Homomorphic encryption-based spatiotemporal big data distributed privacy computing system and device

Abstract

The application provides a space-time big data distributed privacy computing system and equipment based on homomorphic encryption, and relates to the technical field of data processing. It comprises a space privacy computing operator, which is used for the computing characteristics of vector data and raster data, and realizes homomorphic encryption computing of vector data and raster data by using TFHE homomorphic encryption algorithm and CKKS homomorphic encryption algorithm respectively; a distributed space privacy computing framework is used, each node of the distributed space privacy computing framework is respectively provided with a space privacy computing operator, and the distributed space privacy computing framework realizes multi-node task parallelization and safe cooperation based on a key management mechanism under a distributed environment and a remote Boolean decryption protocol based on inadvertent transmission; each node in the distributed space privacy computing framework is provided with a GPU acceleration architecture, the GPU acceleration architecture is respectively provided with a corresponding CUDA heterogeneous computing optimization scheme for the Boolean logic operation of vector calculation and the matrix operation characteristics of raster calculation, and the processing efficiency of a single node is improved.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a spatiotemporal big data distributed privacy computing system and device based on homomorphic encryption. Background Technology

[0002] With the development of satellite navigation and remote sensing technologies and the widespread use of mobile devices, spatiotemporal big data has become a key production factor in fields such as smart cities and traffic management. However, the individual trajectories and sensitive areas contained in spatiotemporal data significantly increase the risk of leakage of highly sensitive privacy data. Therefore, how to break down data silos and achieve efficient computation of massive spatiotemporal data while effectively protecting data privacy and security has become an important issue that urgently needs in-depth research. Spatial privacy computing technology can perform computational processing on spatiotemporally sensitive privacy data while ensuring data security, achieving "computable but invisible data." However, existing spatial privacy computing still faces many limitations: most existing work is designed and optimized for specific spatial operations, making it difficult to support general complex spatial operations; the high temporal and spatial complexity of encrypted computation leads to severe single-machine performance bottlenecks during computation, and existing spatiotemporal distributed computing frameworks do not provide support for privacy computing.

[0003] Therefore, there is an urgent need for a spatiotemporal big data distributed privacy computing system. Summary of the Invention

[0004] This invention provides a spatiotemporal big data distributed privacy computing system and device based on homomorphic encryption, to at least partially solve the above-mentioned problems. This invention aims to construct a spatiotemporal big data privacy computing framework based on a fully homomorphic encryption algorithm and the Spark computing engine. By implementing spatial privacy computing operators and constructing a multi-layered acceleration mechanism, it achieves an efficient and secure spatial computing paradigm.

[0005] The first aspect of this invention provides a spatiotemporal big data distributed privacy computing system based on homomorphic encryption, the system comprising:

[0006] Spatial privacy computation operators are used to achieve homomorphic encrypted computation of vector data and raster data by employing the TFHE homomorphic encryption algorithm and the CKKS homomorphic encryption algorithm, respectively, based on the computational characteristics of vector data and raster data.

[0007] The system employs a distributed spatial privacy computing framework. Each node of this framework is equipped with a spatial privacy computing operator. The framework utilizes a distributed key management mechanism and a remote Boolean decryption protocol based on unintentional transmission to achieve multi-node task parallelization and secure collaboration. Each node in the framework is equipped with a GPU-accelerated architecture. This architecture features corresponding CUDA heterogeneous computing optimization schemes tailored to the Boolean logic operations of vector computing and the matrix operations of raster computing, thereby improving the processing efficiency of a single node.

[0008] Optionally, a space privacy computation operator based on fully homomorphic encryption is used for:

[0009] For vector data, plaintext data is encrypted using the TFHE type system. The TFHE type system provides an abstraction of encrypted data types for the implementation of the TFHE OGR class library, and the TFHE OGR class library provides an abstraction of encrypted geometric types for TFHE spatial predicate computation. By constructing an encrypted nine-intersection model and a spatial relation computation algorithm, the TFHE spatial predicate computation algorithm is realized.

[0010] The TFHE type system includes the following basic types: TFHEBool, TFHEInt64, Int32, and TFHEDecimal.

[0011] Optionally, plaintext data is encrypted using the TFHE type system, including:

[0012] The type registration framework based on the TFHERegisteredType generic interface provides a dynamic type registration mechanism, enabling encrypted data types within a process to use a unified encryption context; it manages the context of homomorphic encryption based on a process-level globally unique TFHEContext singleton structure; it constructs a type system based on TFHE encryption by encapsulating basic FHE ciphertext pointers; and for spatial data, it uses a floating-point encryption method based on scaling factors to convert floating-point operations into integer field operations through scaling.

[0013] Optionally, the TFHE type system provides an abstraction of cryptographic data types for the implementation of the TFHE OGR class library, including:

[0014] The TFHEGeometry interface class, corresponding to the functionality of the Geometry class, defines the construction and calculation operations of geometric objects in the encrypted domain. Subclasses of the TFHEGeometry interface class implement the specific expression of spatial data through polymorphism: TFHEPoint encapsulates the coordinates of encrypted single points, TFHELineString manages the ordered encrypted point sequences to form line elements, and the TFHEPolygon class describes the spatial morphology of planar elements through a set of encrypted linear rings. Among them, TFHELinearRing, as a subclass of TFHELineString, additionally implements the ring closure verification method.

[0015] For coordinate data, the TFHECoordinate class is used to uniformly manage encrypted coordinate data. Its internal coordinate values ​​are stored and calculated using the TFHEDecimal class in the TFHE type system, which is based on TFHE and ensures that all spatial operations are performed in encrypted state.

[0016] Optionally, the TFHE OGR library provides an abstraction of cryptographic geometry types for TFHE spatial predicate computation. By constructing a cryptographic nine-intersection model and a spatial relation computation algorithm, it implements the TFHE spatial predicate computation algorithm, including:

[0017] Predicate initialization is based on dimensional filtering and bounding rectangle filtering;

[0018] The computation of homomorphic encryption space is reduced by using a variety of geometric feature-based pruning methods. These methods include point-to-point optimization, point-to-non-point optimization, line-to-polygon endpoint detection, and polygon-to-polygon vertex pre-detection.

[0019] Determine whether the bounding rectangles of two cryptographic geometric objects involved in the operation intersect;

[0020] In the case of intersection, line features are decomposed into line segment geometry, and polygon features are decomposed into rings, marking whether it is an inner ring or an outer ring and the polygon number to which it belongs; calculate the SegmentStrings intersection set of the two encrypted geometry objects to generate topology nodes; each topology node encapsulates an intersection point in a SegmentString and records its dimensional relationship in the two encrypted geometry objects.

