Method and System for Evaluating Performance of a Location Privacy Protection Algorithm in Internet of Vehicles

Abstract

The present invention discloses a method and system for evaluating the performance of a vehicle networking location privacy protection algorithm. The method includes: processing the real location by using the location privacy protection algorithm to be evaluated to obtain an anonymous location; calculating the quantization value between the real location and the anonymous location and the averaged value of the anonymous parameters; the quantization value includes the fusion distance, the Jaccard similarity, and the Hellinger divergence; based on the real location, the anonymous location, the quantization value, and the averaged value of the anonymous parameters, calculating the weight value by using the information gain method and the entropy weight method; the weight value includes the weight of the quantization value and the weight of the averaged value of the anonymous parameters; calculating the privacy protection degree evaluation value and the location service quality evaluation value of the location privacy protection algorithm to be evaluated based on the quantization value, the averaged value of the anonymous parameters, and the weight value. The present invention can evaluate a variety of vehicle networking location privacy protection algorithms and has a wide adaptation range.

Description

Technical Field

[0001] The present invention relates to the field of Internet of Vehicles (IoV) security, and particularly to a method and system for evaluating the performance of an Internet of Vehicles location privacy protection algorithm. Background Art

[0002] The Internet of Vehicles represents the evolution of vehicle ad hoc networks supported by cloud computing towards the Internet of Things paradigm, which is jointly composed of vehicles with sensing and communication capabilities driving on the road, roadside communication units, and back-end servers. Location Based Services (LBS) is a value-added service that obtains the location information of mobile terminal users through the radio communication network of telecom mobile operators or external positioning methods such as GPS, and provides corresponding services for users with the support of a geographic information system platform. LBS usually requires in-vehicle users to report location information of continuous road segments. When the vehicle is in an open physical space, the leakage of privacy such as the driver's identity information, license plate number, location, and itinerary may pose a threat to the life and property safety of the driver and passengers. The diversity of the combination of evaluation subjects, evaluation objects, and evaluation scales determines the non-uniqueness of the evaluation index system. The existing methods for quantifying the performance evaluation indexes of location privacy protection algorithms mainly perform security situation grading and analyze from the perspectives of data accuracy, integrity, availability, energy consumption, etc. Most of these methods are only applicable to specific applications, attack models (such as replay attacks, message modification and generation attacks, denial attacks, simulation attacks, location tracking attacks, etc.), and privacy threats, and are difficult to be generalized. They lack universal applicability in the evaluation of classical algorithms and their improved algorithms, and cannot evaluate multiple Internet of Vehicles location privacy protection algorithms to implement location privacy protection using the optimal Internet of Vehicles location privacy protection algorithm. Summary of the Invention

[0003] Based on this, it is necessary to provide a method and system for evaluating the performance of an Internet of Vehicles location privacy protection algorithm, which can evaluate multiple Internet of Vehicles location privacy protection algorithms and has a wide range of adaptability, thereby improving the security of Internet of Vehicles location privacy.

[0004] To achieve the above object, the present invention provides the following solution:

[0005] A method for evaluating the performance of an Internet of Vehicles location privacy protection algorithm includes:

[0006] Obtaining the real location of a user entity;

[0007] Processing the real location using the location privacy protection algorithm to be evaluated to obtain an anonymous location;

[0008] Calculate the quantization value between the real location and the anonymous location and the averaged value of the anonymous parameters; the quantization value includes the fusion distance, Jaccard similarity, and Hellinger divergence;

[0009] Based on the real location, the anonymous location, the quantization value, and the averaged value of the anonymous parameters, calculate the weight values using the information gain method and the entropy weight method; the weight values include the weight of the quantization value and the weight of the averaged value of the anonymous parameters;

[0010] Based on the quantization value, the averaged value of the anonymous parameters, and the weight values, calculate the evaluation result of the location privacy protection algorithm to be evaluated; the evaluation result includes the privacy protection degree evaluation value and the location service quality evaluation value.

[0011] Optionally, the obtaining of the real location of the user entity specifically includes:

[0012] Obtain the original location data of the user entity;

[0013] Perform data cleaning on the original location data, and determine the two-dimensional location coordinates in the original location data as the real location of the user entity.

[0014] Optionally, the calculating of the quantization value between the real location and the anonymous location and the averaged value of the anonymous parameters specifically includes:

[0015] Calculate the Mahalanobis distance and the Euclidean distance between the real location and the anonymous location;

[0016] Calculate the correlation coefficient matrix of the anonymous location;

[0017] Determine the fusion distance between the real location and the anonymous location from the Mahalanobis distance, the Euclidean distance, and the correlation coefficient matrix;

[0018] Based on the similarity quantization method, calculate the Jaccard similarity between the real location and the anonymous location;

[0019] Determine the probability distribution of the real location and the probability distribution of the anonymous location;

[0020] Based on the divergence quantization method, calculate the Hellinger divergence between the real location and the anonymous location from the probability distribution of the real location and the probability distribution of the anonymous location;

[0021] Based on the scale quantization method, calculate the averaged value of the anonymous parameters; the anonymous parameters include the anonymous set users and the size of the anonymous area.

[0022] Optionally, for the averaging value based on the real location, the anonymous location, the quantization value, and the anonymous parameter, the weight value is calculated by using the information gain method and the entropy weight method, specifically including:

[0023] Classify and assign values to the averaging value of the quantization value and the anonymous parameter to generate an evaluation information matrix, and determine the index type of each index in the evaluation information matrix; the index type includes a positive index and a negative index;

[0024] Calculate the entropy value of each index type in the evaluation information matrix by using the entropy value method;

[0025] Based on the real location and the anonymous location, calculate the privacy gain of the location privacy protection algorithm to be evaluated by using the information gain method;

[0026] Calculate the weight of each index type in the evaluation information matrix from the entropy value and the privacy gain.

[0027] Optionally, for calculating the evaluation result of the location privacy protection algorithm to be evaluated based on the quantization value, the averaging value of the anonymous parameter, and the weight value, specifically including:

[0028] Perform weighted summation on the positive index in the evaluation information matrix and the weight of the corresponding index type to obtain the privacy protection degree evaluation value of the location privacy protection algorithm to be evaluated;

[0029] Perform weighted summation on the negative index in the evaluation information matrix and the weight of the corresponding index type to obtain the location service quality evaluation value of the location privacy protection algorithm to be evaluated.

