A trajectory data-based outsourcing cloud environment privacy protection infection mode mining method

Abstract

The application belongs to the field of cloud computing security, and discloses a trajectory data-based outsourcing cloud environment privacy protection infection mode mining method, which is divided into two stages: the first stage: trajectory data preprocessing, encryption and outsourcing, first, the data owner pre-processes the trajectory data set to be uploaded, generates a corresponding encoding matrix for the trajectory data of each object, then generates a secure index matrix for the encoding matrix through a strong anti-collision one-way hash function, and finally uploads the encrypted trajectory data and the generated secure index matrix to a cloud server, and shares the key with authorized users; the second stage: a privacy protection infection mode mining method, first, the authorized user sends the infected object number to the cloud server, the cloud server performs infection mode mining after receiving the infected object number, and returns the mining result to the authorized user. The application can ensure high accuracy of the mining result, improve the mining efficiency, and is easy to implement.

Description

Technical Field

[0001] This invention belongs to the field of cloud computing security, specifically relating to a method for mining infection patterns in outsourced cloud environments based on trajectory data to protect privacy. Background Technology

[0002] Pattern recognition is a crucial tool for controlling pandemic outbreaks such as COVID-19. Currently, some small and medium-sized service providers outsource large amounts of trajectory data to the cloud. However, security and privacy are primary concerns in this service model because cloud service providers are not entirely trustworthy and may leak or infer confidential geographic data, thereby inferring sensitive user information and compromising user privacy.

[0003] Traditional methods for detecting infection patterns involve using infected individuals' activities via infected trackers to identify infected targets, but this is extremely time-consuming during large-scale outbreaks. Other methods use Bluetooth pairing to protect users' absolute location and avoid exposing location privacy. However, ignoring absolute location leads to low accuracy in pattern detection. Therefore, a method for detecting infection patterns that balances accuracy and user privacy is needed. Summary of the Invention

[0004] To address the issue of cloud service providers accessing users' sensitive information without authorization while ensuring the accuracy of the mining results during infection pattern mining, this invention proposes a privacy-preserving infection pattern mining method based on trajectory data in an outsourced cloud environment. This method can mine infection patterns while ensuring privacy and security, and is simple to implement, yields high-accuracy mining results, and is fast.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solution:

[0006] This invention proposes a privacy-preserving infection pattern mining method for outsourced cloud environments based on trajectory data.

[0007] First, the relevant symbols and the definitions of the infection pattern are given:

[0008] Suppose that the data owner owns a trajectory dataset O = {O1, O2, ..., O} of n objects. n}, where O i ={(x i,1 ,y i,1 , t1), (x i,2 , y i,2 , t2), …, (x i,m , y i,m , t m )} represents object O iThe trajectory, (x i,1 ,y i,1 ,t1) O i Indicates that the i-th object is in t j The two-dimensional coordinates of the time sampling points are =(x i,j ,y i,j ).

[0009] Definition 1 Infection Mode: Given a distance threshold δ and a sliding time window threshold τ, assuming object o is already infected with the virus, if at time t, object o... i If the following conditions are met, then o is called o. i It is infected by o at time t.

[0010] Condition 1:

[0011] Condition 2: |T c |= And T c .begin≥et

[0012] Wherein, D(o, o) i , t j ) represents t j Time o and o i The Euclidean distance between them, T c For object o at time t i Sliding window, T represents the threshold of the sliding time window. c .begin indicates that in T c The occurrence time of the first sampling point within the time window, et indicates the time when object o was determined to be infected.

[0013] The determination of the infected object in the above definition is performed in plaintext, which makes it impossible to securely mine infection patterns. In order to protect the privacy of trajectory data, the plaintext trajectory data needs to be processed, and infection pattern mining also needs to be supported.

[0014] This invention proposes a privacy-preserving infection pattern mining method based on trajectory data in an outsourced cloud environment. The method is characterized by two stages: first, the preprocessing, encryption, and outsourcing of trajectory data; and second, privacy-preserving infection pattern mining.

[0015] The preprocessing, encryption, and outsourcing stages of the trajectory data include the following steps:

[0016] Suppose that the data owner owns a trajectory dataset O = {O1, O2, ..., O} of n objects. n}, where O i ={(x i,1 ,yi,1 , t1), (x i,2 , y i,2 , t2), …, (x i,m , y i,m , t m )} represents the trajectory of the object O i , (x i,1 , y i,1 , t1) O i represents the two-dimensional coordinates of the i-th object at the t j time sampling point = (x i,j , y i,j ).