[0021] Traverse all topological nodes to construct a topological graph model, which includes RelateNode and RelateEdge. RelateNode represents the topological state of the neighborhood of an intersection point, and RelateEdge describes the topological position of the geometric boundary.

[0022] By recursively analyzing the geometric positions on both sides of RelateEdge, the boundary-to-boundary dimensional interaction pattern is derived, and the IM is updated progressively.

[0023] The judgment result is output by the DE-9IM pattern matching rule of predicate P.

[0024] Optionally, for raster data, a basic raster homomorphic encryption operator is implemented based on the SEAL homomorphic encryption library, wherein the basic raster homomorphic encryption operator includes:

[0025] PlainText layer: Completes the conversion of raster data to CKKS plaintext field, and encodes floating-point arrays into polynomial plaintext;

[0026] CipherText layer: Converts CKKS plaintext into CKKS ciphertext using an encryption algorithm;

[0027] HECrypto layer: Implements ciphertext operation operators and key generation management, providing a unified abstract interface.

[0028] Optionally, the distributed spatial privacy computing framework integrates spatial privacy computing capabilities through the SpatialFHE module. The FHEHelper component serves as the core interaction interface, enabling collaboration between Spark and SpatialFHE. Cross-language adaptation technology is used to embed the spatial privacy computing operators into the Spark distributed framework. SpatialFHEManager has two sub-modules, MatFHEManager and VecFHEManager, which respectively integrate SEALCrypto / PhantomCrypto and TFHE_GEOS to support homomorphic encrypted computing of raster and vector data.

[0029] Optionally, at the distributed level, the distributed spatial privacy computing framework is based on the elastic scaling framework of Spark+Sedona. Through RDD data sharding and distributed key sharing mechanism, it realizes parallel processing and secure collaboration of spatial privacy computing tasks; it designs and implements a remote Boolean decryption method based on the OT protocol, which supports the encrypted execution of short-circuit optimization strategy while maintaining the security boundary of the encryption system.

[0030] At the single-node acceleration level, GPU acceleration solutions are set up for the characteristics of vector and raster data respectively: the CUDA architecture based on TFHE-rs is optimized to realize the heterogeneous execution of the vector homomorphic encryption core operation on GPU; the NWT transformation of the Phantom framework and the CUDA kernel fusion technology are used to reduce the polynomial multiplication complexity of raster data and reduce memory and computational overhead.

[0031] A second aspect of the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executed, implements the spatiotemporal big data distributed privacy computing system based on homomorphic encryption as described in the first aspect of the present invention.

[0032] A third aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the spatiotemporal big data distributed privacy computing system based on homomorphic encryption as described in the first aspect of the present invention.

[0033] The fourth aspect of the present invention provides a computer program product, including a computer program / instructions, which are implemented by a processor as described in the first aspect of the present invention in a spatiotemporal big data distributed privacy computing system based on homomorphic encryption.

[0034] This invention constructs a spatial privacy computation operator based on fully homomorphic encryption. Addressing the differences in computational characteristics between vector and raster data, this invention employs different homomorphic encryption schemes to achieve their respective privacy computation capabilities. For vector computation, this invention selects the TFHE fully homomorphic encryption algorithm, constructs a TFHE basic type system, and designs an OGR class library containing encrypted geometric types, capable of supporting ciphertext spatial relationship determination (such as intersection, inclusion, etc.) and complex spatial relationship operations based on the nine-intersection model. Vector spatial privacy computation supports 30 vector construction and computation functions and 10 types of spatial predicate functions. For raster computation, this invention uses the CKKS homomorphic encryption scheme and SIMD encoding method to encode multidimensional raster data into complex polynomials, achieving efficient raster algebraic operations on encrypted rasters, supporting 9 different encrypted raster map algebraic functions.

[0035] To overcome the efficiency bottleneck of homomorphic encryption, this invention constructs a multi-layered acceleration optimization system of "multi-node distributed expansion + single-node heterogeneous acceleration". For distributed expansion, this invention builds a distributed privacy computing framework based on Spark-Sedona, proposing a key management mechanism in a distributed environment and a remote Boolean decryption protocol based on unintentional transfer (OT), achieving multi-node task parallelization and secure collaboration. For single-node acceleration, this invention introduces GPU heterogeneous acceleration technology, integrating different CUDA acceleration schemes for TFHE and CKKS homomorphic encryption schemes respectively, significantly improving the efficiency of single-node spatial privacy computing.

[0036] This invention enables the engineering implementation of a generalized spatiotemporal privacy computing framework, breaking through the technical barriers of complex spatial computation in encrypted environments. It provides an engineering solution for cross-domain data collaboration in scenarios such as smart cities and public health, ensuring that data is "usable but not visible," which is of great significance for releasing the value of data elements and promoting the high-quality development of the digital economy. Attached Figure Description

[0037] To more clearly illustrate the technical solution of the present invention, the accompanying drawings used in the description of the present invention will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0038] Figure 1 This is a schematic diagram of the vector operator computing architecture and process of the spatiotemporal big data distributed privacy computing system based on homomorphic encryption provided by the present invention;

[0039] Figure 2 This is a schematic diagram of the overall process of the privacy space predicate judgment algorithm of the spatiotemporal big data distributed privacy computing system based on homomorphic encryption provided by the present invention;

[0040] Figure 3 This is a schematic diagram of the SIMD encoding and encryption process for raster data in a spatiotemporal big data distributed privacy computing system based on homomorphic encryption provided by the present invention;

[0041] Figure 4 This is a schematic diagram of the distributed spatial privacy computing framework of the spatiotemporal big data distributed privacy computing system based on homomorphic encryption provided by the present invention. Detailed Implementation

[0042] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0043] This invention provides a spatiotemporal big data distributed privacy computing system based on homomorphic encryption, the system comprising:

[0044] Spatial privacy computation operators are used to achieve homomorphic encrypted computation of vector data and raster data by employing the TFHE (Torus-based Fully Homomorphic Encryption) ring-based fully homomorphic encryption algorithm and the CKKS (Cheon-Kim-Kim-Song) homomorphic encryption algorithm, respectively, based on the computational characteristics of vector data and raster data.

[0045] The system employs a distributed spatial privacy computing framework. Each node of this framework is equipped with a spatial privacy computing operator. The framework utilizes a distributed key management mechanism and a remote Boolean decryption protocol based on unintentional transmission to achieve multi-node task parallelization and secure collaboration. Each node in the framework is equipped with a GPU-accelerated architecture. This architecture features corresponding CUDA heterogeneous computing optimization schemes tailored to the Boolean logic operations of vector computing and the matrix operations of raster computing, thereby improving the processing efficiency of a single node.