[0030] The present invention also provides a performance evaluation system for a vehicle networking location privacy protection algorithm, including:

[0031] A data acquisition module, configured to acquire the real location of a user entity;

[0032] An anonymization module, configured to process the real location by using the location privacy protection algorithm to be evaluated to obtain an anonymous location;

[0033] An algorithm performance index quantization module, configured to calculate the quantization value between the real location and the anonymous location and the averaging value of the anonymous parameter; the quantization value includes a fusion distance, a Jaccard similarity, and a Hellinger divergence;

[0034] An index quantization value weight determination module, configured to calculate a weight value by using the information gain method and the entropy weight method based on the real location, the anonymous location, the quantization value, and the averaging value of the anonymous parameter; the weight value includes the weight of the quantization value and the weight of the averaging value of the anonymous parameter;

[0035] An algorithm comprehensive evaluation module, which is used to calculate the evaluation result of the location privacy protection algorithm to be evaluated based on the quantization value, the averaged value of the anonymous parameter, and the weight value; the evaluation result includes a privacy protection degree evaluation value and a location service quality evaluation value.

[0036] Optionally, the data acquisition module specifically includes:

[0037] A raw data acquisition unit, which is used to acquire the raw location data of the user entity;

[0038] A screening unit, which is used to clean the raw location data and determine the two-dimensional location coordinates in the raw location data as the real location of the user entity.

[0039] Optionally, the algorithm performance index quantization module specifically includes:

[0040] A first calculation unit, which is used to calculate the Mahalanobis distance and the Euclidean distance between the real location and the anonymous location;

[0041] A second calculation unit, which is used to calculate the correlation coefficient matrix of the anonymous location;

[0042] A fusion distance calculation unit, which is used to determine the fusion distance between the real location and the anonymous location from the Mahalanobis distance, the Euclidean distance, and the correlation coefficient matrix;

[0043] A similarity calculation unit, which is used to calculate the Jaccard similarity between the real location and the anonymous location based on the similarity quantization method;

[0044] A probability distribution determination unit, which is used to determine the probability distribution of the real location and the probability distribution of the anonymous location;

[0045] A divergence calculation unit, which is used to calculate the Hellinger divergence between the real location and the anonymous location from the probability distribution of the real location and the probability distribution of the anonymous location based on the divergence quantization method;

[0046] An anonymous parameter averaging unit, which is used to calculate the averaged value of the anonymous parameter based on the scale quantization method; the anonymous parameter includes the anonymous set users and the anonymous area size.

[0047] Optionally, the index quantization value weighting module specifically includes:

[0048] An information matrix generation unit, which is used to classify and assign values to the quantization value and the averaged value of the anonymous parameter, generate an evaluation information matrix, and determine the index type of each index in the evaluation information matrix; the index type includes a positive index and a negative index;

[0049] An entropy value calculation unit for calculating the entropy values of each index type in the evaluation information matrix by using the entropy value method;

[0050] A privacy gain calculation unit for calculating the privacy gain of the location privacy protection algorithm to be evaluated based on the real location and the anonymous location by using the information gain method;

[0051] An index weight calculation unit for calculating the weights of each index type in the evaluation information matrix from the entropy value and the privacy gain.

[0052] Optionally, the algorithm comprehensive evaluation module specifically includes:

[0053] A privacy protection degree evaluation unit for weighted summing the positive indexes in the evaluation information matrix and the weights of the corresponding index types to obtain the privacy protection degree evaluation value of the location privacy protection algorithm to be evaluated;

[0054] A location service quality evaluation unit for weighted summing the reverse indexes in the evaluation information matrix and the weights of the corresponding index types to obtain the location service quality evaluation value of the location privacy protection algorithm to be evaluated.

[0055] Compared with the prior art, the beneficial effects of the present invention are:

[0056] The present invention proposes a method and system for evaluating the performance of a location privacy protection algorithm in a vehicle network. The real location is processed by using the location privacy protection algorithm to be evaluated to obtain an anonymous location; the fusion distance, Jaccard similarity, Hellinger divergence quantization value, and the averaged value of the anonymous parameter between the real location and the anonymous location are calculated; the weight values corresponding to the above four indexes are calculated by using the information gain method and the entropy weight method; the privacy protection degree evaluation value and the location service quality evaluation value of the location privacy protection algorithm to be evaluated are calculated based on the weight values. The present invention has universal applicability, solves the problems that the existing evaluation methods are not sufficient to support the performance requirements such as the high-speed movement of vehicles and the storage burden of the system, and most of the indexes are aimed at specific applications, attack models, and privacy threats, and solves the limitations brought by the single distance measurement method. The method or system of the present invention is applicable to the evaluation of various location privacy protection algorithms in a vehicle network, so as to select the location privacy protection algorithm with the optimal performance and improve the security of location privacy in a vehicle network. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required to be used in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention, and those of ordinary skill in the art can also obtain other drawings based on these drawings without creative efforts.

[0058] Figure 1 It is a framework diagram of the vehicle networking system based on LBS provided by an embodiment of the present invention;

[0059] Figure 2 It is a flowchart of the performance evaluation method for the vehicle networking location privacy protection algorithm provided by an embodiment of the present invention;

[0060] Figure 3 It is a corresponding evaluation flowchart of privacy protection degree and service quality provided by an embodiment of the present invention;

[0061] Figure 4 It is a structure diagram of the vehicle networking location privacy protection algorithm performance evaluation system provided by an embodiment of the present invention;

[0062] Figure 5 It is a schematic diagram of the evaluation results for evaluating three location privacy protection algorithms derived from PPA, P2P-IS-CA-HL, and SCAPGID k-anonymity provided by an embodiment of the present invention; among them, Figure 5 (a) is a schematic diagram of the change of privacy protection degree with the number of set vehicles when evaluating three location privacy protection algorithms derived from k-anonymity; Figure 5 (b) is a schematic diagram of the change of service quality with the number of set vehicles when evaluating three location privacy protection algorithms derived from k-anonymity;

[0063] Figure 6 It is a schematic diagram of the change of the anonymity success rate with the anonymity degree k when evaluating three location privacy protection algorithms derived from PPA, P2P-IS-CA-HL, and SCAPGID k-anonymity provided by an embodiment of the present invention. Detailed implementation manners

[0064] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0065] The purpose of the present invention is to provide a method and system for evaluating the performance of a vehicle networking location privacy protection algorithm, to solve the limitation that traditional evaluation algorithms are only applicable to the evaluation of a single privacy protection algorithm, to provide a theoretical basis and technical support for subsequent researchers to select and improve appropriate-level privacy protection schemes, and to improve the security of vehicle networking location privacy.

