operating system to set the physical memory slot position pointed to by the physical slot pointer before the arithmetic modulo operation to zero.

10. A data exchange method based on secure multi-party computation, applied to a data exchange system based on secure multi-party computation as described in any one of claims 1-9, characterized in that, The method includes the following steps: The second mechanism node generates a static topology template and a scrambled truth table bound with global circuit instance numbers, and sends the static topology template and the scrambled truth table to the first mechanism node. The first institutional node receives and stores the static topology template and the garbled truth table, and maintains a physical slot pointer for locating the local cache address; The first institutional node obtains the business rule parameters, the physical slot pointer, and the global circuit instance number to generate a single authorization token. The first agency node encapsulates the single authorization token and the global circuit instance number into a handshake connection request and sends it to the second agency node. The second mechanism node receives the handshake connection request, compares the global circuit instance number in the handshake connection request with the consumption expectation pointer maintained by the second mechanism node, and sends the input wire tag set to the first mechanism node after the comparison is successful. The first mechanism node obtains the garbled truth table based on the physical slot pointer, performs dense logic calculation on the input wire label set in conjunction with the static topology template, and outputs the calculation result.

Images