[0046] In this embodiment of the invention, different homomorphic encryption schemes are employed to achieve privacy-preserving computation capabilities for vector and raster data, taking into account their different computational characteristics. For vector computation, the TFHE fully homomorphic encryption algorithm is selected, a basic TFHE type system is constructed, and an OGR class library containing encrypted geometric types is designed based on this. This library supports ciphertext spatial relationship determination (such as intersection, inclusion, etc.) and complex spatial relationship operations based on the nine-intersection model. Vector spatial privacy computation supports 30 vector construction and computation functions and 10 types of spatial predicate functions. For raster computation, based on the CKKS homomorphic encryption scheme and SIMD encoding, multidimensional raster data is encoded into complex polynomials, enabling efficient raster algebraic operations on encrypted rasters. A total of 9 different encrypted raster map algebraic functions are supported.

[0047] Vector data, as a fundamental data type in Geographic Information Systems (GIS), is primarily used to represent discrete geographic features such as points, lines, and polygons. Conventional computations of vector data encompass buffer construction, overlay analysis between vector data, and spatial relationship determination. These computational processes involve various nonlinear operations (such as division, square root, and size comparison). TFHE can encapsulate various nonlinear unit functions through PBS, demonstrating significant advantages in handling vector computation scenarios. Therefore, this invention achieves privacy-preserving computation of vector data through the modification and encapsulation of the TFHE-rs homomorphic encryption computation library. The overall architecture and computational process are as follows: Figure 1 As shown, plaintext data is encrypted using the Fully Homomorphic Encryption Type System (TFHE Type System). This type system provides an abstraction of encrypted data types for the implementation of the Fully Homomorphic Encryption Geometry Library (TFHE OGR Library). The OGR Library further provides an abstraction of encrypted geometry types for the more complex Fully Homomorphic Encryption Spatial Predicate Computation (TFHE Spatial Predicate Computation).

[0048] To achieve homomorphic encrypted computation of spatial data, it is first necessary to establish a type system adapted to the encrypted computation paradigm. In traditional programming paradigms, a series of basic data types (such as integers and floating-point numbers) are typically used to represent spatial data. However, in a homomorphic encryption environment, these traditional data types need to be redesigned to adapt to the computational characteristics under encryption. Based on the TFHE-rs homomorphic encryption library, this invention implements a memory-safe and easy-to-use TFHE basic type system (such as integers and floating-point numbers). Figure 1 As shown), specifically, it consists of the following three parts:

[0049] Type Registration Framework: This invention proposes a type registration framework based on the fully homomorphic encrypted type registration mechanism (TFHERegisteredType) generic interface. This framework provides a dynamic type registration mechanism, enabling encrypted data types within a process to use a unified encryption context and avoiding repeated reading of encryption keys.

[0050] Fully Homomorphic Encryption Context (TFHEContext) Management: This invention designs a process-level globally unique TFHEContext singleton structure to manage the homomorphic encryption context, realizing centralized control of key parameters and ensuring the security and consistency of the key during the calculation process.

[0051] Basic Encryption Type System: This invention constructs a type system based on TFHE encryption by encapsulating basic FHE ciphertext pointers. This type system ensures thread and memory safety during homomorphic encryption operations and hides these details, significantly improving the usability of privacy computing systems.

[0052] Coordinates used in spatial computation are mostly floating-point types, but the TFHE (Fully Homomorphic Encryption) scheme natively only supports integer operations and cannot directly process floating-point data. To solve this problem, this invention proposes a floating-point encryption method based on scaling factors. This method converts floating-point operations into integer field operations through scaling, while maintaining a certain level of computational precision. Given a floating-point number... f and scaling factor Δ=10 k The encryption and decryption process is as follows:

[0053] encryption: ;

[0054] Deciphering: ;

[0055] in, This indicates TFHE encryption. This indicates TFHE decryption. c f Indicate plaintext f The corresponding ciphertext.

[0056] In this way, the present invention ensures that the precision of some decimal places is preserved during encryption and decryption. When performing various operations, the present invention also needs to pay attention to the uniformity and handling of scaling factors, and formulate different rules for different operations:

[0057] Addition / subtraction / comparison operations: Performed after aligning with scaling factors. ;

[0058] Multiplication and division require scaling adjustments: ;in, , Represent plaintext , The corresponding ciphertext.

[0059] The actual scaling process uses a binary scaling factor Δ=2. N This transforms ciphertext multiplication and division in scaling operations into fast bitwise operations, resulting in a 25-fold performance improvement in a single calculation. The specific implementation of encryption, decryption, and arithmetic operations in TFHEDecimal is shown in Algorithm 1.

[0060] In summary, the basic types of the TFHE core implemented by this invention include:

[0061] Fully Homomorphic Encrypted Boolean (TFHEBool): An encrypted Boolean type that supports basic Boolean operations such as AND, OR, and NOT in an encrypted environment.

[0062] Fully homomorphic encrypted integer types (TFHEInt64, Int32): Encrypted integer types that support arithmetic operations, size comparisons, and bitwise operations in an encrypted environment.

[0063] Fully Homomorphic Encrypted Fixed-Point Number (TFHEDecimal): An encrypted fixed-point number implemented by wrapping and scaling an Int64.

[0064] OGR (Open GIS Simple Features Reference Implementation) is a standard library for geospatial data processing. To achieve spatial data processing in an encrypted state, this invention implements the core functions of OGR based on TFHE, referred to as the Fully Homomorphic Encrypted Geometry Library (TFHE-OGR Library) (e.g.) Figure 1 As mentioned above, while inheriting the traditional OGR interface, it integrates the TFHE type system proposed above to realize geometric type abstraction in the encrypted state.

[0065] At the implementation level, this embodiment of the invention first introduces the fully homomorphic encrypted geometry (TFHEGeometry) interface class, which corresponds to the functionality of the Geometry class. This abstract base class defines the construction and calculation operations of geometric objects in the encrypted domain. Its subclasses implement the specific expression of spatial data through polymorphism: the fully homomorphic encrypted point (TFHEPoint) encapsulates the coordinates of encrypted single points; the fully homomorphic encrypted polyline (TFHELineString) manages the ordered sequence of encrypted points to form line features; and the fully homomorphic encrypted polygon (TFHEPolygon) class describes the spatial shape of planar features through the encrypted closed polyline (TFHELinearRing) collection. Among them, TFHELinearRing, as a subclass of TFHELineString, additionally implements the loop closure verification method isClosed().