[0017] (1a) The data owner pre-processes each object O i in the trajectory data set O: first, the corresponding encoding matrix is generated for the trajectory data set of the object O i ; then, the encoding matrix is unidirectionally privacy encoded using the HMAC function to generate a secure index hash matrix , and finally, the trajectory data is encrypted using the key k f .

[0018] (1a1) Construct a projection function h : mapping a two-dimensional vector to an integer

[0019]

[0020] where the vector is a two-dimensional vector in polar form, where the radius , the angle is randomly selected in the range , the parameter , is the infection distance threshold, and the parameter b is a constant value determined by the parameters , the vector , and the parameter D, and its value is

[0021]

[0022] (1a2) Select l groups of unit vectors for each sampling point of the trajectory data of the object O i , and each group of unit vectors includes m sub-vectors

[0023]

[0024] …

[0025]

[0026] Each set of vectors Between each adjacent sub-vector All included angles .

[0027] Given parameter δ, according to parameter b in step (1a1) i The calculation formula can be used to calculate the parameter vector group A. i1 A i2 ,…,A il The corresponding parameter set B i1 B i2 ,…,B il

[0028]

[0029] …

[0030]

[0031] (1a3) Based on object O i Trajectory dataset and parameter group A i B i The projection function value matrix is ​​calculated.

[0032]

[0033] Among them, H ij O represents the trajectory data of the i-th object. i Regarding parameter group A corresponding to the j-th object i1 A i2 ,…,A il B i1 B i2 ,…,B il The projection function matrix, Represents trajectory data O j The t-th sub-vector in the corresponding k-th vector group and parameters Corresponding projection function value

[0034]

[0035] For trajectory data O i The position vector corresponding to the kth sampling point.

[0036] (1a4) Use a fixed-length encoding method (such as MD5 code) to project the function value The encoding is performed, and the encoding of the order t is connected with a fixed bit (such as 32 bits) after the last bit is encoded, and the projection function value The encoding value is . Constructing the trajectory data O i The encoding matrix S i

[0037]

[0038] Wherein, "||" is the encoding connector.

[0039] (1a5) Using a strong anti-collision one-way hash function hash (such as SHA-1, MD5, etc.), the encoding matrix S i is mapped to a secure index hash matrix

[0040]

[0041] Wherein,

[0042] (1b) The data owner uploads all encrypted trajectory data and secure index hash matrix to the cloud server, and shares the key k f with authorized users.

[0043] The privacy-preserving infection pattern mining method stage includes the following steps:

[0044] (1c) The authorized user sends the infected object number to the cloud server, and then waits for the server to return the mining result;

[0045] (1c1) The cloud server receives the infection object number uploaded by the authorized user, and stores the infection pattern mining result in the form of infection event triple e = <origin, object, t>, wherein origin is the infection source object number, object represents the infected object number, and t is the tth time sampling point of object object infected by origin object; If origin does not exist, object is the initial infection source object, that is, i. P is the current set of infection events to be analyzed, and Final is the set of determined infection events. The cloud server adds the initial infection source object i uploaded by the authorized user to P and Final in the form of infection event triple.

[0046] (1c2) Investigate the infected object e.object of each infection event e in P, given the infection distance threshold and the sliding time window threshold , the set of all objects is denoted as O, and the object set to be processed is denoted as O' = O-{e'.object | e' Final}. For each object o in O', if there exists a time window T c satisfying the following conditions, then add the infection event triple (e.object, o, T c .end) to Upgrate.

[0047] Condition 1: T c .begin≥e.t

[0048] Condition 2: T c .begin T T c .end (Ha o [e.object][T]=Ha e.object [e.object][T])

[0049] where, T c .begin and T c .end respectively represent the occurrence times of the first sampling point and the last sampling point within the time window T c , T represents all sampling points within the time window T c . Ha o represents the hash matrix corresponding to the object o, and Ha e.object represents the hash matrix corresponding to the object e.object.

[0050] It is not difficult to prove that when Condition 1 holds, the Euclidean distance between the object o and the object e.object at time T is not greater than the infection determination distance δ. Therefore, this condition can be used to determine whether the object o and the object e.object satisfy the infection pattern definition at time T, and then correctly perform infection pattern mining.

