Police information recommendation method and system based on face recognition

Abstract

This invention relates to the field of police information technology, specifically disclosing a police information recommendation method and system based on facial recognition. After successful facial recognition, multi-dimensional security policy parameters, including the operating entity, task type, and environmental tags, are collected. These parameters are then combined with a target identifier to construct a joint arbitration request and homomorphically encrypted. The encrypted request is distributed to at least two independent data access control nodes. Each node performs direct computation on the ciphertext based on its local private rules, generating and returning an encrypted partial arbitration result. A secure arbitration service aggregates all encrypted results to form a final encrypted arbitration result. This result is decrypted to obtain the plaintext arbitration conclusion, which is then used to filter and output a customized information list matching the current scenario from the associated database. This invention achieves secure collaborative adjudication while fully protecting the private policies and data of all parties, ensuring the accuracy and security of information recommendations.

Description

Technical Field

[0001] This invention relates to the field of police information technology, specifically to a method and system for recommending police information based on facial recognition. Background Technology

[0002] Existing facial recognition-based police information recommendation methods typically employ a centralized or simple parallel architecture to achieve data association queries. The specific process involves facial recognition confirming identity, then using that identity as the core, directly or through middleware, initiating query requests to multiple backend business databases, and merging and recommending the returned results. While such methods can achieve information aggregation, their access control largely relies on static identity authentication before the query is initiated, or on simple filtering based on fixed rules (such as IP address or tags) at each data service entry point. When making collaborative judgments across highly sensitive data sources, it is often necessary to centrally integrate or simplify some access rules to achieve inter-system collaboration.

[0003] Existing technologies lack a way to achieve secure and reliable joint real-time arbitration when collaboratively recommending information across highly sensitive and data-sovereign police data sources, while fully protecting the private permission determination rules and sensitive data tags of each participating party, thereby eliminating the risk of unauthorized information access caused by rule exposure, blurred boundaries, or leakage of intermediate calculations. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for recommending police information based on facial recognition, so as to solve the problems mentioned above.

[0005] The objective of this invention can be achieved through the following technical solutions: A method for recommending police information based on facial recognition includes the following steps: S1: Based on the successful facial recognition trigger event, collect multi-dimensional security policy parameters in the current police operation scenario. The multi-dimensional security policy parameters include the operation subject attributes, the type of task to be performed, and the on-site environment labels. S2: Generate a joint arbitration request based on multi-dimensional security policy parameters and target personnel identifiers; encrypt the joint arbitration request using homomorphic encryption to obtain an encrypted arbitration request; S3: Send the encrypted arbitration request to the secure arbitration service, which then distributes the encrypted arbitration request to at least two independent data access control nodes; S4: Locally on each data access control node, based on the access control rules of each data privatization, the encrypted arbitration request is processed by ciphertext logic to generate the respective encrypted partial arbitration result and return it to the security arbitration service. The security arbitration service performs ciphertext aggregation calculation on all encrypted partial arbitration results to generate the final encrypted arbitration result. S5: Decrypt the final encrypted arbitration result to obtain the plaintext arbitration conclusion. Based on the predefined information security level corresponding to the plaintext arbitration conclusion, select information entries of the corresponding security level from the associated database, generate and output a customized information recommendation list for the current police operation scenario.

[0006] As a further aspect of the present invention: S2 specifically includes: The operator's attributes, task type, and on-site environment labels are serialized and assembled according to a predefined data structure, and bound to the target personnel identifier to generate a structured joint arbitration request plaintext. Based on the session key temporarily generated for this arbitration session, and combined with an encryption library that supports homomorphic operations of addition and multiplication, a homomorphic encryption component specifically for this arbitration request is instantiated. Using a homomorphic encryption component, the plaintext of the structured joint arbitration request is encrypted as a whole to obtain an encrypted arbitration request.

[0007] As a further aspect of the present invention: the use of a homomorphic encryption component to perform a total encryption operation on the structured joint arbitration request plaintext to obtain an encrypted arbitration request specifically includes: The structured joint arbitration request plaintext is divided into multiple data fragments according to predetermined fields, and each data fragment is converted into an integer vector; Generate a corresponding random blinding vector for each integer vector, and add the integer vector and the random blinding vector element by element to obtain the blinded intermediate vector; Using a homomorphic encryption component, homomorphic encryption operations supporting addition and multiplication are performed on each blinded intermediate vector to generate the corresponding encrypted vector; All encrypted vectors are assembled according to the original field order of the encrypted vectors in the structured joint arbitration request plaintext, and the assembled result is used as the encrypted arbitration request.

[0008] As a further aspect of the present invention: S3 specifically includes: Add a timestamp and a one-way progressive sequence identifier to the encrypted arbitration request to form a request packet to be distributed with verification information; The security arbitration service filters data access control nodes that are currently in a responsive state based on a pre-configured node status table and generates a list of target nodes. Based on the target node list, the request packet to be distributed is synchronously sent to each data access control node in the list through an independent secure channel, and a unique transaction fingerprint of this distribution transaction is recorded after the sending is completed; Listen for the confirmation signals returned by each data access control node. When the number of confirmation signals received within the preset timeout threshold is consistent with the number of nodes in the target node list, the encrypted arbitration request is determined to have been successfully distributed.

[0009] As a further aspect of the present invention: the generation of the target node list specifically includes: Real-time acquisition of multi-dimensional status parameters of each data access control node. These multi-dimensional status parameters include at least the current computing load rate, security level identifier, and historical response success rate. The multidimensional state parameters are compared with predefined node response thresholds, and candidate nodes that match the security level identifier and whose parameters meet the threshold requirements are dynamically filtered based on the implicit adjudication security level in the request packet to be distributed. Candidate nodes are scored using a weighted average of their current computational load rate and historical response success rate, and then sorted from highest to lowest score. Based on the predetermined number of nodes, select the top-ranked candidate nodes from the sorted list to generate the target node list.