[0066] In terms of coordinate representation, this invention uses TFHECoordinate to uniformly manage encrypted coordinate data. Its internal coordinate values ​​are stored and calculated using the TFHEDecimal class in the TFHE type system, ensuring that all spatial operations are performed in encrypted form.

[0067] The TFHE-OGR library supports 7 privacy vector constructors and 21 privacy space computation functions, as detailed in Tables 1 and 2:

[0068] Table 1 Privacy Vector Constructor Table

[0069]

[0070] Table 2 Vector Privacy Calculation Functions

[0071]

[0072] Based on the basic type system and basic spatial types of TFHE, spatial predicate computation based on TFHE can be further implemented, such as... Figure 1 As shown, this embodiment of the invention constructs an encrypted spatial relationship matrix and combines it with a multi-layer optimization strategy to achieve efficient spatial relationship determination of ciphertext geometric objects in a homomorphic encryption environment.

[0073] The core of the spatial relation computation algorithm is to construct a nine-intersection model of the two geometric objects involved in the computation and an intersection pattern matrix (IM). The IM describes the intersection relationship between the interiors, boundaries, and exteriors of two geometric objects. Its core idea is to encode the topological relationship between two geometric objects a and b into a 3×3 matrix. The matrix elements represent the intersection dimension (dim) of the interiors, boundaries, and exteriors of a and b, respectively. The dimension values ​​are -1 (empty set), 0 (point), 1 (line), and 2 (surface), or simplified using the logical symbols T (not empty), F (empty), and * (arbitrary). Based on the DE-9IM matrix, spatial predicates can be abstracted as logical combinations of specific pattern matrices to support efficient semantic queries of spatial relations. In the homomorphic encryption implementation, this invention uses TFHEInt32 to encrypt each cell of the pattern matrix and performs pattern matching in the ciphertext environment to obtain a ciphertext spatial predicate judgment result of type TFHEBool. The spatial predicate functions and their corresponding pattern matrices supported by this invention in the encryption environment are shown in Table 3.

[0074] Table 3 shows the supported privacy vector space predicates and corresponding pattern matrices.

[0075]

[0076] This invention adaptively improves the spatial relationship judgment algorithm in the GEOS spatial computing library, enabling it to run efficiently in encrypted computing environments. Considering the high computational overhead of spatial operations in homomorphic encryption environments, avoiding the construction of the full topology during spatial relationship judgment is crucial for algorithm performance. Therefore, this invention improves algorithm efficiency by setting multiple pruning exits to reduce spatial computation in homomorphic encryption. Experimental data shows that the time cost of a single pruning operation is only about 1 / 1000 of that of a single data operation in a homomorphic encryption environment.

[0077] Based on the above optimization ideas, the privacy space predicate judgment algorithm designed in this invention (such as...) Figure 2 The algorithm (as shown) consists of four core modules: predicate initialization, geometric feature pruning, topology construction, and topology node evaluation. It constructs a multi-pruning mechanism through four dimensions: dimensional constraints of spatial predicates, outer rectangle relationship analysis, dimensional relationships between geometric parameters, and predicate IM pattern matching, effectively reducing topology construction and evaluation in encrypted environments. The detailed algorithm implementation is as follows:

[0078] 1) Predicate initialization

[0079] Predicate initialization is divided into two parts: dimension filtering and outer rectangle filtering.

[0080] Dimension Filtering: For specific predicates such as Covers, Contains, Crosses, Overlaps, and Touches, geometric dimension constraints are first verified for rapid filtering. Some geometric dimension constraints are shown in Table 3. Covers / Contains requires that the dimension of geometry A is strictly greater than that of B (dim(A)>dim(B)), while Within / CoveredBy does the opposite; Crosses requires that the two geometries are not both points or faces; Overlaps requires that the dimensions be the same; Touches excludes the case where the two encrypted geometries are two points. When the dimension constraints are not satisfied, the predicate calculation result is returned directly.

[0081] Enclosing Rectangle Filtering: Computationally find the minimum enclosing rectangle of geometric A and B, and make a preliminary judgment based on the envelope relation constraints of the predicate (e.g., Disjoint requires disjoint envelopes, Contains requires contained envelopes). If the relationship between the enclosing rectangles does not meet the necessary conditions of the predicate, the calculation is terminated immediately and a negative result is returned.

[0082] 2) Optimized pruning based on geometric features

[0083] This invention designs several pruning methods based on geometric features to further enhance the algorithm's ability to calculate spatial relationships before constructing the topology. Since the IM matrix of some predicates only requires calculating a few values ​​to obtain the final result (as shown in Table 3), for example, intersection only requires calculating the intersection of any interior and boundary. This characteristic allows the result to be obtained without completing all IM calculations, which is also an effective pruning technique.

[0084] Point-to-point optimization: If both geometric features are point types, coordinate matching is performed directly, the IM matrix is ​​updated, and the result is returned.

[0085] Point-to-non-point optimization: For line features, the point-line position relationship is determined by solving the projection of the point onto the line segment and combining the tolerance; for polygon features, the Ray-Crossing Algorithm is used to determine the position of the point in the polygon feature: first, it is checked whether the point is inside the outer ring. If it is true, the inner ring is checked layer by layer to exclude the hole area. Finally, it is determined whether the point is inside, on the boundary or outside of the polygon, and the predicate judgment result is returned.

[0086] Line-Polygon Endpoint Detection: When it is necessary to verify the Exterior relationship, extract the endpoints of the line features and detect whether they are located outside the polygon features, and update the IM.

[0087] Polygon-Polygon Vertex Pre-detection: For relationships between polygons, prioritize detecting the position of the loop vertex in the other geometry. If a vertex is located outside the other, update the Exterior-Exterior element of the IM.

[0088] 3) Incremental construction of topology

[0089] The specific algorithm for constructing the topology structure includes the following steps: First, before calculating the topology structure, it's possible to determine whether the outer rectangles of the two participating encrypted geometric objects intersect. If they don't intersect, there's no need to calculate the topology structure. Then, SegmentStrings are extracted: line features are decomposed into line segments, and polygon features are decomposed into rings, marked as inner or outer rings and their corresponding polygon numbers. Next, the intersection set of the SegmentStrings of the two geometries is calculated, generating topology nodes (NodeSections). Each topology node encapsulates an intersection point from a SegmentString and records its dimensional relationship (interior, boundary, exterior) in geometries A and B. While generating the topology intersection points, the algorithm also dynamically updates the corresponding cells of the IM matrix. IM pruning is incorporated here, potentially leading to calculation conclusions while constructing the topology structure.