[0051] (1c3) Update the candidate infection event set Select according to Upgrate obtained in step (1c2). The update method is as follows:

[0052] Method 1: If there exists an infection event e' in Select that simultaneously satisfies e.object = e'.object and e.time < e'.time, then update e' with e, making e'.time = e.time and e'.source = e.source;

[0053] Method 2: If there exists an infection event e' in Select that simultaneously satisfies e.object = e'.object and e.time ≥ e'.time, then directly ignore e;

[0054] Method 3: If there is no any infection event e' in Select satisfying e.object = e'.object, the infection event e is directly added to Select.

[0055] (1c4) judge whether Select is empty, if Select is not empty, the infection event with the earliest occurrence time in Select is added to P, at the same time, Final is added, and step (1c2) is transferred; if Select is empty, it indicates that the infection mode mining is ended, Final is the final determined infection mode mining result set and is returned to the authorized user.

[0056] (1d) after the authorized user receives the infection mode mining result returned in (1c), the user trajectory in the mining result set is decrypted by using the key obtained in (1b), and the infection mode mining is ended.

[0057] The method has the advantages that the method constructs a secure index matrix based on a hash technology, and can determine the specific propagation path of a virus and the infected object safely and efficiently under the condition of a given infection source, and has significant improvement in index security, mining efficiency and other aspects.

[0058] The infection mode mining method based on trajectory data and the outsourcing cloud environment privacy protection provided by the application can mine the infection mode under the condition of ensuring privacy security, is simple to implement, has high mining result precision and fast mining speed.

[0059] In summary, the infection mode mining method based on trajectory data and the outsourcing cloud environment privacy protection considers a service model composed of a data owner, a cloud server and an authorized user. Under the model of the cloud server being honest and curious, in order to protect the location privacy, the data owner needs to encrypt the trajectory data before outsourcing, at the same time, based on the combination of a projection function and a hash technology, the secure index matrix is uploaded to the cloud server, the authorized user provides the infection number to the cloud server, the infection mode mining is carried out in the cloud server, and the final mining result is transmitted to the authorized user. BRIEF DESCRIPTION OF DRAWINGS

[0060] Figure 1 It is a system architecture diagram of the application.

[0061] Figure 2 It is a first stage flow chart of the application.

[0062] Figure 3 It is a second stage flow chart of the application.

[0063] Figure 4 It is a theorem illustration diagram.

[0064] Figure 5 Fig. 1 is a schematic diagram of an infection mode. DETAILED DESCRIPTION

[0065] Embodiments of the present application will be described below with reference to drawings, and many practical details will be described in the following description for the purpose of explanation. However, it should be understood that these practical details are not intended to limit the present application. That is, these practical details are not necessary in some embodiments of the present application.

[0066] The present application is a trajectory data-based privacy protection infection mode mining method in an outsourcing cloud environment, as shown in Fig. 1. Figure 2 Fig. 2 is a flowchart of the first stage of the present application, and the first stage is the preprocessing, encryption and outsourcing of trajectory data, and the specific steps are as follows:

[0067] Step 1: The data owner pre-processes each object O in the trajectory data set O i First, generate the corresponding encoding matrix S for the trajectory data set of the object O i ; then, use the HMAC function to perform one-way private encoding on the encoding matrix to generate a secure index hash matrix i ; and finally, encrypt the trajectory data using the key k f .

[0068] (1) Construct a projection function h : Map the two-dimensional vector to an integer

[0069]

[0070] where the vector represents a two-dimensional vector in polar form, where the radius , the angle is randomly selected in the range , the parameter , is the infection distance threshold, and the parameter b is a constant value determined by the parameters , the vector , and the parameter D, which has a value of

[0071]

[0072] (2) Select l groups of unit vectors for each sampling point i of the trajectory data of the object O , and each group of unit vectors includes m sub-vectors

[0073]

[0074] … ​

[0075]

[0076] where each group of vectors is adjacent to each other with an included angle of .

[0077] Given a parameter δ, according to the calculation formula of step (1) parameter b i , the parameter vector group A i1 , A i2 , …, A il corresponding to the parameter set B i1 , B i2 , …, B il can be calculated.