[0010] As a further aspect of the present invention: S4 specifically includes: Each data access control node receives an encrypted arbitration request and parses it locally to restore it to the ciphertext form corresponding to the blinded intermediate vector; Based on the private permission determination rules local to the data permission control node and encoded in the form of logic gate circuits, the encrypted vector is subjected to combinational logic operations consisting of homomorphic addition and homomorphic multiplication to obtain the preliminary encryption decision value inside the node. The corresponding node's digital signature and watermark information are appended to the encrypted preliminary arbitration value to generate an encrypted partial arbitration result and send it to the secure arbitration service. After the secure arbitration service collects all the necessary encrypted local arbitration results, it performs a homomorphic weighted summation operation on them in the ciphertext state according to the predetermined arbitration weights, and uses the result as the final encrypted arbitration result.

[0011] As a further aspect of the present invention: the generation of encrypted local arbitration results and their transmission to the secure arbitration service specifically includes: A unique composite watermark is generated based on the identity identifier of the data access control node and the current timestamp accurate to milliseconds. Using the private key of the data access control node, a nested signature calculation is performed on the encrypted preliminary ruling value and the composite watermark to obtain an indivisible encrypted ruling signature composite, which is the encrypted partial arbitration result. The encrypted local arbitration result is divided into several data fragments that conform to the transmission protocol. After attaching a sequence check code to each data fragment, it is sent to the secure arbitration service through an independent secure channel.

[0012] As a further aspect of the present invention: S5 specifically includes: Using the session key corresponding to the encryption of the joint arbitration request, the final encrypted arbitration result is decrypted through the decryption function of the homomorphic encryption component. During the decryption process, the integrity of the digital signature and watermark information contained in the final encrypted arbitration result is verified simultaneously to obtain the verified plaintext arbitration conclusion. Parse the plaintext arbitration conclusion, extract the security classification identifier and dynamic permission tag, and compare the security classification identifier with the predefined dynamic information security classification mapping table to determine the range of information items that are currently allowed to be accessed and the corresponding data source. Based on the scope of the information item set, combined with dynamic permission tags and current police operation scenario parameters, a structured database query instruction is generated; The system sends the query command to the associated database, receives the returned information entries, sorts and assembles them according to predetermined business logic priorities and timeliness, and generates and outputs a customized information recommendation list with added anti-leakage tracking tags.

[0013] As a further aspect of the present invention: the generation of structured database query instructions specifically includes: The dynamic permission tag is concatenated with the parameters of the current police operation scenario, and encoded using a preset irreversible conversion rule to generate a composite query index code; Based on the data source type pointed to by the range of information entries, select a matching structured query template from the preset instruction template library; The composite query index code and the target personnel identifier are used as fill parameters and injected into the predetermined placeholders of the structured query template to form a basic query instruction. By appending a temporary access token derived from the session key to the header of the basic query command and appending a digest value of the composite query index code to the tail, the structured database query command is constructed.

[0014] A police information recommendation system based on facial recognition includes: The multi-dimensional security policy acquisition module collects multi-dimensional security policy parameters under the current police operation scenario based on the trigger event of successful face recognition. The multi-dimensional security policy parameters include the operation subject attributes, the type of task to be performed, and the on-site environment labels. The arbitration request construction and privacy encryption module generates a joint arbitration request based on multi-dimensional security policy parameters and target personnel identification; it then encrypts the joint arbitration request using homomorphic encryption to obtain an encrypted arbitration request. The secure request distribution and collaborative scheduling module sends the encrypted arbitration request to the secure arbitration service, which then distributes the encrypted arbitration request to at least two independent data access control nodes. The distributed ciphertext arbitration and aggregation module performs ciphertext logic operations on encrypted arbitration requests locally on each data access control node, based on the access control rules of each data privatization, to generate their respective encrypted partial arbitration results and return them to the secure arbitration service. The secure arbitration service performs ciphertext aggregation calculations on all encrypted partial arbitration results to generate the final encrypted arbitration result. The result decryption and graded information recommendation module decrypts the final encrypted arbitration result to obtain the plaintext arbitration conclusion. Based on the predefined information security level corresponding to the plaintext arbitration conclusion, it selects information entries of the corresponding security level from the associated database, generates and outputs a customized information recommendation list for the current police operation scenario.

[0015] The beneficial effects of this invention are: (1) This invention solves the security and trust challenges in collaborative decision-making across highly sensitive police data sources by introducing a joint arbitration mechanism based on homomorphic encryption. This method allows each data access control node to perform distributed arbitration calculations directly on encrypted data without disclosing its own core access rules and sensitive data tags. This not only strictly protects the data sovereignty and policy privacy of each participating party and effectively avoids potential access boundary conflicts and "logic short circuits" caused by rule integration or simplification, but also ensures the reliability of the final arbitration conclusion through encrypted calculation and aggregation, eliminating the risk of unauthorized information leakage from a technical perspective, and providing a solid security foundation for the integrated application of police information across departments and security levels.

[0016] (2) This invention achieves dynamic precision in information recommendation and intelligent resource scheduling. By synchronously collecting multi-dimensional real-time context parameters (specific task, on-site environment) when face recognition is triggered, and encrypting these parameters before integrating them into the arbitration process, the final authorization decision and information recommendation results are highly adapted to the current specific police operation scenario, improving the accuracy and practicality of information support. At the same time, the security arbitration service dynamically filters and weights the nodes based on their real-time load, historical success rate, and other multi-dimensional statuses, ensuring both the efficiency and stability of the arbitration process and optimizing the utilization of overall computing resources. This avoids bottlenecks in single nodes and ensures efficient and reliable response of the recommendation service in complex combat environments. Attached Figure Description

[0017] The invention will now be further described with reference to the accompanying drawings.