[0090] (4) Evaluation of topological relationships

[0091] The topological relationship evaluation process includes: First, traversing all NodeSections to construct a topological graph model, containing topological nodes (RelateNode) and topological edges (RelateEdge): RelateNode represents the topological state of the neighborhood of an intersection point, and RelateEdge describes the topological position of the geometric boundary. By recursively analyzing the geometric positions on both sides of RelateEdge (e.g., the left side belongs to the Interior of A, and the right side belongs to the Exterior of B), the boundary-to-boundary dimensional interaction pattern is derived, and the IM is updated progressively. Finally, the judgment result is output through the DE-9IM pattern matching rule of predicate P.

[0092] Raster data, as an important type of spatial data, divides space into regular grids, with each grid cell storing specific attribute values, and is therefore widely used in fields such as remote sensing imagery and terrain analysis. The CKKS homomorphic encryption algorithm can encode multiple data points into a single vector and store them in the same ciphertext, efficiently handling the computation of repetitive data, which is highly compatible with the common computational patterns of raster data. Therefore, this invention employs the CKKS homomorphic encryption algorithm to achieve homomorphic encrypted computation of raster data.

[0093] At the implementation level, this invention uses the SEAL (SimpleEncryptedArithmeticLibrary) homomorphic encryption library to implement basic grid homomorphic encryption operators. SEAL is an open-source homomorphic encryption library developed by Microsoft Research, supporting mainstream homomorphic encryption schemes such as CKKS, and has wide applications in privacy-preserving machine learning and other fields. To improve computational efficiency, this invention also introduces Phantom, a CUDA-accelerated CKKS implementation framework. This framework optimizes polynomial operations through GPU parallelization (see below), and its API interface design maintains compatibility with SEAL. To uniformly support both SEAL and Phantom encryption backends, this invention designs a three-level architecture:

[0094] PlainText layer: Completes the conversion of raster data to CKKS plaintext field, and encodes floating-point arrays into polynomial plaintext;

[0095] CipherText layer: Converts CKKS plaintext into CKKS ciphertext using an encryption algorithm;

[0096] The HECrypto layer implements ciphertext operation operators (addition / multiplication / rotation) and key generation management, providing a unified abstract interface.

[0097] This architecture is adapted for both the SEAL (CPU) and Phantom (GPU) heterogeneous computing platforms. It maintains consistency in API calls across the key system, encryption / decryption, and homomorphic computation, allowing users to call the homomorphic acceleration libraries for both CPU and GPU using a consistent API. The overall architecture employs a layered abstraction design, ensuring compatibility with different encryption backends while reserving standard interfaces for future expansion with other homomorphic encryption libraries.

[0098] One of the core bottlenecks faced by homomorphic encryption technology in practical applications lies in its high computational complexity. To further optimize computational efficiency, this invention introduces a Single Instruction Multiple Data (SIMD) encoding mechanism, which significantly improves computational efficiency while maintaining approximate computational accuracy. This scheme is based on a circumspersible polynomial ring. (Where R represents the quotient ring, i.e., the polynomial ring) According to ideal ( X N The algebraic properties of the ring obtained by modulo operation (where N is a positive integer) are used to construct a complex vector space. Homomorphic mappings to ciphertext space can encode multiple complex numbers at once.

[0099] The core technological advantages of the SIMD encoding mechanism are reflected in two aspects:

[0100] Computational parallelism: Let the ciphertext space have dimension 1. N Then a single ciphertext unit can hold Each data component is independent, making a single homomorphic operation equivalent to... Order-of-magnitude acceleration of parallel computing;

[0101] Storage communication efficiency: It can reduce the amount of ciphertext that needs to be stored and serialized. This significantly improves storage and communication efficiency.

[0102] Ultimately, the process of raster homomorphic encryption computation is as follows: Figure 4 As shown. This embodiment of the invention can utilize SIMD encoding to encode multiple sampling points of a pixel matrix in parallel into a single plaintext element, then generate ciphertext through encryption operations, and finally integrate them into an encryption matrix (CipherMat). HECrypto serves as the unified interface for raster homomorphic encryption computation, responsible for performing raster addition, subtraction, multiplication, and other computational operations. This invention supports a total of 9 raster map algebraic functions, as shown in Table 4.

[0103] Table 4 Privacy Vector Constructor Table

[0104]

[0105] In this embodiment of the invention, for vector data, a complete computational framework is constructed based on TFHE homomorphic encryption technology. First, a unified security type system is designed, including basic types such as TFHEBool, TFHEInt64, and TFHEDecimal, and a scaling strategy is used to solve the floating-point calculation problem in an encrypted environment. Second, the TFHE-OGR spatial data structure is implemented. This structure follows the traditional OGR interface specification and supports encrypted domain operations, including the TFHEGeometry interface and its various derived classes such as points, lines, and surfaces. Finally, by constructing an encrypted nine-intersection model (DE-9IM) and a spatial relation calculation algorithm, efficient computation of spatial predicates is achieved. This algorithm greatly improves the efficiency of spatial relation calculation in an encrypted environment through multi-layered optimization strategies such as dimension filtering, bounding rectangle filtering, geometric feature optimization pruning, and predicate IM evaluation pruning.

[0106] For raster data, this invention employs the CKKS homomorphic encryption scheme. In its implementation, a three-tier architecture consisting of a PlainText layer, a CipherText layer, and a HECrypto layer is constructed, simultaneously supporting both SEAL and Phantom heterogeneous computing backends to ensure the framework's compatibility and scalability. Through the SIMD encoding mechanism, this invention achieves a parallel processing mode where a single ciphertext can accommodate multiple raster data points, significantly improving ciphertext space utilization and computational efficiency, while reducing computational and communication complexity while ensuring security.

[0107] Current spatial privacy computing solutions face a dual bottleneck: First, while homomorphic encryption computing can build a security guarantee for the entire data lifecycle, encryption operations lead to an exponential increase in computational overhead, resulting in superlinear time overhead in high-dimensional spatial computing scenarios; Second, existing operators are limited by monolithic architectures and have deficiencies in terms of computational scalability, making it difficult to meet the needs of processing massive spatial data.