[0078]

[0079] …

[0080]

[0081] (3) According to the object O i trajectory data set and the parameter group A i , B i , the projection function value matrix

[0082]

[0083] where H ij represents the trajectory data O i of the i-th object. The projection function matrix of the j-th object corresponding to the parameter group A i1 , A i2 , …, A il , B i1 , B i2 , …, B il , represents the t-th sub-vector in the k-th vector group corresponding to the trajectory data O j and the projection function value corresponding to the parameter

[0084]

[0085] is the position vector corresponding to the k-th sampling point in the trajectory data O i .

[0086] (4) The projection function value ​Encode and concatenate the encoding of order t with a fixed number of bits (e.g., 32 bits) after the last encoding, set the projection function value The encoding value is . Construct the trajectory data O i The encoding matrix S i

[0087]

[0088] Wherein, "||" is the encoding connector.

[0089] (5) Using a strong anti-collision one-way hash function hash (such as SHA-1, MD5, etc.), the encoding matrix S i is mapped to a secure index hash matrix

[0090]

[0091] Wherein,

[0092] Step 2: The data owner uploads all encrypted trajectory data and secure index matrix to the cloud server, and shares the key k f with authorized users.

[0093] As Figure 3 , the second phase flowchart of the present application.

[0094] The second phase is a privacy-preserving infection pattern mining method, which specifically includes the following steps:

[0095] Step 1: The authorized user sends the infected object number to the cloud server, and then waits for the server to return the mining results;

[0096] (1) The cloud server receives the infection object number uploaded by the authorized user, and stores the infection pattern mining results in the form of infection event triple e = <origin, object, t>, wherein origin is the infection source object number, object represents the object number of the infected object, and t is the tth time sampling point of the object object infected by the origin object; If origin does not exist, object is the initial infection source object, that is, i. P is the current set of infection events to be analyzed, and Final is the set of determined infection events. The cloud server adds the initial infection source object i uploaded by the authorized user to P and Final in the form of infection event triple.

[0097] (2) Investigate the infected object e.object of each infection event e in P, and give the infection distance threshold and the sliding time window threshold , all objects are denoted as O, and the set of objects to be processed is denoted as O' = O-{e'.object | e' Final}. For each object o in O', if there exists a time window T c that satisfies the following conditions, the infection event triple (e.object, o, T c .end) is added to Upgrate.

[0098] Condition 1: T c .begin≥e.t

[0099] Condition 2: T c .begin T T c .end (Ha o [e.object][T]=Ha e.object [e.object][T])

[0100] where T c .begin and T c .end represent the occurrence times of the first sampling point and the last sampling point within the T c time window, respectively, and T represents all sampling points within the T c time window. Ha o represents the hash matrix corresponding to object o, and Ha e.object represents the hash matrix corresponding to object e.object.

[0101] It is not difficult to prove that when Condition 1 is satisfied, the Euclidean distance between object o and object e.object at time T is not greater than the infection determination distance δ. Therefore, this condition can be used to determine whether object o and object e.object satisfy the infection pattern definition at time T, and thus correctly perform infection pattern mining.

[0102] (3) According to the Upgrate obtained in step (2), the candidate infection event set Select is updated. The updating method is as follows:

[0103] Method 1: If there is an infection event e' in Select, and e.object = e'.object and e.time < e'.time are satisfied, then e is used to update e' such that e'.time = e.time and e'.source = e.source;

[0104] Method 2: If there exists an infection event e' in Select, and e.object = e'.object and e.time >= e'.time are satisfied, then directly ignore e.

[0105] Method 3: If there does not exist any infection event e' in Select satisfying e.object = e'.object, then directly add the infection event e into Select.

[0106] (4) Determine whether Select is empty, if Select is not empty, add the infection event with the earliest time in Select into P, and add it into Final, and go to step (2); if Select is empty, it indicates that the infection mode mining is completed, Final is the final determined infection mode mining result set, and the final determined infection mode mining result set is returned to the authorized user.

[0107] Step 2: After the authorized user receives the infection mode mining result returned in step (4), the user trajectory in the mining result set is decrypted by using the key obtained in step 2 of the first stage, and the infection mode mining is completed.

[0108] The effect of the outsourcing cloud environment privacy protection infection mode mining method of the application is high accuracy of the mining result, and the specific content is as follows:

[0109] Theorem 1: Given a point , for the projection function , The feasible region of the projection function is located between two parallel lines l1 and l2 of the vertical vector, and the distance is equal to D, which is irrelevant to the selection of the point and the value of the parameter b, and the points with equal projection function values are located in the same feasible region.