[0018] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a system module block diagram of the present invention. Detailed Implementation

[0019] The technical solutions of 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.

[0020] Please see Figure 1 As shown, this invention is a method for recommending police information based on facial recognition, comprising the following steps: S1: Based on the successful facial recognition trigger event, collect multi-dimensional security policy parameters in the current police operation scenario. The multi-dimensional security policy parameters include the operation subject attributes, the type of task to be performed, and the on-site environment labels. S2: Generate a joint arbitration request based on multi-dimensional security policy parameters and target personnel identifiers; encrypt the joint arbitration request using homomorphic encryption to obtain an encrypted arbitration request; S3: Send the encrypted arbitration request to the secure arbitration service, which then distributes the encrypted arbitration request to at least two independent data access control nodes; S4: Locally on each data access control node, based on the access judgment rules of each data privatization, the encrypted arbitration request is subjected to ciphertext logic operation to generate its own encrypted partial arbitration result and is returned to the security arbitration service. The security arbitration service performs ciphertext aggregation calculation on all encrypted partial arbitration results to generate the final encrypted arbitration result. S5: Decrypt the final encrypted arbitration result to obtain the plaintext arbitration conclusion. Based on the predefined information security level corresponding to the plaintext arbitration conclusion, select information entries of the corresponding security level from the associated database, generate and output a customized information recommendation list for the current police operation scenario.

[0021] In S1, based on the successful facial recognition trigger event, multi-dimensional security policy parameters are collected for the current police operation scenario. These parameters include the operator's attributes, the type of task being executed, and the on-site environment tags. Specifically, this includes: collecting "operation operator attributes," which uniquely identifies and defines the initiator of this information query and its permission criteria; determining the "type of task being executed," which defines the scope of the current operation's business scenario and is a key basis for judging the rationality of the information request; and obtaining "on-site environment tags," which describe the physical and network environment status at the time the information query request was generated to assess risk.

[0022] In S2, a joint arbitration request is generated based on multi-dimensional security policy parameters and target personnel identifiers; the joint arbitration request is then encrypted using homomorphic encryption to obtain an encrypted arbitration request, which specifically includes: First, a structured joint arbitration request plaintext is generated. The inputs for this step are the operational subject attributes, task type, and on-site environment labels collected in step S1, as well as the target personnel identifier obtained from successful face recognition. The predefined data structure uses a hierarchical combination of key-value pairs and arrays. Specifically, a root object is created; each sub-item of the operational subject attributes (unit code, permission level identifier) ​​is stored as the first set of key-value pairs in this object; the standardized code corresponding to the task type is stored as a separate key-value pair; and each sub-item of the on-site environment label (including the administrative division code mapped to latitude and longitude coordinates, timestamp, network type identifier, and lighting condition identifier) ​​is stored as a nested set of key-value pairs. Finally, the target personnel identifier is bound as an independent key-value pair to the root object. Subsequently, using standard JSON serialization methods, the root object is converted into a continuous, well-formatted string, which is the structured joint arbitration request plaintext.

[0023] Secondly, a homomorphic encryption component specifically for this arbitration request is instantiated. This process begins by temporarily generating a session key for this arbitration session. Specifically, a 256-bit binary sequence generated by a true random number generator is used as the initial seed key. The accompanying encryption library, which supports homomorphic addition and multiplication operations, refers to a pre-integrated and verified software library, such as an implementation library that satisfies a specific fully homomorphic encryption scheme. During instantiation, the initialization function of this encryption library is called, passing the previously generated session key as the master key input parameter. Internally, the encryption library uses this key and, through a series of fixed calculation processes including key expansion and modulus generation, generates and loads all the internal parameters and states necessary for subsequent homomorphic encryption and decryption. Finally, this process outputs a software object that has completed its initial configuration; this object is the dedicated homomorphic encryption component for this arbitration, encapsulating the capabilities of encryption, decryption, and addition and multiplication operations on ciphertext.

[0024] Next, the plaintext joint arbitration request is encrypted to obtain an encrypted arbitration request. This process consists of four consecutive sub-steps. The first sub-step is data segmentation and vector transformation: according to the boundaries of each field in a predefined data structure, the serialized plaintext string of the joint arbitration request is cut into multiple independent data fragments. Each data fragment is converted into a byte array through UTF-8 encoding, and then the value of each byte in the byte array is treated as an integer, arranged in order, thereby converting each data fragment into an integer vector.

[0025] The second sub-step is to generate and blind the vectors: for each integer vector obtained in the first sub-step, an independent random blinding vector is generated. The length of the random blinding vector is the same as the length of its corresponding integer vector, and the element value at each position is obtained by uniformly sampling from zero to a fixed large integer much larger than the range of plaintext data values ​​using an encrypted random number generator. Then, blinding calculation is performed: the integer vector is added element-wise to its corresponding random blinding vector; that is, the first element of the first vector is added to the first element of the second vector, the second element of the first vector is added to the second element of the second vector, and so on, performing this operation on all elements to obtain the blinded intermediate vector. This addition is a standard integer addition.