[0108] To overcome the aforementioned bottlenecks, this invention proposes a multi-layered acceleration system, accelerating spatial privacy computing from two dimensions: distributed multi-node scaling and single-node heterogeneous acceleration. At the distributed computing level, a flexible horizontal scaling framework based on Apache Spark + Apache Sedona is constructed, achieving parallel processing of spatial privacy computing tasks through RDD data sharding and distributed task scheduling. At the single-node acceleration level, a GPU heterogeneous computing architecture is introduced to improve the computational efficiency of raster / vector core homomorphic encryption operators. By constructing a composite acceleration architecture that is both horizontally scalable and vertically accelerated, this invention elevates spatial computing performance to an engineering-practical level while ensuring data privacy and security.

[0109] To overcome the efficiency bottleneck of homomorphic encryption, this invention proposes a multi-layered acceleration system, accelerating spatial privacy computing from two dimensions: distributed multi-node scaling and single-node heterogeneous acceleration. At the distributed computing level, a flexible horizontal scaling framework based on Apache Spark + Apache Sedona is constructed, achieving parallel processing of spatial privacy computing tasks through RDD data sharding and distributed task scheduling. At the single-node acceleration level, a GPU heterogeneous computing architecture is introduced to improve the computational efficiency of raster / vector core homomorphic encryption operators. By constructing a composite acceleration architecture that is both horizontally scalable and vertically accelerated, this invention elevates spatial computing performance to a practical engineering level while ensuring data privacy and security.

[0110] The distributed spatial privacy computing framework proposed in this invention is as follows: Figure 4The framework uses Apache Spark as its distributed computing foundation, integrating the spatial data processing capabilities of the distributed spatial processing engine (Apache Sedona) with the spatial privacy computing capabilities of the spatial privacy computing operator (Spatial FHE). It constructs an architecture of "distributed computing + privacy protection + spatial function support," effectively solving the problem of insufficient single-machine performance in spatial privacy computing, which is unable to handle massive amounts of data. The specific technical implementation is as follows.

[0111] The distributed spatial privacy computing framework integrates spatial privacy computing capabilities through the Spatial FHE module. The homomorphic encryption interface (FHE Helper) component serves as the core interaction interface, enabling collaboration between Spark and Spatial FHE. Cross-language (C++→Java) adaptation technology is used to embed the aforementioned spatial privacy computing operators into the Spark distributed framework. The Spatial Homomorphic Encryption Management (SpatialFHEManager) includes matrix homomorphic encryption management (MatFHEManager) and vector homomorphic encryption management (VecFHEManager) submodules, which respectively integrate CPU raster ciphertext operator (SEALCrypto) / GPU raster ciphertext operator (PhantomCrypto) and vector ciphertext operator (TFHE_GEOS) to support homomorphic encryption computing of raster and vector data.

[0112] A distributed execution environment is built based on the core components (Driver, SparkEnv, Executor) of the distributed computing engine (Apache Spark). Data sharding is achieved through Resilient Distributed Datasets (RDDs), and parallel task scheduling is handled by the distributed core code (Sparkcore). The execution of distributed spatial privacy computing tasks adopts a master-slave model.

[0113] The client-side driver node runs at the computation initiator, decomposing the spatial computation task into multiple stages through the DAG Scheduler, generating an RDD operation chain containing spatial semantics. Simultaneously, the computation initiator has access to the public and private keys, enabling the creation of client-level FHE Helpers to handle information encryption, decryption, and computation.

[0114] The Executor nodes run on the computing server, integrating the Apache Sedona spatial computing engine to perform distributed spatial computations and utilizing the Block Manager to efficiently distribute spatial data across nodes. The computing server has access to the public key, therefore it can only create server-level FHE Helpers to handle homomorphic encryption computations of encrypted information.

[0115] Collaborative computing across multiple clusters can effectively improve the processing speed and data scale of spatial privacy computing, solving the privacy computing challenges of massive spatiotemporal data.

[0116] In this invention, two mode function systems are extended based on Apache Sedona:

[0117] Plaintext processing mode (native Sedona support): Provides raster functions (such as raster addition RS_Add) and geom functions (such as spatial intersection ST_Intersect), and combines RasterUDT and GeometryUDT to realize the serialization and parsing of native spatial data types;

[0118] Cipher processing mode (extended implementation of this invention): Design ciphertext raster functions (such as ciphertext raster addition RS_Add_Private) and ciphertext geometry functions (such as ciphertext spatial intersection ST_Intersects_Private). Encapsulate the ciphertext data operations of the TFHEGeometry and CipherMat classes through CipherMatUDT and the custom type CipherGeometryUDT, respectively implement the serialization / deserialization process of ciphertext raster and vector data in Spark SQL, and realize the encryption and decryption process transparent to the user.

[0119] Based on this, users can use standard Spark SQL syntax to manipulate encrypted data with the help of distributed clusters, which significantly reduces the technical threshold for privacy computing and expands the application scenarios of spatial privacy computing frameworks.

[0120] This invention employs a distributed homomorphic encryption key generation and sharing mechanism. Through the collaboration of a distributed file system and access control policies, it reliably distributes the public key within the computing cluster while ensuring the absolute security of the private key, thus establishing a foundation of trust for subsequent ciphertext computation.

[0121] Client initialization phase: The driver program loads the libSpatialFHE dynamic library through LocalFS (step 1), initializes the homomorphic encryption component, and generates public and private key pairs for two homomorphic encryption schemes, TFHE and CKKS, on the client (the initiator of homomorphic encryption) (step 2). The private key is stored in the local file system after being encrypted with AES-256 (step 3), and other nodes cannot access or read the private key.

[0122] Key distribution phase: The generated public key is further encrypted and compressed before being uploaded to the distributed file system (step 3). In this study, HDFS was used. The public key file in HDFS is protected by access control; only nodes participating in this round of computation have access to this public key. In common distributed jobs, the driver sends the storage location of the public key on HDFS and the corresponding ACL (Access Control List) authorization to the participating Executor processes.

[0123] Worker Node Runtime: Upon startup, the Executor process of a worker node first loads the libSpatialFHE dynamic library (step 5), then retrieves the public key from the specified path in HDFS based on the received public key's storage path (step 6), completing the distributed synchronization of the public key. After reading the public key, the homomorphic encryption component of the homomorphic encryption server is initialized within the process. After completing this series of initializations, the Executor can then perform calculations on various homomorphic encrypted ciphertexts in the task.

[0124] In vector space relation calculations, short-circuit optimization algorithms significantly improve computational efficiency by early termination condition checks. However, in encrypted computing environments, since branch checks must rely on plaintext Boolean values, traditional short-circuit mechanisms face a dilemma: directly decrypting intermediate Boolean values ​​for condition checks, while maintaining computational efficiency, can lead to the leakage of information about the computation mode or intermediate states, potentially compromising the security encapsulation of the encrypted system; completely avoiding decryption means failing to leverage the performance advantages of short-circuit optimization.