[0066] To make the above objects, features, and advantages of the present invention more obvious and understandable, the present invention will be further described in detail below in conjunction with the accompanying drawings and specific implementation manners.

[0067] Figure 1 This is the framework diagram of the vehicle networking system based on LBS provided by the embodiments of the present invention. Refer to Figure 1 , the constructed performance evaluation model of the privacy protection algorithm includes a vehicle networking system and a security structure framework. This model also follows the three-layer architecture of the sensing layer, network layer, and application layer of Mobile Ad-hoc Networks (MANETs).

[0068] Among them, the security threats faced by the sensing layer are as follows: The sensing nodes arranged in the environment for data collection are easily modified and maliciously exploited, stealing the location privacy of communication vehicles and other user-sensitive information. Usually, mechanisms such as node authentication and security management are adopted to avoid privacy leakage; multi-sensor data fusion is to perform data authenticity verification and abnormal node detection, etc.

[0069] The security threats faced by the network layer are as follows: It is mainly divided into three modules: access, transmission, and service. The present invention does not discuss the access security risks of malicious node attacks for the time being. The transmission layer is responsible for effectively transmitting the collected data information, and it is necessary to ensure the accuracy, integrity, and availability of the information; location services such as point-of-interest queries belong to the scope of the service layer.

[0070] The security threats faced by the application layer are as follows: It involves the storage and calculation of a large amount of user privacy information, and technical supports such as big data processing and service management control are required behind various service types.

[0071] The vehicle networking system constructed by the present invention and the privacy protection technology evaluation index measurement draw on the three-layer architecture model in MANETs, and can make full use of the existing cloud service computing indicators. Starting from the authoritative national standardization rules, cc standards, BS7799 standards, and network security risk assessment standards such as ISO / IEC21827-2002 (SSE-CMM), the performance evaluation indicators of the vehicle networking privacy protection algorithm with universality are respectively extracted from the above three-layer architecture, namely, privacy protection degree, service quality, and overhead. In this embodiment, the privacy protection degree and service quality are used as evaluation indicators for evaluation.

[0072] The performance evaluation method of the vehicle networking location privacy protection algorithm provided by this embodiment first anonymizes the data through the location privacy protection algorithm to be evaluated during initialization; then, quantifies the performance indicators of the privacy protection algorithm based on four quantization methods: distance, similarity, divergence, and scale; then weights the quantization values, and finally obtains the comprehensive evaluation value of the algorithm. Figure 2 This is the flowchart of the performance evaluation method of the vehicle networking location privacy protection algorithm provided by the embodiments of the present invention. Refer to Figure 2 , the performance evaluation method of the vehicle networking location privacy protection algorithm in this embodiment specifically includes:

[0073] Step 101: Obtain the real location of the user entity. Real location X ← [x1, x2,..., x i ,..., x n T All elements in are two-dimensional position coordinates.

[0074] The said Step 101 specifically includes:

[0075] Obtain the original location data of the user entity; perform data cleaning (data screening) on the original location data, and determine the two-dimensional position coordinates in the original location data as the real location of the user entity.

[0076] Step 102: Process the real location using the location privacy protection algorithm to be evaluated to obtain an anonymous location. Anonymous location Y ← [y1, y2,..., y j ,..., y n T After Step 102, the data preparation work before evaluation is completed.

[0077] Step 103: Calculate the quantization value between the real location and the anonymous location and the averaged value of the anonymity parameter; the quantization value includes the fusion distance, Jaccard similarity, and Hellinger divergence.

[0078] The said Step 103 specifically includes:

[0079] 1. Quantify based on the method of distance (Distance, D)

[0080] (1) Calculate the Mahalanobis distance between the real location X and the anonymous location Y and the correlation coefficient matrix of the anonymous location Y, as shown in Formulas (1) to (3).

[0081]

[0082]

[0083] In Formulas (1) to (2), MD(X, Y) is the Mahalanobis distance between the real location X and the anonymous location Y; Cov(X, Y) is the covariance matrix between the real location X and the anonymous location Y; the real location X is used as the clustering sample; the anonymous location Y is used as the sample real location space set, also called the Mahalanobis space; is the mean matrix of sample Y; T represents transpose; ∑ Y is the covariance matrix of sample Y; is the inverse matrix of ∑ Y ; E represents expectation; E(X) is the expectation of the real location X; E(Y) is the expectation of the anonymous location Y; Cov(x​​i , y j ) is the covariance of x corresponding to the (i, j)-th element in the covariance matrix i and y j ; when ∑ is the expected matrix of the covariance matrix of variable Y and it is an identity matrix, the Mahalanobis distance is simplified to the Euclidean distance.

[0084]

[0085] Among them, ρ in formula (3) XY is the correlation coefficient between X and Y, D(X) is the variance of the true position, D(Y) is the variance of the anonymous position, and ρ XY The larger it is, the higher the correlation degree between X and Y, and ρ XY is proportional to Coy(X, Y).

[0086] Generate the correlation coefficient matrix C of vector Y from the correlation coefficient ρ XY , and the weight coefficient ω of the fusion distance can be deduced from the correlation coefficient matrix C Y . Subsequently, based on the fusion weight determination idea and the weight coefficient ω, calculate the fusion distance d of the Euclidean distance and the Mahalanobis distance Y , taking into account the correlation and independence of the characteristic variables. Mix

[0087] (2) Calculate the Euclidean distance between the true position X and the anonymous position Y, as shown in formula (4).