[0026] The third sub-step is to perform homomorphic encryption: using the previously instantiated homomorphic encryption component, an encryption operation is performed on each blinded intermediate vector. This encryption operation is implemented internally by the encryption component, and its process is a well-encapsulated function for the caller. Upon invocation, a blinded intermediate vector is input, and the encryption component, based on its loaded key and parameters, applies a specific mathematical function based on lattice theory to each element of the vector, ultimately outputting a ciphertext vector, i.e., the encrypted vector. This encryption process guarantees that the result obtained by decrypting two encrypted vectors after performing specific addition or multiplication operations is consistent with the result of performing the same addition or multiplication operations on their corresponding original plaintext vectors.

[0027] The fourth sub-step is assembling the encrypted arbitration request: All the encrypted vectors obtained in the third sub-step are sequentially concatenated according to the field order of their corresponding data segments in the original structured federated arbitration request plaintext. During concatenation, a predetermined, unique delimiter sequence is inserted between each encrypted vector to identify the boundaries. The resulting complete concatenated data block constitutes the encrypted arbitration request. All sensitive policy parameters and identity identifiers in this request are homomorphically encrypted.

[0028] In S3, the cryptographic arbitration request is sent to the security arbitration service, which then distributes the request to at least two independent data access control nodes. Specifically, this includes: First, a request packet with verification information is constructed for distribution. This step takes the encrypted arbitration request output from step S2 as input. The timestamp is obtained by calling the precise clock of the server hosting the secure arbitration service. The one-way progressive sequence identifier is a globally monotonically increasing integer value maintained in memory by the secure arbitration service; this value increments by 1 each time a new arbitration request is processed. The timestamp value and the one-way progressive sequence identifier value are appended in plaintext as header information before the encrypted arbitration request to form the request packet to be distributed. This operation ensures that each request packet is unique and time-sensitive, facilitating subsequent traceability and defense against replay attacks.

[0029] Secondly, a target node list is generated. This step dynamically selects suitable data access control nodes based on a pre-configured node status table and the implicit requirements of the request packets to be distributed. The pre-configured node status table is a data structure stored locally on the security arbitration service, recording the network address, port, and other basic connection information and static attributes of all registered data access control nodes. Its implementation includes four consecutive sub-operations. The first sub-operation is to obtain the multi-dimensional status parameters of each data access control node in real time: this is done by periodically (e.g., every 5 seconds) sending lightweight query heartbeat packets to each node and receiving their responses. The current computing load rate refers to the percentage of CPU usage reported by the node when responding to the heartbeat packet; the security level identifier is the level code declared by the node when registering with the security arbitration service, representing the level of sensitivity of the data it is authorized to process; the historical response success rate is calculated based on the proportion of times the node returned a valid response within the specified timeout threshold in the past 100 requests.

[0030] The second sub-operation is threshold comparison and preliminary filtering. Predefined node response thresholds include a set of fixed values: the computational load rate threshold is set to 85 (i.e., 85%), and the historical response success rate threshold is set to 95 (i.e., 95%). The implicit arbitration security level is an attribute parsed from the metadata of the encrypted arbitration request carried in the request packet to be distributed; for example, a level code "A" indicates that a high-security-level node is required. Dynamic filtering refers to iterating through all data access control nodes recorded in the status table and selecting only those nodes whose security level identifier exactly matches the arbitration security level code, whose real-time computational load rate is less than or equal to 85, and whose historical response success rate is greater than or equal to 95, marking these nodes as candidate nodes.

[0031] The third sub-operation is to calculate and sort the weighted scores. A comprehensive score is calculated for each candidate node. The specific calculation process is as follows: First, divide the node's current computational load rate by 100 to obtain a decimal between 0 and 1. Then, subtract this decimal from 1 to obtain the load score. Multiply the load score by the first weighting coefficient 0.4 to obtain the load score item. Next, divide the node's historical response success rate by 100 to obtain another decimal between 0 and 1. Multiply this decimal by the second weighting coefficient 0.6 to obtain the success rate score item. Finally, add the load score item and the success rate score item; the sum is the node's final weighted score. Based on this score, all candidate nodes are sorted in descending order of weight.

[0032] The fourth sub-operation is to determine the final list. The predetermined number of nodes is set to at least two. From the sorted candidate node list, nodes are selected sequentially, starting with the first-ranked node, until the required number of selected nodes is reached (e.g., two or three). The network address identifiers of these selected nodes are collected to form the final target node list.

[0033] Next, secure synchronous distribution is performed and transaction fingerprints are recorded. Based on the obtained list of target nodes, the secure arbitration service simultaneously sends a complete data copy of the request packet to be distributed to each node through an independent point-to-point network connection (i.e., a secure channel) pre-established for each target node in the list, based on a transport layer security protocol (version no lower than 1.2). After all transmission actions are completed, the secure arbitration service concatenates all target node identifiers involved in this distribution, the timestamp value in the request packet to be distributed, and the sequence identifier value into a string, and then applies a secure hash algorithm (e.g., SHA-256) to this string to calculate the hash value, which is recorded as the unique transaction fingerprint of this distribution transaction.

[0034] Finally, the system listens for confirmation and determines successful distribution. Immediately after sending, the secure arbitration service starts a timer with a preset timeout threshold typically set to 3000 milliseconds (3 seconds). Simultaneously, it listens for reception confirmation signals returned by all target nodes through their respective secure channels. This signal is a short data packet containing the identifier of the corresponding node and the sequence identifier value from the received request packet, encrypted using the key for this session. If, within 3000 milliseconds, the number of valid confirmation signals received by the secure arbitration service is exactly equal to the number of nodes in the target node list, the encrypted arbitration request is considered successfully distributed. If the timer times out or the number of valid confirmation signals received is insufficient, the distribution is considered a failure, and the corresponding exception handling process is initiated according to a predetermined strategy (e.g., selecting retransmission or activating a backup node).