[0125] To address the aforementioned trade-off between privacy and efficiency, this invention proposes a remote Boolean decryption protocol based on Oblivious Transfer (OT). This protocol establishes a secure interaction mechanism between the client and server, achieving two core objectives in a distributed spatial privacy computing scenario: (1) allowing the server to selectively decrypt specific Boolean values ​​to support short-circuit computation; and (2) ensuring that the specific decrypted Boolean value is not leaked during network transmission through the privacy protection features of the OT protocol, while also preventing the client from obtaining the actual intermediate results. This dual-protection mechanism maintains the security boundaries of the encryption system while inheriting the computational efficiency advantages of short-circuit optimization. The specific steps are as follows:

[0126] 1. Initialization Phase

[0127] The initiator of the client-side homomorphic cryptographic computation holds the homomorphic encryption key pair (public key). pk Private key sk The server, acting as the computation party, only holds the public key. After generating TFHEBool type ciphertext that needs to be remotely decrypted during the computation process, the server performs the following operations:

[0128] Generate random bits ;

[0129] For ciphertext b Perform a homomorphic NOT operation to obtain the obfuscated ciphertext. ;

[0130] Constructing the obfuscated input array B ,satisfy .

[0131] 2. OT Protocol Interaction

[0132] The server will obfuscate the ciphertext array B Send to the client. The client uses the private key. sk Decryption B Obtain plaintext obfuscation results M In this process, regardless of whether it is truly... b The client decrypts the result into an obfuscated value {0,1}, thus protecting the privacy of intermediate results during the computation process.

[0133] 3. Result Validation and Output

[0134] The client obfuscates the plaintext result. M The data is sent to the server. The server then selects the target index for decryption via the OT protocol. The server obtains the plaintext result, but it does not receive the client's encrypted private key during this process. sk This achieves two-way privacy protection.

[0135] Regarding single-node computing performance improvement, this invention integrates two different GPU acceleration schemes to handle GPU acceleration for vector homomorphic encryption (TFHE) and raster homomorphic encryption (RFE) computing, respectively. TFHE actually encrypts input data bit by bit, and computation and bootstrapping on such encrypted data are very costly. To further improve the single-machine vector privacy computing performance, this invention introduces the tfhe-cudabackend module to implement a GPU acceleration scheme for vector homomorphic encryption computing. This scheme adopts a hybrid computing architecture design, in which non-intensive computing tasks such as key generation, ciphertext encryption and decryption are still executed on the host side, while the computationally intensive homomorphic operation module is ported to the CUDA architecture. This acceleration scheme significantly improves the parallel processing efficiency of fully homomorphic encryption computing. The CUDA-based TFHE accelerated computing process specifically includes the following four stages:

[0136] 1. Data preprocessing and encryption: The input data is standardized and encoded in the CPU computing engine and encrypted into LWE (Learning With Errors) ciphertext.

[0137] 2. Heterogeneous data transmission: The generated LWE ciphertext and the corresponding bootstrap key and GLWE ciphertext are transferred from host memory to GPU device memory.

[0138] 3. GPU-accelerated computing: Homomorphic encryption operations are executed in parallel under the CUDA architecture, with a focus on optimizing the parallel computing mode of PBS operations. The deep integration of instruction-level parallelism and data-level parallelism is achieved through the multi-stream processor architecture of the GPU.

[0139] 4. Result Decryption and Output: The LWE ciphertext processed by the GPU is sent back to the host memory, and the final decryption operation is performed on the CPU to output the plaintext result.

[0140] This invention addresses the dual performance bottlenecks of spatial privacy computing by proposing a multi-layered acceleration system that improves computational efficiency from two dimensions: multi-node distributed horizontal scaling and single-node heterogeneous computing acceleration. At the distributed level, a Spark+Sedona-based elastic scaling framework is constructed, achieving parallel processing and secure collaboration of spatial privacy computing tasks through RDD data sharding and distributed key sharing mechanisms. A remote Boolean decryption method based on the OT protocol is designed and implemented, supporting encrypted execution of short-circuit optimization strategies while maintaining the security boundaries of the encryption system. At the single-node acceleration level, GPU acceleration schemes are proposed for vector and raster data characteristics: GPU heterogeneous execution of vector homomorphic encryption core operations is achieved based on the CUDA architecture optimized for TFHE-rs; and the NWT transformation of the Phantom framework is fused with the CUDA kernel technology to reduce the polynomial multiplication complexity of raster data to... This significantly reduces memory and computational overhead. By constructing a composite acceleration architecture that is both horizontally scalable and vertically accelerated, this research elevates the performance of spatial privacy computing to an engineering-practical level, providing a systematic solution for the privacy and security processing of large-scale geospatial data.

[0141] Based on the same inventive concept, the present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the steps of the spatiotemporal big data distributed privacy computing system based on homomorphic encryption as described in any of the above embodiments.

[0142] Based on the same inventive concept, the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in the spatiotemporal big data distributed privacy computing system based on homomorphic encryption as described in any of the above embodiments.

[0143] Based on the same inventive concept, the present invention provides a computer program product, including a computer program / instruction, which, when executed by a processor, implements the steps in the spatiotemporal big data distributed privacy computing system based on homomorphic encryption as described in any of the above embodiments.

[0144] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0145] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, apparatus, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0146] This invention is described with reference to flowchart illustrations and / or block diagrams of the method, terminal device (apparatus), and computer program product according to the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable terminal device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable terminal device, generate instructions for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0147] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable terminal device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0148] These computer program instructions can also be loaded onto a computer or other programmable terminal device to cause a series of operational steps to be performed on the computer or other programmable terminal device to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable terminal device for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0149] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the invention.

[0150] Finally, it should be noted that in this invention, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, article, or terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, article, or terminal device that includes said element.

[0151] The above provides a detailed description of a spatiotemporal big data distributed privacy computing system based on homomorphic encryption provided by the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A spatiotemporal big data distributed privacy computing system based on homomorphic encryption, characterized in that, The system includes: Spatial privacy computation operators are used to achieve homomorphic encrypted computation of vector data and raster data by employing the TFHE homomorphic encryption algorithm and the CKKS homomorphic encryption algorithm, respectively, based on the computational characteristics of vector data and raster data. The system employs a distributed spatial privacy computing framework. Each node of this framework is equipped with a spatial privacy computing operator. The framework utilizes a distributed key management mechanism and a remote Boolean decryption protocol based on unintentional transmission to achieve multi-node task parallelization and secure collaboration. Each node in the framework is equipped with a GPU-accelerated architecture. This architecture features corresponding CUDA heterogeneous computing optimization schemes tailored to the Boolean logic operations of vector computing and the matrix operations of raster computing, thereby improving the processing efficiency of a single node.