[0110] Proof: Taking Figure 4 as an example, considering a projection function with b = 0, given a point , a parameter vector , and assuming , then Therefore, the projection function values of the points in the feasible region are all 1, and the function values of the points not in the feasible region are not equal to 1. Therefore, if there are two points with equal projection function values, they must be located in the same feasible region.

[0111] Theorem 2: The parameter b selected in step (1b1) ensures that the given point is in the feasible region with the point ​The circumscribed polygon is centered at a point with a radius of δ=D / 2. The number of sides depends on the number of basis vectors m chosen. When m is large enough, two points can be considered to be in the same feasible region if the distance between the two points is less than δ.

[0112] Proof: The parameter b acts as a leftward shift of b units when b=0. Assume b=D / 3. Then when When b=0, its function value is 1. When the function value is 1, the point is... The feasible region was shifted to the left by D / 3 units. At that time, its feasible region is translated by exactly D / 2 units, thus guaranteeing the point It is located at the midpoint of its feasible region.

[0113] According to Theorem 1, points with equal projection functions lie in the same feasible region. This applies when only one parameter vector is considered. At this point, the feasible region is an infinitely large region between two parallel lines with a distance of D / 2. Furthermore, consider two parameter vectors... ,and ,like Figure 4 If given a point Existence point satisfy and This explains the point. All are located at points Within the feasible region of the centered square. And so on, if the parameter vector is gradually increased... The number of m, and ensure that its angle is within If the value is uniformly selected within the range, and a point exists... Each satisfies ,…, This explains the point. All are located at points The feasible region of the central polygon, which is a polygon centered on the point Let m be the center of the circle, and let m be the radius of the circumscribed polygon of the circle. If m is large enough, then the point... The circumscribed polygon gradually approaches the point A circle with center δ and radius δ = D / 2 can be considered as follows: The distance between them is no greater than δ.

[0114] According to Theorems 1 and 2, it can be shown that if a given point... Existence point respectively ,…, This explains the point. All are located at points The feasible region of the polygon centered at the point is approximately considered as the point The distance between the points is not greater than δ. After the projection function value is encoded in step (1a4), the equivalence relation becomes that if the encodings of the points are the same, it indicates that the distance between the points is not greater than δ. Furthermore, the hash function selected in step (1b5) is a strong anti-collision one-way hash function, which greatly ensures that if the encodings of two points are the same, the hash values after hash mapping are equal; similarly, it greatly ensures that if the hash values of two points are equal, the encodings of the two points are the same. Therefore, the equivalence relation becomes that if the hash values of the points after security indexing are equal, it indicates that the distance between the points is not greater than δ. Therefore, in step (1c2), directly comparing the hash values of two points in the sliding window is equivalent to judging whether the distance between the two points is not greater than δ.

[0115] In summary, the present application can ensure high accuracy of the mining result and there is no missed judgment. The above description is only an embodiment of the present application and is not used to limit the present application. For those skilled in the art, the present application can have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the scope of the claims of the present application.

Claims

1. A method for mining infection patterns in an outsourced cloud environment based on trajectory data while protecting privacy, characterized in that: This method comprises two stages: the first stage involves preprocessing, encryption, and outsourcing of trajectory data; the second stage involves privacy-preserving infection pattern mining. The first stage of trajectory data preprocessing, encryption, and outsourcing specifically involves: Assuming the data owner possesses a trajectory dataset O = {O1, O2, ..., O...} of n objects. n }, where O i ={(x i,1 , y i,1 , t1), (x i,2 , y i,2 , t2), …,(x i,m ,y i,m , t m )} represents object O i The trajectory, (x i,1 ,y i,1 ,t1) O i Indicates that the i-th object is in t j The two-dimensional coordinates of the time sampling points are =(x i,j ,y i,j The process includes the following steps: Step 1-1: The data owner manages each object O in the trajectory dataset O. i Preprocessing: First, for object O i Generate corresponding encoding matrices from trajectory datasets. Then, the HMAC function is used to perform one-way privacy encoding on the encoding matrix to generate a secure index hash matrix. Finally, using the key k f Encrypt the trajectory data; Steps 1-2: The data owner uploads all encrypted trajectory data and the security index hash matrix to the cloud server, and simultaneously sends the key k f Share with authorized users; Step 1-1 specifically includes the following steps: Step 1-1-1: Construct the projection function h: Two-dimensional vector Mapped to a certain integer Where, vector Let represent a two-dimensional vector in polar coordinates, where the radius is . ,horn In scope Choose any parameter , The infection distance threshold is given by parameter b, which is based on parameter b. ,vector The parameter D is a fixed value, and its value is... Step 1-1-2: For object O i Each sampling point of the trajectory data Select l groups of unit vectors, each group of unit vectors including m sub-vectors. … Each set of vectors Between each adjacent sub-vector All included angles ; Given the parameter δ, according to the parameter in step 1-1-1 The calculation formula can be used to calculate the parameter vector group. Corresponding parameter set , ,…, … Step 1-1-3: Based on object O i Trajectory dataset and parameter set , The projection function value matrix is ​​calculated. in, Indicates the first Trajectory data of one object O i Regarding the first Each object corresponds to a parameter group , , ,…, The projection function matrix, Represents trajectory data O j The corresponding number In the vector group, the first... Subvectors and parameters Corresponding projection function value For trajectory data O i The Middle The position vector corresponding to each sampling point; Step 1-1-4: Use a fixed-length encoding method to encode the projection function values. Encode the data and then concatenate the bits in a fixed order after the last encoded bit. The encoding, let the projection function value The encoded value is Construct trajectory data O i The encoding matrix S i Wherein, "||" is the encoded connector; Step 1-1-5: Use a collision-resistant one-way hash function to encode the matrix S. i Mapped to a secure index hash matrix in, .