[0035] In S4, each data access control node performs ciphertext logical operations on encrypted arbitration requests locally, based on the access control rules for each data privatization, generating its own encrypted partial arbitration results and returning them to the security arbitration service. The security arbitration service then performs ciphertext aggregation calculations on all encrypted partial arbitration results to generate the final encrypted arbitration result, which specifically includes: Each data access control node receives encrypted arbitration requests distributed by the security arbitration service and parses and restores them locally. This process takes the received complete encrypted arbitration request data packet as input. The node first identifies and removes the plaintext timestamp and sequence identifier appended to the packet header. Then, based on a predetermined delimiter sequence, it segments the remaining core ciphertext data block into multiple independent encrypted vector fragments. These encrypted vector fragments correspond to the encrypted vectors generated in step S2. Since these encrypted vectors are obtained by encrypting the blinded intermediate vector using a homomorphic encryption component, the object parsed in this step is the "ciphertext form corresponding to the blinded intermediate vector." The node caches these ciphertext vector forms in local memory, preparing them for subsequent computation.

[0036] Based on the private permission determination rules encoded in logic gate circuits locally on the data permission control node, combinational logic operations are performed on the aforementioned ciphertext vector to obtain the node's initial encryption decision value. The private permission determination rules are a set of judgment logic preset within each node and not publicly disclosed. They map policy parameters and identity identifiers to a Boolean decision of whether access is allowed or a decision score. To accommodate homomorphic encryption computation, these rules are encoded as logic gates consisting of AND, OR, and NOT gates, and further transformed mathematically into a polynomial expression containing only addition and multiplication. Specifically, for a ciphertext input vector protected by homomorphic encryption (whose plaintext counterpart is a blinded intermediate vector), the node alternately performs homomorphic addition and homomorphic multiplication operations on the ciphertext data in the order specified by this polynomial expression. Homomorphic addition refers to operating on two encrypted values ​​to generate a new encrypted value; the result of decrypting this ciphertext is equal to the sum of the two corresponding plaintext values. Homomorphic multiplication refers to operating on two encrypted values ​​to generate a new encrypted value. The decrypted result of this new encrypted value is equal to the product of the two corresponding plaintext values. The entire calculation process is completed in ciphertext mode, and nodes cannot know the plaintext content of the intermediate or final result. After the calculation is completed, one or more encrypted values ​​are obtained, which are the initial encryption decision values ​​inside the node, denoted as . The calculation process can be expressed in the following mathematical form: ;in, and The representative parsed from the encrypted arbitration request The or the first The encrypted value corresponding to a vector (or its fragments) in ciphertext form. and It is a positive integer, depending on the polynomial form transformed by the private access control rules. and These are pre-calculated integer coefficients determined by the node according to its private rules, and they correspond to specific permission determination logic. It is an index set, representing the set of indices at the 1st digit. The set of ciphertext inputs that need to participate in the homomorphic multiplication in the term multiplication. The symbol " "" indicates that the integer coefficient will be applied to the subsequent encrypted value, which is achieved by performing a specified number of homomorphic additions between the encrypted value and itself (e.g., Indicates will With oneself (Addition of homomorphisms). Symbol " " indicates a set All indexes Corresponding ciphertext value Perform consecutive homomorphic multiplication operations. The symbol " This indicates that homomorphic addition is performed on each result. The entire formula describes how, in the ciphertext state, a series of operations—weighted coefficients, ciphertext addition, and ciphertext multiplication—are used to calculate the initial encryption decision value. The process.

[0037] The encrypted preliminary ruling value is appended with the corresponding node's digital signature and watermark information to generate an encrypted partial arbitration result, which is then sent to the secure arbitration service. This step first generates a unique composite watermark: the unique identifier string of the data access control node is concatenated with the current timestamp string accurate to milliseconds, and then a secure hash algorithm is applied to this concatenated string to calculate its hash value, which is the composite watermark. Next, nested signature calculation is performed: the node uses its own asymmetric encryption private key to encrypt the preliminary ruling value. Digital signatures are generated by concatenating data with composite watermarks. The signature calculation process includes: first, calculating the hash value of the concatenated data; then, using the private key to perform an encryption operation on the hash value based on a specific mathematical problem (such as large integer factorization or elliptic curve discrete logarithm) to generate a digital signature; finally, the encrypted preliminary judgment value is... The composite watermark and the generated digital signature are bound together to form an indivisible data structure called the encrypted arbitration signature composite, which is the encrypted partial arbitration result. Finally, data fragmentation and transmission are performed: to accommodate network transmission, the encrypted partial arbitration result is divided into several data fragments of a fixed size. A sequence check code is calculated for each data fragment, obtained by performing a cyclic redundancy check on the fragment's content, and appended to the end of the fragment. All data fragments with appended check codes are sent sequentially to the secure arbitration service through a pre-established, independent secure channel between the node and the secure arbitration service.