[0088]

[0089] When using the method of the Euclidean distance to analyze the differences in multi-dimensional numerical values among individuals, it is necessary to ensure that each dimension index is at the same scale level. Let X and Y be the true position and the fuzzy position (anonymous position) of the vehicle node respectively, and clarify the position coordinates x i , y i The Euclidean distance between them is used to measure the privacy protection degree in the vehicle network, etc.

[0090] (3) Determine the fusion distance between the true position and the anonymous position from the Mahalanobis distance, the Euclidean distance, and the correlation coefficient matrix, as shown in formulas (5) to (6).

[0091] d Mix = ω × MD(X, Y) + (1 - ω) × ED(X, Y) (5)

[0092]

[0093] ​Among them, in formulas (5)-(6), MD(X, Y) is the Mahalanobis distance from vector X to vector Y; ED(X, Y) is the Euclidean distance from vector X to vector Y; C Y is the correlation coefficient matrix of vector Y; |·| is the value of the matrix determinant.

[0094] 2. Quantification is performed using the method based on similarity (Similarity, S).

[0095] Based on the similarity quantification method, calculate the Jaccard similarity between the true location and the anonymous location. Specifically:

[0096] Calculate the Jaccard similarity from dataset X to dataset Y. Consider the set composed of multiple location points as a trajectory, and use the trajectory coincidence degree for evaluation. Compare the similarities and differences between finite sample sets. The larger the Jaccard similarity coefficient, the higher the sample similarity. The Jaccard similarity coefficient formula is as shown in (7).

[0097]

[0098] 3. Quantification is performed using the method based on divergence (F-divergence, F).

[0099] (1) Determine the probability distribution of the true location and the probability distribution of the anonymous location.

[0100] (2) Based on the divergence quantification method, calculate the Hellinger divergence between the true location and the anonymous location from the probability distribution of the true location and the probability distribution of the anonymous location. By the probability difference of accurately locating a certain record with the same attribute value in the original dataset and the anonymous dataset, specifically:

[0101] By comparing measurability, convergence, and sensitivity, calculate the Hellinger divergence, which is used as the divergence function for evaluation; the definition formula of f-divergence is usually transformed into a discrete form. For the Hel divergence of discrete distributions, its definition formula is as shown in formula (8).

[0102]

[0103] The parameters in formula (8) can be explained by the basic definition of f-divergence, that is, assume that there are two probability distributions P = {p1, p2,..., p n} and Q = {q1, q2,..., q n} in the probability space Ω. P is the probability distribution of the true location X, Q is the probability distribution of the anonymous location Y, p i is the i-th probability value in P, q iIf it is the i-th probability value in Q, then the f-divergence between Q and P is shown in Equation (9).

[0104]

[0105] In this embodiment, the Hellinger divergence is calculated using Equation (8).

[0106] 4. Quantization is performed based on the method of the scale (Scale, S).

[0107] Based on the scale quantization method, the averaged value of the anonymous parameters is calculated; the anonymous parameters include anonymous set users (or the number of trajectories) and parameters such as the size of the anonymous area, etc. According to the actual situation of the algorithm to be evaluated, these objective index parameters other than the degree of location anonymization are numbered, that is, j = 1, 2,.... Based on the averaging processing method, combined with the idea of linear dimensionless, the basic indicators are uniformly processed to obtain the distribution value in the range of 0 to 1. Specifically:[[]]END]]

[0108] Based on the averaging processing method, the formula is shown in Equation (10).

[0109]

[0110] In Equation (10), C ij is the original data of the j-th index of the i-th sample in the anonymous parameter dataset; is the processed data, is between 0 and 1, and the numerical distribution is the same as before processing; is the average value of the original data of the j-th index. This method well maintains the overall consistency of the original data, eliminates the influence of dimension and order of magnitude, and retains the information of the variation degree of each original index.

[0111] Step 104: Based on the true position, the anonymous position, the quantization value, and the averaged value of the anonymous parameters, the weight values are calculated using the information gain method and the entropy weight method; the weight values include the weight of the quantization value and the weight of the averaged value of the anonymous parameters.

[0112] To illustrate the problem more clearly, first, the information entropy and privacy gain are defined as follows:

[0113] Definition 1: For N messages with the same probability, the probability p of each message is 1 / N, then the amount of information carried by each message is shown in Equation (11).

[0114] -logp = log(1 / N) (11)

[0115] Definition 2: For a given probability distribution P = {p1, p2,..., p n}, the amount of information carried by this distribution is called the distribution entropy of P, as shown in formula (12).

[0116]

[0117] When P takes (0.5, 0.5), H(P) is 1. If P takes (1, 0), then H(P) is 0. It can be seen that adjusting the weight formula through information entropy can overcome the defect of ignoring the intra-class distribution in weight calculation.

[0118] When for P, the uncertainty of the result before obtaining M is H(P), and the uncertainty of the result after obtaining the information M is H(P|M), then the amount of information about P in the information M is: IG(P, M) = H(P) - H(P|M), where IG represents information gain.

[0119] Definition 3: The degree of privacy improvement after the original location information is anonymized, that is, the probability difference between the original sample set and the anonymized sample set that can accurately locate a certain record through the same attribute value. The calculation formula of the information gain (Information Gain, IG) applied in the present invention is as shown in (13).

[0120]

[0121] In the above formula (13), IG(L, k) represents the privacy gain of the vehicle terminal v after being encrypted by the location privacy protection algorithm with a privacy protection degree of k in the road network. l represents the location coordinate class variable, and L represents the location set, where L = (l1, l2,..., l k ,..., l i ,... n ), H(L) represents the entropy of the probability space that a random location of an entity before anonymization belongs to a certain category, and H(L|k) is the entropy of the probability space that the true location belongs to a certain category after obtaining the returned privacy protection degree k by using the corresponding anonymization algorithm, which implies the amount of information that k can provide in the weight determination process.

[0122] For the privacy protection degree k, its information gain value is the difference in the amount of information of the entire evaluation system when it exists and when it does not exist (this is the so-called "privacy gain"). If it exists, the system entropy is equivalent to formula (11); if it does not exist, it can be considered that k is included in the algorithm to be evaluated, and the value of k is fixed and cannot be changed. Therefore, at this time, it can be considered that its value has two cases: k represents the occurrence of k; represents the non-occurrence of k.