2. The spatiotemporal big data distributed privacy computing system based on homomorphic encryption according to claim 1, characterized in that, Spatial privacy computation operators based on fully homomorphic encryption are used for: For vector data, plaintext data is encrypted using the TFHE type system. The TFHE type system provides an abstraction of encrypted data types for the implementation of the TFHEOGR class library, and the TFHE OGR class library provides an abstraction of encrypted geometric types for TFHE spatial predicate computation. By constructing an encrypted nine-intersection model and a spatial relation computation algorithm, the TFHE spatial predicate computation algorithm is realized. The TFHE type system includes the following basic types: TFHEBool, TFHEInt64, Int32, and TFHEDecimal.

3. The spatiotemporal big data distributed privacy computing system based on homomorphic encryption according to claim 2, characterized in that, Encrypting plaintext data using the TFHE type system includes: The type registration framework based on the TFHERegisteredType generic interface provides a dynamic type registration mechanism, enabling encrypted data types within a process to use a unified encryption context; it manages the context of homomorphic encryption based on a process-level globally unique TFHEContext singleton structure; and it constructs a type system based on TFHE encryption by encapsulating basic FHE ciphertext pointers. Among them, for spatial data, the floating-point encryption method based on scaling factor converts floating-point operations into integer field operations by scaling.

4. The spatiotemporal big data distributed privacy computing system based on homomorphic encryption according to claim 2, characterized in that, The TFHE type system provides an abstraction for cryptographic data types for the implementation of the TFHE OGR class library, including: The TFHEGeometry interface class, corresponding to the functionality of the Geometry class, defines the construction and calculation operations of geometric objects in the encrypted domain. Subclasses of the TFHEGeometry interface class implement the specific expression of spatial data through polymorphism: TFHEPoint encapsulates the coordinates of encrypted single points, TFHELineString manages the ordered encrypted point sequences to form line elements, and the TFHEPolygon class describes the spatial morphology of planar elements through a set of encrypted linear rings. Among them, TFHELinearRing, as a subclass of TFHELineString, additionally implements the ring closure verification method. For coordinate data, the TFHECoordinate class is used to uniformly manage encrypted coordinate data. Its internal coordinate values ​​are stored and calculated using the TFHEDecimal class in the TFHE type system, which is based on TFHE and ensures that all spatial operations are performed in encrypted state.

5. The spatiotemporal big data distributed privacy computing system based on homomorphic encryption according to claim 2, characterized in that, The TFHE OGR library provides an abstraction of encrypted geometry types for TFHE spatial predicate computation. It implements the TFHE spatial predicate computation algorithm by constructing an encrypted nine-intersection model and a spatial relation computation algorithm, including: Predicate initialization is based on dimensional filtering and bounding rectangle filtering; The computation of homomorphic encryption space is reduced by using a variety of geometric feature-based pruning methods. These methods include point-to-point optimization, point-to-non-point optimization, line-to-polygon endpoint detection, and polygon-to-polygon vertex pre-detection. Determine whether the bounding rectangles of two cryptographic geometric objects involved in the operation intersect; In the case of intersection, line features are decomposed into line segment geometry, and polygon features are decomposed into rings, marking whether it is an inner ring or an outer ring and the polygon number to which it belongs; calculate the SegmentStrings intersection set of the two encrypted geometry objects to generate topology nodes; each topology node encapsulates an intersection point in a SegmentString and records its dimensional relationship in the two encrypted geometry objects. Traverse all topological nodes to construct a topological graph model, which includes RelateNode and RelateEdge. RelateNode represents the topological state of the neighborhood of an intersection point, and RelateEdge describes the topological position of the geometric boundary. By recursively analyzing the geometric positions on both sides of RelateEdge, the boundary-to-boundary dimensional interaction pattern is derived, and the IM is updated progressively. The judgment result is output by the DE-9IM pattern matching rule of predicate P.

6. The spatiotemporal big data distributed privacy computing system based on homomorphic encryption according to claim 1, characterized in that, For raster data, a basic raster homomorphic encryption operator is implemented based on the SEAL homomorphic encryption library. This basic raster homomorphic encryption operator includes: PlainText layer: Completes the conversion of raster data to CKKS plaintext field, and encodes floating-point arrays into polynomial plaintext; CipherText layer: Converts CKKS plaintext into CKKS ciphertext using an encryption algorithm; HECrypto layer: Implements ciphertext operation operators and key generation management, providing a unified abstract interface.

7. The spatiotemporal big data distributed privacy computing system based on homomorphic encryption according to claim 1, characterized in that, The distributed spatial privacy computing framework integrates spatial privacy computing capabilities through spatial privacy computing operators. The FHEHelper component serves as the core interaction interface, enabling collaboration between Spark and SpatialFHE. Cross-language adaptation technology is used to embed the spatial privacy computing operators into the Spark distributed framework. SpatialFHEManager has two sub-modules, MatFHEManager and VecFHEManager, which integrate SEALCrypto / PhantomCrypto and TFHE_GEOS respectively to support homomorphic encrypted computing of raster and vector data.

8. The spatiotemporal big data distributed privacy computing system based on homomorphic encryption according to claim 1, characterized in that, At the distributed level, the distributed spatial privacy computing framework is based on the elastic scaling framework of Spark+Sedona. Through RDD data sharding and distributed key sharing mechanism, it realizes parallel processing and secure collaboration of spatial privacy computing tasks. It designs and implements a remote Boolean decryption method based on the OT protocol, which supports the encrypted execution of short-circuit optimization strategy while maintaining the security boundary of the encryption system. At the single-node acceleration level, GPU acceleration solutions are set up for the characteristics of vector and raster data respectively: the CUDA architecture based on TFHE-rs is optimized to realize the heterogeneous execution of the vector homomorphic encryption core operation on GPU; the NWT transformation of the Phantom framework and the CUDA kernel fusion technology are used to reduce the polynomial multiplication complexity of raster data and reduce memory and computational overhead.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the spatiotemporal big data distributed privacy computing system based on homomorphic encryption as described in any one of claims 1-8.

10. A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the spatiotemporal big data distributed privacy computing system based on homomorphic encryption as described in any one of claims 1-8.

Images