A vector high-precision map nonlinear encryption method and system

Abstract

The application discloses a kind of vector high-precision map nonlinear encryption method and system, belong to high-precision electronic map making field, comprising: obtaining longitude and latitude coordinates from vector high-precision map;The longitude and latitude coordinates are input into deep neural network to be encrypted once, and the longitude and latitude coordinates after once encryption are obtained;Wherein, the deep neural network includes: input layer, for input longitude and latitude coordinates;Hidden layer, for nonlinear encryption is carried out by network depth and activation function;Output layer, for output encrypted longitude and latitude coordinates;Based on the longitude and latitude coordinates after once encryption, using piecewise bilinear piola transformation carries out secondary encryption, and the longitude and latitude coordinates after final encryption are obtained.The application uses deep network model to replace trigonometric function model to calculate longitude and latitude variation, improves the degree of freedom and intensity of encryption, improves encryption intensity, real-time performance, flexibility, scalability by deep neural network and piecewise bilinear piola transformation.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of high-precision electronic map making, and particularly relates to a vector high-precision map nonlinear encryption method and system. BACKGROUND

[0002] High-precision maps have important applications in the fields of automatic driving, intelligent transportation, geographic information, etc. High-precision map security processing technology is a key technology for ensuring the security of classified geographic information at the present stage. The currently recognized topographic map security processing technology has greatly promoted the healthy and orderly development of the navigation electronic map industry since it was put into use in 2003, and has been applied in digital cities and geographic information public service platforms, etc., achieving good social and economic benefits. With the rapid development of surveying and mapping geographic information technology in recent years, new industry patterns, new products, and new applications of geographic information have put forward higher demands for geographic information security processing technology. High-precision map encryption uses the original geographic coordinates after offset or conversion, aiming to protect the security of geographic information and prevent unauthorized use and leakage. The independently developed effective encryption algorithm can significantly enhance data protection and achieve precise control of data sovereignty.

[0003] Traditional encryption algorithms do not consider the spatial correlation of map data, resulting in problems such as geometric shape distortion of roads after encryption (e.g., unreasonable bending of straight road segments after encryption), and distortion of the distance between adjacent roads, affecting the positioning and matching accuracy of the vehicle end. In addition, highly nonlinear encryption has strong security, but the computational complexity is high, putting pressure on the limited computing power of the vehicle end. SUMMARY

[0004] The purpose of the present application is to address the problem that traditional encryption algorithms do not consider the spatial correlation of map data, resulting in geometric shape distortion of roads after encryption (e.g., unreasonable bending of straight road segments after encryption), and distortion of the distance between adjacent roads. The present application provides a vector high-precision map nonlinear encryption method that uses deep neural networks and piecewise bilinear Piola transformation for encryption, improving encryption strength, real-time performance, flexibility, and scalability. Additionally, user-defined encryption model parameters make it easier to implement decryption.

[0005] According to an aspect of the present application, a vector high-precision map nonlinear encryption method is provided, comprising:

[0006] Obtaining latitude and longitude coordinates from a vector high-precision map;

[0007] The latitude and longitude coordinates are input into a deep neural network (DNN) for one-time encryption to obtain one-time encrypted latitude and longitude coordinates; wherein the deep neural network comprises: an input layer for inputting latitude and longitude coordinates; a hidden layer for nonlinear encryption through network depth and an activation function; and an output layer for outputting one-time encrypted latitude and longitude coordinates.

[0008] Based on the one-time encrypted latitude and longitude coordinates, a piecewise bilinear Piola transformation is used for two-time encryption to obtain final encrypted latitude and longitude coordinates.

[0009] Further, inputting latitude and longitude coordinates into a deep neural network for one-time encryption comprises:

[0010] The latitude and longitude coordinates are taken as an initial feature space;

[0011] Based on the initial feature space, a specified rotation matrix is used for rotation transformation to obtain a rotated feature space; then a set translation vector is used to perform linear displacement superposition on the rotated feature space to obtain a translated feature space;

[0012] The translated feature space is input into an activation function to obtain a nonlinear transformed feature space, i.e., a latitude and longitude coordinate change.