[0123] p(k) represents the probability of the privacy protection degree k appearing in the dataset; p(l) represents the probability of the position coordinate variable l appearing in the total query call process; p(l|k) represents the conditional probability that contains k and belongs to the coordinate anonymization class privacy algorithm; represents the probability that the dataset does not contain k; represents the conditional probability that the dataset does not contain k but there is l.

[0124] Definition 4: The weight calculation formula based on privacy gain and information entropy is shown in (14).

[0125] w ik =IG(L, k)×S i (14)

[0126] In formula (14), w ik is the weight of the evaluation index for the vehicle networking location privacy protection technology, and S i represents the information entropy weighting factor of the intra-class distribution, as shown in formula (18).

[0127] The said step 104 specifically includes:

[0128] 1. Classify and assign values to the quantization value and the averaged value of the anonymization parameter to generate an evaluation information matrix (r ij ) m×n , and i = 1, 2,..., m; j = 1, 2,..., n, and determine the index types of each index in the evaluation information matrix; the index types include positive indexes and negative indexes. Specifically:

[0129] Classify and assign values to the quantization value and the averaged value of the anonymization parameter respectively, a←d Mix ; b←Jaccard similarity; c←Hellinger divergence; According to the positive and negative directions of the judgment and division indexes a, b, c, d for the privacy protection degree and service quality, that is, they are divided into two categories: positive indexes (indexes with better performance when the index is larger) and negative indexes (indexes with better performance when the index is smaller). In this embodiment, there are four indexes a, b, c, d, so m = 2, n = 2.

[0130] 2. Calculate the entropy values of each index type in the evaluation information matrix by using the entropy value method. Specifically:

[0131] (1) Normalize each index in the evaluation information matrix according to formulas (15) - (16). If the index r ij in the evaluation information matrix is a positive index, it is processed according to formula (15), and if the index r ij in the evaluation information matrix is a negative index, it is processed according to formula (16).

[0132] For positive indicators

[0133] For negative indicators

[0134] Wherein, is r ij r after normalization when it is a positive indicator ij ; is r ij r after normalization when it is a negative indicator ij ; and are respectively the maximum and minimum values of the i-th row in the matrix.

[0135] (2) Calculate the proportion of the indicator in the i-th row and j-th column, as shown in formula (17).

[0136]

[0137] (3) Calculate the entropy value of the indicators in the i-th row based on the entropy value method, as shown in formula (18).

[0138]

[0139] Wherein, in formulas (17) to (18), and it is stipulated that when P ij = 0, P ij lnP ij = 0.

[0140] In this embodiment, the evaluation information matrix is It is a 2×2 matrix, a and c are positive indicators, b and d are negative indicators, the entropy value S1 of the indicators in the first row represents the amount of information of the privacy protection degree, and the entropy value S2 of the indicators in the second row represents the amount of information of the location service quality.

[0141] 3. Based on the real location and the anonymous location, calculate the privacy gain of the location privacy protection algorithm to be evaluated by using the information gain method; the calculation formula of the privacy gain is as shown in formula (13).

[0142] 4. Calculate the weights of each indicator type in the evaluation information matrix from the entropy value and the privacy gain. The calculation formula of the weights is as shown in formula (14), so as to obtain the weight w 1k of the positive indicator and the weight w 2k .

[0143] Step 105: Calculate the evaluation result of the location privacy protection algorithm to be evaluated based on the quantization value, the averaged value of the anonymous parameter, and the weight value; the evaluation result includes a privacy protection degree evaluation value and a location service quality evaluation value.

[0144] The specific steps of Step 105 include:

[0145] Sum the positive indicators in the evaluation information matrix weighted by the weights of the corresponding indicator types to obtain the privacy protection degree evaluation value E of the location privacy protection algorithm to be evaluated p ; sum the reverse indicators in the evaluation information matrix weighted by the weights of the corresponding indicator types to obtain the location service quality evaluation value E of the location privacy protection algorithm to be evaluated q . Among them, E p = w 1k * (a + c), E q = w 2k * (b + d).

[0146] Figure 3 This is the corresponding evaluation flowchart of the privacy protection degree and service quality provided by the embodiment of the present invention. See Figure 3 , the quantization methods for measuring the differences between individuals mainly include quantization based on distance, similarity, divergence, and scale; through comparative analysis, it is selected: fuse distance, Jaccard similarity, Hellinger divergence, and averaging method to quantitatively describe the differences.

[0147] Based on Figure 1 the steps 101-103 in the performance evaluation method framework, complete the operation of distinguishing the positive and negative directions of the indicators before Step 104; classify and assign the quantization results respectively, a ← d Mix ; b ← Jaccard similarity; c ← Hellinger divergence; Distinguish the positive and negative directions of the indicators a, b, c, d, that is, distinguish the indicators (a, c) with better performance as the indicators increase and the indicators (b, d) with better performance as the indicators decrease, and establish an evaluation information matrix.

[0148] The present invention also provides a performance evaluation system for a vehicle networking location privacy protection algorithm, Figure 4 This is the structural diagram of the performance evaluation system for the vehicle networking location privacy protection algorithm provided by the embodiment of the present invention. See Figure 4 , this system includes:

[0149] The location data processing module 401 is used for the preparatory work before evaluation; the preparatory work refers to data screening of the set of user entity location coordinates; verifying that the elements in the coordinate set are all two-dimensional coordinates; and obtaining an anonymized set of location coordinates through the algorithm to be evaluated for the processed data. The location data processing module 401 includes: a data acquisition module for acquiring the real location of the user entity; and an anonymization module for processing the real location using the location privacy protection algorithm to be evaluated to obtain an anonymized location.

[0150] The algorithm performance metric quantification module 402 is used for calculating the quantification value between the real location and the anonymized location and the averaged value of the anonymization parameters; the quantification values include the fusion distance, Jaccard similarity, and Hellinger divergence; specifically, the algorithm performance metric quantification module 402 measures the quality of the performance metrics; discusses the location privacy evaluation metrics extracted from the vehicle networking architecture, and intends to use the same method for quantification in the quantification method, applying the methods of fusion distance, Jaccard similarity, Hellinger divergence, and averaging processing, and forming a weight matrix after distinguishing the quantification values in the positive and negative directions; thereby quantifying the accuracy, integrity, availability of the data, and the impact of objective factors on the privacy protection degree and service quality.