[0038] After collecting all necessary encrypted partial arbitration results, the secure arbitration service performs a homomorphic weighted summation operation on them in ciphertext state according to predetermined arbitration weights, and uses the result as the final encrypted arbitration result. In this step, "all necessary encrypted partial arbitration results" refers to the results required to be collected according to predetermined arbitration logic (e.g., requiring the participation of all specified nodes, or reaching a certain minimum threshold). The secure arbitration service receives data fragments from each independent secure channel, verifies the integrity of each fragment using a checksum, and then reassembles the complete encrypted arbitration signature complex sent by each node. Subsequently, the service extracts the encrypted preliminary arbitration value portion from each complex. The predetermined arbitration weights are a set of predefined real numbers, each weight corresponding to a specific data access control node, and the sum of all weights is 1. The secure arbitration service performs a homomorphic weighted summation operation on each collected encrypted preliminary arbitration value in ciphertext state. (Assuming there are a total of) (each), according to its corresponding weight Perform homomorphic weighted summation. The specific calculation process is as follows: ;in, Indicates from the first The encrypted preliminary ruling value received by each data access control node. Is it pre-assigned to the first The arbitration weight of each node (a real number). This represents the total number of nodes participating in the aggregation. The operation " "Indicates that the encrypted value" With oneself Adding homomorphisms. It might be a non-integer value. In practice, it's often multiplied by a large scaling factor to convert it to an integer before performing homomorphic scalar multiplication, or handled directly in solutions that support floating-point homomorphic operations. The "+" operator represents homomorphic addition. The final calculated value is... This is the final encrypted arbitration result, an encrypted value whose decrypted plaintext reflects the comprehensive arbitration conclusion based on each node's private rules and preset weights. Throughout the entire aggregation computation process, all data involved in the computation remains encrypted. The secure arbitration service cannot know the specific preliminary ruling plaintext value of any node, nor can it know the final arbitration conclusion plaintext.

[0039] In S5, the final encrypted arbitration result is decrypted to obtain the plaintext arbitration conclusion. Based on the predefined information security level corresponding to the plaintext arbitration conclusion, information entries of the corresponding security level are selected from the associated database, generating and outputting a customized information recommendation list for the current police operation scenario, specifically including: First, the final encrypted arbitration result is decrypted and verified to obtain the verified plaintext arbitration conclusion. This step takes the final encrypted arbitration result output in step S4 as input. The decryption function of the homomorphic encryption component instantiated for this session in step S2 is invoked. During decryption, the input parameters are the final encrypted arbitration result and the session key corresponding to this arbitration session. The decryption operation is performed internally by applying a decryption algorithm paired with the session key to perform a series of mathematical transformations on the input ciphertext data, outputting a decrypted plaintext data block. This data block contains the preliminary plaintext arbitration conclusion, a digest of the composite watermark set aggregated from each node, and the aggregated digital signature. To verify integrity, the secure arbitration service uses the public keys of each data access control node, either pre-shared or obtained from certificates, to verify the aggregated digital signature. The verification process includes decrypting the signature using the public key to obtain the signature hash value, and independently calculating the hash value of the decrypted plaintext arbitration conclusion and the hash value of the concatenated composite watermark digest, comparing whether these two hash values ​​are identical. If they are completely identical and the timestamp watermark is within its validity period, the integrity verification is deemed successful, and the plaintext arbitration conclusion is considered valid; otherwise, it is considered invalid and discarded, and the process is terminated.

[0040] Next, the plaintext arbitration conclusion is parsed to determine the scope of accessible information. The input to this step is the verified plaintext arbitration conclusion obtained in the previous step. The parsing operation extracts key fields from the conclusion according to a predefined data format. The security classification identifier is a number or character code, such as "3" or "C," indicating the highest security level permitted in this arbitration. The dynamic access control label is a string or code containing finer-grained access control instructions, such as "trajectory can be checked, communication cannot be checked." The predefined dynamic information security classification mapping table is a data structure stored in memory or a configuration file that maps security classification identifiers to specific data source access permissions. By querying this mapping table, using the security classification identifier as the key, a corresponding detailed description of the range of information entries and one or more specific data source identifiers associated with that range can be retrieved.

[0041] Next, based on the determined information range, and combining the dynamic permission tags with the parameters of the current police operation scenario, a structured database query instruction is generated. This step includes four consecutive sub-operations. The first sub-operation is to generate a composite query index code. The dynamic permission tag string is concatenated with the main identifier (e.g., the task type code) of the current police operation scenario parameters obtained from step S1 to form a combined string. This combined string is encoded using a preset irreversible conversion rule, which employs a secure hash algorithm to calculate the hash value of the combined string. This hash value is the composite query index code. The second sub-operation is to select a matching structured query template. Based on the data source type (e.g., a relational database or a dedicated retrieval interface) pointed to by the information item set range determined in the previous step, a corresponding query template is selected from a preset instruction template library. This template is a query statement skeleton containing specific placeholders. The third sub-operation is to fill in the parameters to form a basic query instruction. The composite query index code and the target personnel identifier are used as filler parameters to replace the predefined placeholders in the selected query template. For example, "{ID}" in the template is replaced with the target personnel identifier, and "{INDEX}" is replaced with the composite query index code, thus forming a specific basic query instruction. The fourth sub-operation is to construct the complete structured query instruction. A temporary access token is appended to the header of the basic query instruction. This token is derived from the first sixteen characters of the hash value of the current session key. A digest value obtained by recalculating the hash of the composite query index code is appended to the tail of the basic query instruction. After adding the header and tail information, the final structured database query instruction is constructed.

[0042] Finally, the query is executed and a customized information recommendation list is generated. The structured database query command constructed in the previous step is sent to one or more associated databases through the corresponding database connection interface. After the database verifies the validity of the temporary access token, it executes the query logic and returns the information entries that meet the conditions. After receiving the returned set of information entries, they are sorted according to the predetermined business logic priority and timeliness. The sorting rules are as follows: first, the information entries are sorted in descending order according to their predefined priority scores; for entries with the same priority scores, they are then sorted in descending order according to the newness of their timestamps. After sorting, the entries are assembled into a list in a predefined format, such as a JSON array. Before outputting the list, a leak-proof tracking tag is attached to the list. This tag is obtained by performing a double hash calculation on the hash value of the list content, the current timestamp, and the query initiator identifier, and then output together. Finally, the list with the leak-proof tracking tag attached is the customized information recommendation list for the current police operation scenario, which is sent to the police terminal that initiated the request for display.