[0013] Further, the one-time encryption comprises:

[0014]

[0015] wherein, is the input layer, is a two-dimensional vector of latitude and longitude coordinates, is the output of the previous layer, is an activation function corresponding to the hidden layer, represents the hidden layer, represents a rotation matrix, represents a translation vector, represents an activation function; represents a latitude and longitude coordinate change, represents network depth.

[0016] Further, the method further comprises:

[0017] The latitude and longitude coordinate change is converted to obtain one-time encrypted latitude and longitude coordinates, expressed as:

[0018]

[0019]

[0020] wherein, is the encrypted latitude coordinate, c is the coordinate scale conversion coefficient m / °, used to convert the latitude change from radian to meter unit, is the ellipsoid long radius, and e is the ellipsoid eccentricity.

[0021] Further, the twice encryption is performed by using the sliced bilinear Pierrot transformation, including:

[0022] The encrypted latitude coordinate is matched to the preset rectangular unit K, and the center coordinates of the rectangular unit K are set as ;

[0023] Each rectangular unit K corresponds to a perturbed four-side unit , with the vertex coordinates , and the sliced bilinear Pierrot transformation expression is obtained as:

[0024]

[0025] wherein, is the final encrypted latitude coordinate, the vertex coordinates and the node basis function are one-to-one corresponding, satisfying .

[0026] Further, the method further includes:

[0027] The rectangular frame E is calculated based on the latitude and longitude range of the map, and the rectangular frame E is divided into a rectangular grid T;

[0028] Any one rectangular unit K is obtained based on the rectangular grid T.

[0029] According to an aspect of the present specification, a vector high-precision map nonlinear encryption system is provided, including:

[0030] A data acquisition module is configured to acquire latitude and longitude coordinates from a vector high-precision map;

[0031] A primary encryption module is configured to input the latitude and longitude coordinates into a deep neural network for primary encryption to obtain a primary encrypted latitude and longitude coordinate; wherein the deep neural network includes: an input layer configured to input the latitude and longitude coordinates; a hidden layer configured to perform nonlinear encryption through network depth and an activation function; and an output layer configured to output the primary encrypted latitude and longitude coordinate;

[0032] A twice encryption module is configured to perform twice encryption by using a sliced bilinear Pierrot transformation based on the encrypted latitude and longitude coordinate to obtain a final encrypted latitude and longitude coordinate.

[0033] According to an aspect of the present specification, there is provided an electronic device including a memory storing a computer program and a processor implementing the steps of the vector high-precision map nonlinear encryption method when executing the computer program.

[0034] According to an aspect of the present specification, there is provided a computer-readable storage medium having a computer program stored thereon, the computer program implementing the steps of the vector high-precision map nonlinear encryption method when executed by a processor.

[0035] According to an aspect of the present specification, there is provided a computer program product containing instructions which, when executed on a computer, cause the computer to perform the steps of the vector high-precision map nonlinear encryption method.

[0036] Compared with the prior art, the present application has the following beneficial effects:

[0037] 1. The embodiment of the present application uses a deep network model to replace the calculation of latitude and longitude changes of the trigonometric function model, thereby improving the encryption degree and strength.

[0038] 2. The embodiment of the present application uses a piecewise bilinear Piola transformation for encryption, avoiding the overall deduction function for local offset inversion algorithm.

[0039] 3. The embodiment of the present application uses a deep neural network and a piecewise bilinear Piola transformation for encryption, thereby improving the encryption strength, real-time performance, flexibility, and scalability, and through user-defined encryption model parameters, it is easier to implement in the decryption process. BRIEF DESCRIPTION OF DRAWINGS

[0040] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiment or prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of these drawings.

[0041] Figure 1 The method flowchart of the embodiment of the present application;

[0042] Figure 2 The rectangular grid / quadrilateral grid diagram of the embodiment of the present application;

[0043] Figure 3 The piecewise bilinear Piola transformation encryption diagram of the embodiment of the present application. DETAILED DESCRIPTION

[0044] 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.

[0045] like Figure 1 As shown, this embodiment of the invention provides a nonlinear encryption method for high-precision vector maps, comprising: obtaining latitude and longitude coordinates from a high-precision vector map; inputting the latitude and longitude coordinates into a deep neural network for primary encryption to obtain primary encrypted latitude and longitude coordinates; wherein the deep neural network comprises: an input layer for inputting latitude and longitude coordinates; a hidden layer for nonlinear encryption using network depth and activation function; an output layer for outputting the primary encrypted latitude and longitude coordinates; and, based on the primary encrypted latitude and longitude coordinates, performing secondary encryption using piecewise bilinear Piola transform to obtain the final encrypted latitude and longitude coordinates.