[0151] The metric quantification value weighting module 403 is used for calculating the weight values based on the real location, the anonymized location, the quantification value, and the averaged value of the anonymization parameters using the information gain method and the entropy weight method; the weight values include the weight of the quantification value and the weight of the averaged value of the anonymization parameters. Specifically, the metric quantification value weighting module 403 further improves the vehicle networking location privacy protection algorithm performance evaluation metric system; adjusts the weight formula through information entropy to overcome the defect of ignoring the intra-class distribution in weight calculation; introduces the concept of vehicle networking location privacy gain, that is, the degree of privacy improvement after the original location information is anonymized; and models considering the respective applicable scenarios of privacy gain and the entropy weight method.

[0152] The algorithm comprehensive evaluation module 404 is used for calculating the evaluation result of the location privacy protection algorithm to be evaluated based on the quantification value, the averaged value of the anonymization parameters, and the weight value; the evaluation result includes the privacy protection degree evaluation value and the location service quality evaluation value. Specifically, the algorithm comprehensive evaluation module 404 conducts a comprehensive evaluation of the vehicle networking privacy protection algorithm performance; distinguishes the results of the above four quantification methods in the positive and negative directions and normalizes the original data; matches the information gain value of the metric and the weight matrix to calculate the comprehensive evaluation values of the two evaluation metrics of privacy protection degree and location service quality.

[0153] As an optional implementation manner, the data acquisition module specifically includes:

[0154] An original data acquisition unit for acquiring the original location data of a user entity.

[0155] A screening unit for performing data cleaning on the original location data and determining the two-dimensional position coordinates in the original location data as the real location of the user entity.

[0156] As an optional implementation manner, the algorithm performance index quantization module specifically includes:

[0157] A first calculation unit for calculating the Mahalanobis distance and the Euclidean distance between the real location and the anonymous location. A second calculation unit for calculating the correlation coefficient matrix of the anonymous location. A fusion distance calculation unit for determining the fusion distance between the real location and the anonymous location from the Mahalanobis distance, the Euclidean distance, and the correlation coefficient matrix. The first calculation unit, the second calculation unit, and the fusion distance calculation unit constitute a distance-based quantization unit. The distance-based quantization unit measures the individual differences between the position coordinates before and after anonymization, and selects the fusion distance weighted by the Euclidean distance and the Mahalanobis distance for quantization; the weighting weight of the fusion distance is determined by the correlation coefficient matrix generated from the correlation coefficient between the position coordinate sets output by the position data processing module, that is, the vectors before and after anonymization.

[0158] A similarity calculation unit for calculating the Jaccard similarity between the real location and the anonymous location based on the similarity quantization method. Specifically, the similarity calculation unit compares the similarities and differences between finite sample sets, selects the Jaccard similarity, regards the set composed of multiple position points as a trajectory, and uses the trajectory coincidence degree for evaluation.

[0159] A probability distribution determination unit for determining the probability distribution of the real location and the probability distribution of the anonymous location. A divergence calculation unit for calculating the Hellinger divergence between the real location and the anonymous location from the probability distribution of the real location and the probability distribution of the anonymous location based on the divergence quantization method. The probability distribution determination unit and the divergence calculation unit constitute a divergence-based quantization unit. The divergence-based quantization unit selects the Hellinger divergence as the divergence function for evaluation through the probability difference of accurately locating a certain record in the original data set and the anonymous data set for the same attribute value.

[0160] An anonymous parameter averaging unit for calculating the averaged processing value of anonymous parameters based on the scale quantization method; the anonymous parameters include the users in the anonymous set and the size of the anonymous area. Specifically, the anonymous parameter averaging unit uniformly processes each basic index, selects parameters such as the number of users or trajectories in the anonymous set and the size of the anonymous area for quantization; based on the averaging processing method, combined with the idea of linear dimensionless, a distribution value in the range of 0 to 1 is obtained.

[0161] As an alternative implementation, the index quantization value weighting module specifically includes:

[0162] An information matrix generation unit, configured to classify and assign values to the quantization value and the averaged value of the anonymous parameter to generate an evaluation information matrix, and determine the index type of each index in the evaluation information matrix; the index type includes a positive index and a negative index.

[0163] An entropy value calculation unit, configured to calculate the entropy value of each index type in the evaluation information matrix by using the entropy value method;

[0164] A privacy gain calculation unit, configured to calculate the privacy gain of the location privacy protection algorithm to be evaluated based on the real location and the anonymous location by using the information gain method.

[0165] An index weight calculation unit, configured to calculate the weight of each index type in the evaluation information matrix from the entropy value and the privacy gain.

[0166] As an alternative implementation, the algorithm comprehensive evaluation module specifically includes:

[0167] A privacy protection degree evaluation unit, configured to perform weighted summation of the positive indexes in the evaluation information matrix and the weights of the corresponding index types to obtain an evaluation value of the privacy protection degree of the location privacy protection algorithm to be evaluated.

[0168] A location service quality evaluation unit, configured to perform weighted summation of the negative indexes in the evaluation information matrix and the weights of the corresponding index types to obtain an evaluation value of the location service quality of the location privacy protection algorithm to be evaluated.

[0169] According to the dynamic balance requirement between the privacy protection degree and the location service quality, the present invention compares and analyzes classical location privacy protection algorithms, and verifies and optimizes evaluation indexes and strategies. The superiority of the performance of the present invention can be further illustrated by the following simulation experiments:

[0170] Using the python3 compilation environment, install Ubuntu on CW-KNN and VMware virtual machines, and build a corresponding road network environment for experimental simulation.

[0171] Load the vehicle movement model generated in sumo into ns-2 in the form of a trace file to obtain vehicle nodes.

[0172] By simulating a total of 6 two-way straight lanes with intersections, adjust the node distribution density within the RSU signal coverage range of 1500m * 1500m, and discuss the location privacy protection degree and the location service quality during the vehicle driving process.