[0043] Please see Figure 2 As shown, the police information recommendation system based on facial recognition includes: The multi-dimensional security policy acquisition module collects multi-dimensional security policy parameters under the current police operation scenario based on the trigger event of successful face recognition. The multi-dimensional security policy parameters include the operation subject attributes, the type of task to be performed, and the on-site environment labels. The arbitration request construction and privacy encryption module generates a joint arbitration request based on multi-dimensional security policy parameters and target personnel identification; it then encrypts the joint arbitration request using homomorphic encryption to obtain an encrypted arbitration request. The secure request distribution and collaborative scheduling module sends the encrypted arbitration request to the secure arbitration service, which then distributes the encrypted arbitration request to at least two independent data access control nodes. The distributed ciphertext arbitration and aggregation module performs ciphertext logic operations on encrypted arbitration requests locally on each data access control node, based on the access control rules of each data privatization, to generate their respective encrypted partial arbitration results and return them to the secure arbitration service. The secure arbitration service performs ciphertext aggregation calculations on all encrypted partial arbitration results to generate the final encrypted arbitration result. The result decryption and graded information recommendation module decrypts the final encrypted arbitration result to obtain the plaintext arbitration conclusion. Based on the predefined information security level corresponding to the plaintext arbitration conclusion, it selects information entries of the corresponding security level from the associated database, generates and outputs a customized information recommendation list for the current police operation scenario.

[0044] The working principle of this invention is as follows: First, after successful facial recognition, multi-dimensional security strategy parameters such as the operator's attributes, the type of task being performed, and on-site environmental labels are collected. Then, these parameters, along with the target personnel identifier, are used to construct a joint arbitration request, which is then encrypted and blinded using homomorphic encryption technology, forming an encrypted arbitration request. This request is sent to a secure arbitration service, dynamically filtered, and distributed to at least two independent data access control nodes. Each node, based on its private access control rules, directly performs homomorphic addition and multiplication operations on the encrypted data, generating encrypted partial arbitration results and returning them with signatures. The secure arbitration service performs weighted aggregation of all partial results in the encrypted state to generate the final encrypted arbitration result. Finally, the result is decrypted and verified. Based on the security level identifier in the obtained plaintext arbitration conclusion, a customized information recommendation list matching the current police scenario is securely retrieved from the associated database and output, achieving accurate and secure information recommendation while protecting the data privacy and access control rules of all parties.

[0045] The specific implementation of this invention system shall strictly comply with national laws and regulations concerning personal information protection, cybersecurity, and data security, ensuring the legality, legitimacy, and necessity of data processing. This reflects consideration for legal compliance.

[0046] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A method for recommending police information based on facial recognition, characterized in that, Includes the following steps: S1: Based on the successful facial recognition trigger event, collect multi-dimensional security policy parameters in the current police operation scenario. The multi-dimensional security policy parameters include the operation subject attributes, the type of task to be performed, and the on-site environment labels. S2: Generate a joint arbitration request based on multi-dimensional security policy parameters and target personnel identification; The joint arbitration request is encrypted using homomorphic encryption to obtain the encrypted arbitration request; S3: Send the encrypted arbitration request to the secure arbitration service, which then distributes the encrypted arbitration request to at least two independent data access control nodes; S4: Locally on each data access control node, based on the access control rules of each data privatization, the encrypted arbitration request is processed by ciphertext logic to generate the respective encrypted partial arbitration result and return it to the security arbitration service. The security arbitration service performs ciphertext aggregation calculation on all encrypted partial arbitration results to generate the final encrypted arbitration result. S5: Decrypt the final encrypted arbitration result to obtain the plaintext arbitration conclusion. Based on the predefined information security level corresponding to the plaintext arbitration conclusion, select information entries of the corresponding security level from the associated database, generate and output a customized information recommendation list for the current police operation scenario.

2. The police information recommendation method based on face recognition according to claim 1, characterized in that, S2 specifically includes: The operator's attributes, task type, and on-site environment labels are serialized and assembled according to a predefined data structure, and bound to the target personnel identifier to generate a structured joint arbitration request plaintext. Based on the session key temporarily generated for this arbitration session, and combined with an encryption library that supports homomorphic operations of addition and multiplication, a homomorphic encryption component specifically for this arbitration request is instantiated. Using a homomorphic encryption component, the plaintext of the structured joint arbitration request is encrypted as a whole to obtain an encrypted arbitration request.

3. The police information recommendation method based on face recognition according to claim 2, characterized in that, The process of using a homomorphic encryption component to perform a comprehensive encryption operation on the plaintext of the structured joint arbitration request to obtain an encrypted arbitration request specifically includes: The structured joint arbitration request plaintext is divided into multiple data fragments according to predetermined fields, and each data fragment is converted into an integer vector; Generate a corresponding random blinding vector for each integer vector, and add the integer vector and the random blinding vector element by element to obtain the blinded intermediate vector; Using a homomorphic encryption component, homomorphic encryption operations supporting addition and multiplication are performed on each blinded intermediate vector to generate the corresponding encrypted vector; All encrypted vectors are assembled according to the original field order of the encrypted vectors in the structured joint arbitration request plaintext, and the assembled result is used as the encrypted arbitration request.