[0046] Specifically, the input layer of the deep network model takes the longitude and latitude of WGS84 coordinates as input, the output layer takes the coordinate changes as output, and the hidden layer is the core of non-linear encryption. Its functionality depends on two configurable parameters: network depth and activation function sequence. Users can freely select these parameters in the following process according to their needs. The output layer of the deep network model undergoes post-processing to obtain the encrypted longitude and latitude of the coordinates.

[0047] Specifically, the longitude and latitude of the WGS84 coordinates are initialized as follows:

[0048] , (1)

[0049] in, The longitude is in WGS84 coordinates. The dimension of the WGS84 coordinate system.

[0050] Specifically, the user selects the network depth (i.e., the number of hidden layers). Theoretically, users can choose deeper networks to increase the nonlinearity, but this also means an increase in computational cost. Each hidden layer consists of linear transformations and nonlinear activations, where linear transformations refer to rotations and translations. For each layer... Randomly generate the corresponding rotation angle and two-dimensional translation vector ,in The range of values ​​is , The value range of each component is [-100, 100]. For example, for the first layer, the following values can be selected

[0051] , (2)

[0052] The weight matrix (i.e., the rotation matrix) is defined as:

[0053] (3)

[0054] Then, encryption is performed, and the expression is as follows:

[0055] (4)

[0056] wherein, is the input layer, i.e., a two-dimensional vector composed of the calculated initial coordinates. is the output of the previous layer, is the activation function corresponding to the hidden layer, represents the hidden layer , the output of the previous layer is taken as the input of the layer (the input of the first hidden layer is the input layer), and then, after linear transformation of the rotation matrix and the translation vector , the output result of the layer is obtained through the activation function . is a two-dimensional vector of longitude and latitude coordinates, and after the last hidden layer, the output longitude and latitude coordinate change , represents the network depth.

[0057] Specifically, the present application also provides an embodiment one, if the selected network depth is , the activation function sequence can be selected as follows:

[0058] (5)

[0059] wherein, the expression is:

[0060] (6)

[0061] if the selected network depth is , the activation function sequence can be selected as follows:

[0062] (7)

[0063] Because tanh restricts the value domain to (-1, 1), the local approximate rigid body property can be maintained.

[0064] Specifically, the embodiment of the present application also provides the embodiment two,

[0065] (8)

[0066] For the first layer , the randomly generated , , the linear transformation is:

[0067] (9)

[0068] After the activation function (f(x)) is activated, it becomes:

[0069] (10)

[0070] The output is:

[0071] (11)

[0072] Then the second layer, the third layer and the like are calculated in turn.

[0073] Specifically, the coordinates obtained by post-processing are longitude and latitude:

[0074] (12)

[0075] (13)

[0076] Wherein , c is the coordinate scale conversion coefficient (dimension m / °), which is used to convert the longitude and latitude change from radian to meter unit, is the long radius of the ellipsoid, and e is the eccentricity of the ellipsoid.

[0077] Specifically, the embodiment of the present application provides a piecewise bilinear piecewise bilinear Piola transformation encryption. The input is the result of the deep neural network encryption in the previous step And , the output is the final encrypted longitude and latitude coordinates. In this step, the data is encrypted for the second time, improving the security of the algorithm, wherein the map point is framed by the map longitude and latitude range. The piecewise bilinear Piola transformation realizes quadrilateral grid mapping through bilinear basis function, which belongs to the isoparametric element transformation in the finite element method.