[0173] The simulation experiment process is divided into two parts: traffic scenario and network communication. The relevant experimental configuration parameters are shown in Table 1.

[0174] Table 1

[0175]

[0176] There are three embodiments selected for simulation by the evaluation method proposed in the present invention as follows:

[0177] The improved P2P privacy protection algorithm (Privacy Preserving Algorithm, PPA) based on PSSM;

[0178] Peer to Peer-Information Sharing-Cloaked Area-Historical Location (P2P-IS-CA-HL) algorithm;

[0179] A comparative experiment is carried out with the Spatial Cloaking Algorithm Based on P2P and Grid ID (SCAPGID).

[0180] Among them, the P2P-IS-CA-HL algorithm is an algorithm for sharing node information and adjusting the area of the anonymous area; the SCAPGID algorithm is an algorithm that divides the plane with grids and completes location anonymity through the dynamic amplification of grids and the cooperation of users in the same grid.

[0181] Furthermore, the evaluation algorithm is reversely verified by analyzing the evaluation results of the effectiveness and authenticity of PPA.

[0182] When the present invention evaluates the location privacy protection algorithms derived from the three k-anonymities of the embodiments PPA, P2P-IS-CA-HL, and SCAPGID, the schematic diagram of the change of the privacy protection degree and service quality with the number of set vehicles is as Figure 5 shown.

[0183] In the experiment, the privacy protection degree and the evaluation value E of the location service quality in the intervals of the number of vehicles [100, 500] and in the interval [0, 1] are respectively compared p 、E q .

[0184] Through analysis, it is obtained that Figure 5 the change situations of the three privacy protection algorithms conform to the objective laws and expected effects, verifying the effectiveness and usability of the present invention.

[0185] It can be seen from this that there is a certain inverse relationship between the two indicators of privacy protection degree and location service quality, that is, the optimal values cannot be achieved simultaneously, and only a balance point can be found according to the user's privacy needs and the amount of information contained in the indicators.

[0186] Through calculation, when the number of vehicles is in the range of [200, 300], the privacy gain IG(l, v k ) ≈ 0.82.

[0187] The comprehensive evaluation value of the algorithm of Embodiment P2P-IS-CA-HL is better than that of SCAPGID, as shown in Table 2, and except for the case where the sample value is extremely small, PPA has obvious advantages and basically meets the improvement requirements;

[0188] Table 2

[0189]

[0190] In the traffic scenario based on the same parameters, the more the number of vehicles, the larger the number of assisting neighbor nodes in the formed anonymous area, thereby improving the anonymity success rate of the PPA algorithm.

[0191] Furthermore, when the number of vehicles increases to a certain scale, the minimum anonymous area of the algorithm cannot meet the anonymity protection degree, and it is necessary to amplify the area of the anonymous area through the other two types of algorithms to make the entire communication network tend to be stable.

[0192] When evaluating the location privacy protection algorithms derived from the three k-anonymities of Embodiment PPA, P2P-IS-CA-HL, and SCAPGID of the present invention, the schematic diagram of the change of the anonymity success rate with the privacy protection degree k is shown in Figure 6.

[0193] The number of vehicles is set to 200. When k increases, the vehicle terminal needs to recruit more peers to help collect enough peer location information to anonymize its location. The pseudonym node and the amplification scheme of the anonymous area in PPA can effectively avoid problems such as network partitioning, so it shows a steady upward state in the figure.

[0194] The method and system for evaluating the performance of the vehicle networking location privacy protection algorithm provided by the present invention have the following advantages:

[0195] (1) The present invention designs a unified performance evaluation method and standard, and develops a generally applicable evaluation index and evaluation system to objectively and reasonably evaluate the anonymization technology. Existing privacy evaluation methods are not sufficient to support performance requirements such as the high-speed movement of vehicles and the storage burden of the system, and most of the indicators are for specific applications, attack models, and privacy threats.

[0196] (2) The present invention solves the limitations brought by a single distance measurement method. In the environment of the high mobility and instantaneity characteristics of the vehicle networking, using only the Euclidean distance as the measurement method may make it difficult to explain the actual meaning of the dimension between variables, ignore the distribution of samples, and be unable to explain the correlation between multivariate data. Selecting the fusion distance to quantify the indicators can take into account both the correlation and independence of feature variables at the same time.

[0197] (3) The present invention can efficiently achieve dynamic evaluation. The quality of location services and the quality of location privacy protection are almost inversely proportional. According to the privacy protection requirements of users in different context environments, the weights of performance indicators are adjusted targeted, so that the privacy protection algorithm can adapt to the privacy protection requirements at different levels.

[0198] (4) The present invention constructs a scientific and reasonable mathematical model and performs multiple nestings according to the quantitatively described features to achieve complementary advantages. The two relevant evaluation indicators of the privacy protection degree and the location service quality are quantified one by one, which not only reflects the proportional relationship between the two, but also facilitates the consistent magnitude adjustment.

[0199] In this specification, each embodiment is described in a progressive manner. The key point of each embodiment is to illustrate the differences from other embodiments. For the same or similar parts among the embodiments, reference can be made to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and reference can be made to the description in the method part for the relevant parts.

[0200] In this article, specific examples are used to elaborate on the principle and implementation manner of the present invention. The description of the above embodiments is only used to help understand the method and its core idea of the present invention; at the same time, for those of ordinary skill in the art, according to the idea of the present invention, there will be changes in the specific implementation manner and application scope. In summary, the content of this specification should not be construed as a limitation to the present invention.