2. The method for mining infection patterns in an outsourced cloud environment based on trajectory data according to claim 1, characterized in that: The privacy-preserving infection pattern mining phase includes the following steps: Step 2-1: Authorize users to number the infected individuals. Send the data to the cloud server and then wait for the server to return the mining results; Step 2-2: After receiving the infection pattern mining results returned in Step 2-1, the authorized user uses the key obtained in Step 1-2 to decrypt the user trajectory in the mining result set, and the infection pattern mining ends.

3. The method for mining infection patterns in an outsourced cloud environment based on trajectory data, as described in claim 2, is characterized in that: Step 2-1 includes the following steps: Step 2-1-1: The cloud server receives the infected object ID uploaded by the authorized user, and then processes the infection pattern mining results according to the infection event triple e=<origin, object, t> The data is stored in the form of an object, where origin is the source object ID and object represents the infected object ID. The first instance of an object being infected by an origin object. There are 10 time sampling points; if origin does not exist, then object is the initial source of infection object, i.e. P represents the set of infection events to be analyzed, and Final represents the set of confirmed infection events. The cloud server will authorize the user to upload the initial infection source object. Following the pattern of the infectious event triplet, add P and Final; Step 2-1-2: Examine each infection event in P. Let the infected object be e.object, given the infection distance threshold δ and the sliding time window threshold τ, the set of all objects is denoted as O, and the set of objects to be processed is denoted as O' = O-{e'.object | e' Final}, for each object o in O', if there exists a time window The infection event triple (e.object, o, T) will be generated if the following conditions are met. c Add Upgrate to .end: Condition 1: T c .begin≥et Condition 2: T c .begin T T c .end (Ha o [e.object][T]=Ha e.object [e.object][T]) Among them, T c .begin and T c .end respectively represent T c The occurrence times of the first and last sampling points within the time window. Indicates in All sampling points within the time window, Ha o Ha represents the hash matrix corresponding to object o. e.object This represents the hash matrix corresponding to the object e.object; Step 2-1-3: Update the candidate infection event set Select based on the Upgrate obtained in Step 2-1-2; Step 2-1-4: Determine if Select is empty. If Select is not empty, add the earliest occurrence of the infection event in Select to P, and also add it to Final, and go to step 2-1-2. If Select is empty, it indicates that the infection pattern mining is over. Final is the final determined infection pattern mining result set and is returned to the authorized user.

4. The method for mining infection patterns in an outsourced cloud environment based on trajectory data, as described in claim 3, is characterized in that: The update method in step 2-1-3 is as follows: Method 1: If there is an infection event in Select , and at the same time e.object = e'.object and e.time < e'.time are satisfied, then use to update such that e'.time = e.time and e'.source = e.source; Method 2: If a contagion event exists in Select. If both e.object = e'.object and e.time ≥ e'.time are satisfied, then it is ignored. ; Method 3: If there are no contagion events in Select. If e.object = e'.object, then the contagion event will occur. Add it directly to the Select list.

Images