[0078] Specifically, the rectangular frame of the map longitude and latitude range is defined as E, and the set of map points is . E is divided into a rectangular grid T, which contains g vertices, saves the coordinate matrix G (g x 2, g is the total number of grid vertices), and performs global-local double-layer numbering on the grid vertices and elements. The user inputs the perturbation matrix of the grid vertices​ (g x 2), a set of quadrilateral grids can be obtained The vertex coordinate matrix expression is as follows:

[0079] (14)

[0080] The quadrilateral after adding the disturbance is generally not rectangular, and the quadrilateral element after the disturbance breaks the global regularity, as shown in the following formula: Figure 2

[0081] Specifically, the embodiment of the present application also provides the steps of the unit cycle, assuming that the quadrilateral element is K, and the quadrilateral element after adding the disturbance is , wherein the four vertices of K are A1, A2, A3, and A4, , the four vertices of are B1, B2, B3, and B4, as shown in the following formula: Figure 3 The forward transformation is performed to obtain the following expression:

[0082] (15)

[0083] Specifically, the embodiment of the present application also provides the steps of screening the grid, calculating the latitude and longitude range of the input data, and judging which rectangular grid elements can cover the range. At the same time, these grid elements are recorded as the element set L={L1, L2,..., Lm} (wherein L1, L2,..., Lm are subsets of the rectangular grid T, and if the range of the input data is small, the Piola transformation of the data on the element set L is more time-saving than the transformation on the element set T). The map data points are judged point by point The element K belongs to L (if exactly on the grid vertex or the common edge, the element is uniformly attributed to the quadrilateral element with the smallest global number), and the expression is calculated as follows:

[0084] (16)

[0085] The encrypted map point set is saved .

[0086] Specifically, the embodiment of the present application provides The forward transformation is performed, assuming that the center coordinates of the quadrilateral element K are , as shown in the following formula: Figure 2 , the length and the width are 2h and 2w respectively, and the expression of the node basis function is as follows:

[0087] (17)

[0088] , wherein is the point to be encrypted, ​​​satisfy . Suppose that the coordinates of the vertices of the quadrilateral element are , then The positive formula of the transformation is:

[0089] (18)

[0090] where, are the encrypted coordinates, and the vertex coordinates correspond to the node basis function one by one.

[0091] Specifically, according to the three principles of "high nonlinearity, approximate rigidity, and relative irreversibility" of high-precision map encryption, the embodiment of the application provides a high-precision map encryption method that is efficient and easy to implement under different parameter systems, so as to improve the security protection level of geographic information data. The "approximate rigidity" solves the distortion and distortion problem, and the high efficiency and ease of implementation guarantee the real-time application of the vehicle end. Here, "approximate rigidity" refers to the combination of multi-layer linear transformation (rotation + translation) and nonlinear activation function, and the local area transformation is close to the characteristics of rigid transformation.

[0092] Specifically, the embodiment of the application also provides an explanation of relative irreversibility. First, explain reversibility, and then explain relative irreversibility. Among them, the reversibility ensures that two geographically different points do not coincide in the encrypted coordinate system. The specific content is as follows:

[0093] 1. The mathematical reversibility of deep neural network encryption. F(Lon, Lat) represents the encrypted longitude and latitude data obtained by DNN transformation. The way to construct the deep neural network is to couple a specific activation function (LeakyReLU, arctan, and tanh, etc.) through rigid transformation (rotation + translation) for each layer. It can be proved that The norm is bounded and less than 1, so the above transformation is a monomorphism, thereby ensuring that the encryption method is a monomorphism and reversible in mathematics.

[0094] 2. The mathematical reversibility of the piecewise bilinear transformation encryption. This point can be obtained from the geometric properties of bilinear transformation. Note that given the mapping coordinates of the four vertices of the quadrilateral parallel to the coordinate system, the bilinear transformation within the rectangular region is restricted to a straight line parallel to the coordinate system, which degenerates into a linear transformation, thereby ensuring the uniqueness of the transformation in the rectangular region. Finally, using the mapping relationship between the vertices, plus the monotonicity along any coordinate in each rectangular region, it can be concluded that the piecewise bilinear transformation is mathematically reversible.

[0095] 3. Relatively irreversible. The non-rigid characteristics of local DNN encryption can lead to the inaccuracy of parameter inversion; the piecewise bilinear encryption can cause large deviation of the approximate rigid parameters in different regions. When the encryption process is known, DNN inversion needs to solve a nonlinear equation set (the complexity increases exponentially with D), and Piola inversion needs global optimization (the time complexity is too high), and in actual calculation, the approximate irreversibility is caused by floating point error.