Claims

1. A method for evaluating the performance of a location privacy protection algorithm in a vehicle networking system, characterized in that, Including: Obtain the real location of the user entity; Process the real location using the location privacy protection algorithm to be evaluated to obtain an anonymous location; Calculate the quantization value between the real location and the anonymous location and the averaged value of the anonymous parameter; the quantization value includes the fusion distance, Jaccard similarity, and Hellinger divergence; Based on the real location, the anonymous location, the quantization value, and the averaged value of the anonymous parameter, calculate the weight value using the information gain method and the entropy weight method; the weight value includes the weight of the quantization value and the weight of the averaged value of the anonymous parameter; Calculate the evaluation result of the location privacy protection algorithm to be evaluated based on the quantization value, the averaged value of the anonymous parameter, and the weight value; the evaluation result includes the privacy protection degree evaluation value and the location service quality evaluation value; The calculating the weight value using the information gain method and the entropy weight method based on the real location, the anonymous location, the quantization value, and the averaged value of the anonymous parameter specifically includes: Classify and assign values to the quantization value and the averaged value of the anonymous parameter to generate an evaluation information matrix, and determine the index types of each index in the evaluation information matrix; the index types include positive indexes and reverse indexes; Calculate the entropy value of each index type in the evaluation information matrix using the entropy value method; Calculate the privacy gain of the location privacy protection algorithm to be evaluated using the information gain method based on the real location and the anonymous location; Calculate the weight of each index type in the evaluation information matrix from the entropy value and the privacy gain.

2. The method for evaluating the performance of a location privacy protection algorithm in a vehicle networking system according to claim 1, characterized in that, The obtaining the real location of the user entity specifically includes: Obtain the original location data of the user entity; Clean the original location data, and determine the two-dimensional location coordinates in the original location data as the real location of the user entity.

3. The method for evaluating the performance of a location privacy protection algorithm in a vehicle networking system according to claim 1, characterized in that, The calculating the quantization value between the real location and the anonymous location and the averaged value of the anonymous parameter specifically includes: Calculate the Mahalanobis distance and Euclidean distance between the real location and the anonymous location; Calculate the correlation coefficient matrix of the anonymous location; Determine the fusion distance between the real location and the anonymous location from the Mahalanobis distance, the Euclidean distance, and the correlation coefficient matrix; Calculate the Jaccard similarity between the real location and the anonymous location based on the similarity quantization method; Determine the probability distribution of the real location and the probability distribution of the anonymous location; Calculate the Hellinger divergence between the real location and the anonymous location from the probability distribution of the real location and the probability distribution of the anonymous location based on the divergence quantization method; Calculate the averaged value of the anonymous parameter based on the scale quantization method; the anonymous parameter includes the anonymous set users and the size of the anonymous area.

4. The method for evaluating the performance of a location privacy protection algorithm in a vehicle networking system according to claim 1, characterized in that, The calculating the evaluation result of the location privacy protection algorithm to be evaluated based on the quantization value, the averaged value of the anonymous parameter, and the weight value specifically includes: Sum the positive indexes in the evaluation information matrix and the weights of the corresponding index types to obtain the privacy protection degree evaluation value of the location privacy protection algorithm to be evaluated; The reverse indicators in the evaluation information matrix are weighted and summed with the weights of the corresponding indicator types to obtain the location service quality evaluation value of the location privacy protection algorithm to be evaluated.

5. A system for evaluating the performance of a location privacy protection algorithm in a vehicle networking system, characterized in that, It includes: A data acquisition module for acquiring the real location of the user entity; An anonymization module for processing the real location using the location privacy protection algorithm to be evaluated to obtain an anonymized location; An algorithm performance indicator quantification module for calculating the quantification value between the real location and the anonymized location and the averaged value of the anonymization parameters; the quantification values include the fusion distance, Jaccard similarity, and Hellinger divergence; An indicator quantification value weight determination module for calculating the weight values using the information gain method and entropy weight method based on the real location, the anonymized location, the quantification value, and the averaged value of the anonymization parameters; the weight values include the weight of the quantification value and the weight of the averaged value of the anonymization parameters; An algorithm comprehensive evaluation module for calculating the evaluation result of the location privacy protection algorithm to be evaluated based on the quantification value, the averaged value of the anonymization parameters, and the weight values; the evaluation results include the privacy protection degree evaluation value and the location service quality evaluation value; The indicator quantification value weight determination module specifically includes: An information matrix generation unit for classifying and assigning values to the quantification value and the averaged value of the anonymization parameters to generate an evaluation information matrix and determining the indicator types of each indicator in the evaluation information matrix; the indicator types include positive indicators and reverse indicators; An entropy value calculation unit for calculating the entropy values of each indicator type in the evaluation information matrix using the entropy value method; A privacy gain calculation unit for calculating the privacy gain of the location privacy protection algorithm to be evaluated using the information gain method based on the real location and the anonymized location; An indicator weight calculation unit for calculating the weights of each indicator type in the evaluation information matrix from the entropy value and the privacy gain.

6. A performance evaluation system for a vehicle networking location privacy protection algorithm according to claim 5, wherein, The data acquisition module specifically includes: An original data acquisition unit for acquiring the original location data of the user entity; A screening unit for cleaning the original location data and determining the two-dimensional location coordinates in the original location data as the real location of the user entity.

7. A performance evaluation system for a vehicle networking location privacy protection algorithm according to claim 5, wherein, The algorithm performance indicator quantification module specifically includes: A first calculation unit for calculating the Mahalanobis distance and Euclidean distance between the real location and the anonymized location; A second calculation unit for calculating the correlation coefficient matrix of the anonymized location; A fusion distance calculation unit for determining the fusion distance between the real location and the anonymized location from the Mahalanobis distance, the Euclidean distance, and the correlation coefficient matrix; A similarity calculation unit for calculating the Jaccard similarity between the real location and the anonymized location based on the similarity quantification method; A probability distribution determination unit for determining the probability distribution of the real location and the probability distribution of the anonymized location; A divergence calculation unit for calculating the Hellinger divergence between the real location and the anonymized location based on the divergence quantification method from the probability distribution of the real location and the probability distribution of the anonymized location; An anonymous parameter averaging unit for calculating an averaged value of anonymous parameters based on a scale quantization method; the anonymous parameters include anonymous set users and the size of an anonymous area.

8. A performance evaluation system for a vehicle networking location privacy protection algorithm according to claim 5, wherein, The algorithm comprehensive evaluation module specifically includes: A privacy protection degree evaluation unit for weighted summing the positive indicators in the evaluation information matrix with the weights of the corresponding indicator types to obtain a privacy protection degree evaluation value of the location privacy protection algorithm to be evaluated; A location service quality evaluation unit for weighted summing the reverse indicators in the evaluation information matrix with the weights of the corresponding indicator types to obtain a location service quality evaluation value of the location privacy protection algorithm to be evaluated.

Images