4. The police information recommendation method based on face recognition according to claim 1, characterized in that, S3 specifically includes: Add a timestamp and a one-way progressive sequence identifier to the encrypted arbitration request to form a request packet to be distributed with verification information; The security arbitration service filters data access control nodes that are currently in a responsive state based on a pre-configured node status table and generates a list of target nodes. Based on the target node list, the request packet to be distributed is synchronously sent to each data access control node in the list through an independent secure channel, and a unique transaction fingerprint of this distribution transaction is recorded after the sending is completed; Listen for the confirmation signals returned by each data access control node. When the number of confirmation signals received within the preset timeout threshold is consistent with the number of nodes in the target node list, the encrypted arbitration request is determined to have been successfully distributed.

5. The police information recommendation method based on face recognition according to claim 4, characterized in that, The generated target node list specifically includes: Real-time acquisition of multi-dimensional status parameters of each data access control node. These multi-dimensional status parameters include at least the current computing load rate, security level identifier, and historical response success rate. The multidimensional state parameters are compared with predefined node response thresholds, and candidate nodes that match the security level identifier and whose parameters meet the threshold requirements are dynamically filtered based on the implicit adjudication security level in the request packet to be distributed. Candidate nodes are scored using a weighted average of their current computational load rate and historical response success rate, and then sorted from highest to lowest score. Based on the predetermined number of nodes, select the top-ranked candidate nodes from the sorted list to generate the target node list.

6. The police information recommendation method based on face recognition according to claim 1, characterized in that, S4 specifically includes: Each data access control node receives an encrypted arbitration request and parses it locally to restore it to the ciphertext form corresponding to the blinded intermediate vector; Based on the private permission determination rules local to the data permission control node and encoded in the form of logic gate circuits, the encrypted vector is subjected to combinational logic operations consisting of homomorphic addition and homomorphic multiplication to obtain the preliminary encryption decision value inside the node. The corresponding node's digital signature and watermark information are appended to the encrypted preliminary arbitration value to generate an encrypted partial arbitration result and send it to the secure arbitration service. After the secure arbitration service collects all the necessary encrypted local arbitration results, it performs a homomorphic weighted summation operation on them in the ciphertext state according to the predetermined arbitration weights, and uses the result as the final encrypted arbitration result.

7. The police information recommendation method based on face recognition according to claim 6, characterized in that, The process of generating an encrypted local arbitration result and sending it to the secure arbitration service specifically includes: A unique composite watermark is generated based on the identity identifier of the data access control node and the current timestamp accurate to milliseconds. Using the private key of the data access control node, a nested signature calculation is performed on the encrypted preliminary ruling value and the composite watermark to obtain an indivisible encrypted ruling signature composite, which is the encrypted partial arbitration result. The encrypted local arbitration result is divided into several data fragments that conform to the transmission protocol. After attaching a sequence check code to each data fragment, it is sent to the secure arbitration service through an independent secure channel.

8. The police information recommendation method based on face recognition according to claim 1, characterized in that, S5 specifically includes: Using the session key corresponding to the encryption of the joint arbitration request, the final encrypted arbitration result is decrypted through the decryption function of the homomorphic encryption component. During the decryption process, the integrity of the digital signature and watermark information contained in the final encrypted arbitration result is verified simultaneously to obtain the verified plaintext arbitration conclusion. Parse the plaintext arbitration conclusion, extract the security classification identifier and dynamic permission tag, and compare the security classification identifier with the predefined dynamic information security classification mapping table to determine the range of information items that are currently allowed to be accessed and the corresponding data source. Based on the scope of the information item set, combined with dynamic permission tags and current police operation scenario parameters, a structured database query instruction is generated; The system sends the query command to the associated database, receives the returned information entries, sorts and assembles them according to predetermined business logic priorities and timeliness, and generates and outputs a customized information recommendation list with added anti-leakage tracking tags.

9. The police information recommendation method based on face recognition according to claim 8, characterized in that, The generation of structured database query instructions specifically includes: The dynamic permission tag is concatenated with the parameters of the current police operation scenario, and encoded using a preset irreversible conversion rule to generate a composite query index code; Based on the data source type pointed to by the range of information entries, select a matching structured query template from the preset instruction template library; The composite query index code and the target personnel identifier are used as fill parameters and injected into the predetermined placeholders of the structured query template to form a basic query instruction. By appending a temporary access token derived from the session key to the header of the basic query command and appending a digest value of the composite query index code to the tail, the structured database query command is constructed.

10. A police information recommendation system based on facial recognition, characterized in that, The method for performing the police information recommendation method based on face recognition as described in any one of claims 1-9 includes: The multi-dimensional security policy acquisition module collects multi-dimensional security policy parameters under the current police operation scenario based on the trigger event of successful face recognition. The multi-dimensional security policy parameters include the operation subject attributes, the type of task to be performed, and the on-site environment labels. The arbitration request construction and privacy encryption module generates a joint arbitration request based on multi-dimensional security policy parameters and target personnel identification; it then encrypts the joint arbitration request using homomorphic encryption to obtain an encrypted arbitration request. The secure request distribution and collaborative scheduling module sends the encrypted arbitration request to the secure arbitration service, which then distributes the encrypted arbitration request to at least two independent data access control nodes. The distributed ciphertext arbitration and aggregation module performs ciphertext logic operations on encrypted arbitration requests locally on each data access control node, based on the access control rules of each data privatization, to generate their respective encrypted partial arbitration results and return them to the secure arbitration service. The secure arbitration service performs ciphertext aggregation calculations on all encrypted partial arbitration results to generate the final encrypted arbitration result. The result decryption and graded information recommendation module decrypts the final encrypted arbitration result to obtain the plaintext arbitration conclusion. Based on the predefined information security level corresponding to the plaintext arbitration conclusion, it selects information entries of the corresponding security level from the associated database, generates and outputs a customized information recommendation list for the current police operation scenario.

Images