[0096] Specifically, the encryption process is: first encrypted by DNN, and then encrypted by piecewise bilinear transformation. At this time, in order to accurately obtain the inverse process of the encryption process, the following parameters need to be estimated: the architecture of DNN (network width, depth, activation function, network parameter), the control point position and change of piecewise bilinear transformation. It is noted that the architecture of DNN has a very large degree of freedom, although the last layer is close to a rigid transformation and can be more accurately estimated, but if accurate estimation is desired, as the estimation accuracy increases, the complexity of inverse calculation also increases exponentially (at this time, the network architecture needs to be estimated very accurately). And the piecewise bilinear transformation seems to be regular, but in order to get an accurate estimation formula, the position of each vertex and the transformed coordinates need to be estimated. At this time, the calculation complexity increases with the increase of the number of vertices, especially the parameters brought by each vertex position and change are 4 times the number of vertices. Therefore, even if the encryption process is known, the system shows strong security due to the double restrictions of floating point error accumulation and calculation complexity explosion.

[0097] The implementation basis of each embodiment of the present application is realized by the programmed processing of a device with processor function. Therefore, in engineering practice, the technical solutions and functions of each embodiment of the present application are packaged into various modules. Based on this actual situation, on the basis of the above embodiments, the embodiments of the present application provide a vector high-precision map nonlinear encryption system, which is used to execute a vector high-precision map nonlinear encryption method in the above method embodiments.

[0098] The system comprises a data acquisition module configured to acquire latitude and longitude coordinates from a vector high-precision map; a primary encryption module configured to input the latitude and longitude coordinates into a deep neural network for primary encryption to obtain primary encrypted latitude and longitude coordinates; wherein the deep neural network comprises an input layer configured to input the latitude and longitude coordinates; a hidden layer configured to perform nonlinear encryption through network depth and an activation function; and an output layer configured to output the primary encrypted latitude and longitude coordinates; and a secondary encryption module configured to perform secondary encryption on the primary encrypted latitude and longitude coordinates based on a piecewise bilinear Piola transformation to obtain final encrypted latitude and longitude coordinates.

[0099] The vector high-precision map nonlinear encryption system provided by the embodiment of the application, in view of the fact that the traditional encryption algorithm does not consider the spatial correlation of map data, leading to the problems of road geometric shape distortion (such as unreasonable bending after encryption of a straight road segment) and adjacent road distance distortion after encryption, adopts a plurality of modules to perform encryption through a deep neural network and a piecewise bilinear Piola transformation, and improves the encryption strength, real-time performance, flexibility, and scalability, and through user-defined encryption model parameters, the decryption process is more easily implemented.

[0100] Based on the same inventive concept as the foregoing embodiments, the embodiment of the application also provides an electronic device including a memory and a processor, the memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions to implement the vector high-precision map nonlinear encryption method as proposed in the foregoing embodiments.

[0101] The embodiment of the application also provides a computer readable storage medium having a computer program stored thereon. When the program is executed by a processor, the problems of road geometric shape distortion (such as unreasonable bending after encryption of a straight road segment) and adjacent road distance distortion after encryption caused by the fact that the traditional encryption algorithm does not consider the spatial correlation of map data are overcome, the encryption strength, real-time performance, flexibility, and scalability are improved, and through user-defined encryption model parameters, the decryption process is more easily implemented.

[0102] The storage medium can be a hard disk, a solid state disk, a flash disk, an optical disk, or any non-volatile storage device, used to store computer program codes and necessary data files, and the stored computer program includes a data acquisition module, a primary encryption module, and a secondary encryption module.

[0103] The embodiment of the application also provides a computer program product containing instructions, which, when running on a computer, generates the vector high-precision map nonlinear encryption method proposed in the foregoing embodiments in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices.

[0104] In summary of the inventive embodiments, the input latitude and longitude coordinates are taken as an initial feature space (two-dimensional vector), in each hidden layer, a user-specified rotation matrix is first used for rotation transformation to obtain a rotated feature space, and then a user-specified translation vector is used for linear displacement superposition to further obtain a translated feature space. Finally, the feature space after rotation and translation geometric transformation is input to an activation function to obtain a nonlinear transformed feature space. After the hidden layer, the feature space is no longer the latitude and longitude coordinates, but the change amount of the latitude and longitude coordinates.

[0105] Finally, it should be pointed out that the above specific embodiments are only representative examples of the present application. Obviously, the present application is not limited to the above specific embodiments, and there can be many variations. Any simple modification, equivalent change and modification made to the above specific embodiments according to the technical essence of the present application should be considered to be within the protection scope of the present application.

Claims

1. A method for non-linear encryption of a vector high-precision map, characterized in that, The method comprises the following steps: Obtaining latitude and longitude coordinates from a vector high-precision map; Inputting the latitude and longitude coordinates into a deep neural network for one-time encryption to obtain one-time encrypted latitude and longitude coordinates; wherein the deep neural network comprises: an input layer for inputting latitude and longitude coordinates; a hidden layer for nonlinear encryption through network depth and an activation function; and an output layer for outputting one-time encrypted latitude and longitude coordinates; Based on the one-time encrypted latitude and longitude coordinates, performing two-time encryption using a piecewise bilinear Piole transformation to obtain final encrypted latitude and longitude coordinates.

2. The method of claim 1, wherein the vector high-precision map is encrypted by using a non-linear encryption algorithm. Inputting the latitude and longitude coordinates into a deep neural network for one-time encryption comprises: Taking the latitude and longitude coordinates as an initial feature space; Based on the initial feature space, performing rotation transformation using a specified rotation matrix to obtain a rotated feature space; then performing linear displacement superposition on the rotated feature space using a set translation vector to obtain a translated feature space; Inputting the translated feature space into an activation function to obtain a nonlinearly transformed feature space, i.e., latitude and longitude coordinate changes.

3. The method of claim 2, wherein the vector high-precision map is encrypted by using a non-linear encryption algorithm. The one-time encryption comprises: ， wherein, is an input layer, is a two-dimensional vector of latitude and longitude coordinates, is an output of a previous layer, is an activation function corresponding to a hidden layer, denotes a hidden layer, denotes a rotation matrix, denotes a translation vector, denotes an activation function; denotes a change in latitude and longitude coordinates, denotes a network depth.

4. The method of claim 2, wherein the vector high-precision map is encrypted by using a non-linear encryption algorithm. The method further comprises: Converting the latitude and longitude coordinate changes to obtain one-time encrypted latitude and longitude coordinates, expressed as: , , wherein, is the encrypted latitude coordinate, , c is a coordinate scale conversion coefficient m / °, used to convert the latitude change from radian to meter unit, is the ellipsoid long radius, and e is the ellipsoid eccentricity.

5. The method of claim 1, wherein the vector high-precision map is non-linearly encrypted by using a non-linear function. Performing two-time encryption using a piecewise bilinear Piole transformation comprises: Match the encrypted latitude and longitude coordinates to the preset rectangular unit K, and set the center coordinates of the rectangular unit K ; Each rectangular element K corresponds to one , The vertex coordinates of the rectangular element K are The expression of the piecewise bilinear Pioola transformation is obtained as , wherein, are the final encrypted latitude and longitude coordinates, vertex coordinates with node basis functions one-to-one correspondence, satisfies .

6. The method of claim 5, wherein the vector high-precision map is encrypted by using a non-linear encryption algorithm. The method further comprises: Based on the latitude and longitude range of the map, calculating a rectangular frame E and dividing the rectangular frame E into a rectangular grid T; Based on the rectangular grid T, obtaining any one rectangular element K.

7. A system for non-linear encryption of a vector high-precision map, characterized in that, The method comprises the following steps: A data acquisition module for obtaining latitude and longitude coordinates from a vector high-precision map; A one-time encryption module for inputting the latitude and longitude coordinates into a deep neural network for one-time encryption to obtain one-time encrypted latitude and longitude coordinates; wherein the deep neural network comprises: an input layer for inputting latitude and longitude coordinates; a hidden layer for nonlinear encryption through network depth and an activation function; and an output layer for outputting one-time encrypted latitude and longitude coordinates; A two-time encryption module for performing two-time encryption based on the one-time encrypted latitude and longitude coordinates using a piecewise bilinear Piole transformation to obtain final encrypted latitude and longitude coordinates. 8.An electronic device comprising a memory and a processor, the memory storing a computer program, wherein, The processor executes the computer program to implement the steps of the vector high-precision map nonlinear encryption method of any one of claims 1-6.

9. A computer readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the vector high-precision map nonlinear encryption method of any one of claims 1-6.

10. A computer program product comprising instructions, characterized in that, When it runs on a computer, it makes the computer execute the steps of the vector high-precision map nonlinear encryption method of any one of claims 1-6.

Images