Privacy protection smart medical service recommendation method based on weight matching

Abstract

The invention discloses a privacy protection smart medical service recommendation method based on weight matching. The method comprises the steps that a medical management center generates materials needed by a security key and sends the materials to a patient user and a server cluster; the medical management center carries out doctor registration service and patient user registration service through copy secret sharing and the server group; after registration is completed, each server in the server cluster obtains a pair of shared values; screening doctors by calculating a similarity vector between the user demand vector and the doctor attribute vector; and after the doctor service is completed, calculating the weight of each dimension of the score vector of the user, and updating the score of the doctor based on each dimension value of the score vector and the corresponding weight. According to the method, the most suitable doctor is recommended by calculating the similarity among the patient demand, the doctor attribute and the score vector; multi-party calculation of data is realized through a function secret sharing technology, and the data security is enhanced. Calculation and auxiliary servers are introduced, so that the online calculation cost is reduced, and malicious scoring is effectively resisted.

Description

Technical Field

[0001] The present invention belongs to the technical field of information security, and particularly relates to a privacy protection intelligent medical service recommendation method based on weight matching. Background Art

[0002] With the rapid development of the medical industry, various innovative technologies are changing the face of the medical industry, and the amount of medical data is also showing an explosive growth. From the wide application of electronic medical records and information management systems to the popularization of telemedicine services and mobile medical applications, the communication and cooperation between patients and medical staff have become more efficient and convenient. However, with the increase in electronic medical data, the issues of data security and privacy protection have become increasingly prominent. First of all, medical institutions need to strengthen network security measures to ensure the security and privacy of patients' medical records, examination reports, drug information, etc.; at the same time, doctors also face the challenge of quickly and accurately finding the most beneficial treatment plan for patients from these massive amounts of data; in addition, patients also face the problem of information asymmetry when choosing doctors, hospitals, and treatment plans. The proposed online medical recommendation system can effectively solve these problems. It provides an efficient doctor-patient matching service, thereby significantly improving the convenience of users and saving doctors' data processing time. However, this also raises serious privacy issues because sensitive health information such as users' medical information and medical treatment demands is vulnerable to unauthorized access and abuse. Existing privacy protection solutions still face huge challenges in achieving matching accuracy and efficiency. Summary of the Invention

[0003] To solve the above technical problems, the present invention proposes a privacy protection intelligent medical service recommendation method based on weight matching to solve the problems existing in the above prior art.

[0004] To achieve the above object, the present invention provides a privacy protection intelligent medical service recommendation method based on weight matching, including:

[0005] The medical management center generates the materials required for the security key and sends them to the patient user and the server group;

[0006] The medical management center performs doctor registration services and patient user registration services with the server group through replicated secret sharing; after the registration is completed, each server in the server group obtains a pair of shared values;

[0007] The patient user conducts identity verification through the medical management center. After the verification is passed, the medical management center sends the multi-point function secret sharing key to the patient user; based on the multi-point function secret sharing key and the user demand vector, an encrypted index vector is obtained; the server group calculates the similarity vector between the user demand vector and the doctor attribute vector based on the key evaluation algorithm, screens doctors based on the similarity vector, and sends the corresponding information to the patient user;

[0008] Obtain the scoring vector and decompose it. Combine two by two the three sharing shares obtained from the decomposition to obtain three pairs of shared values, which are held by two computing servers and an auxiliary server respectively; Based on protocol and protocol, calculate the weights of each dimension of the user scoring vector, and update the doctor's score based on the values of each dimension of the scoring vector and the corresponding weights.

[0009] Optionally, the process of the medical management center generating the materials required for the security key and sending them to the patient user and the server group includes:

[0010] The medical management center initializes the system through security parameters and generates a random t-point function; Generate two pairs of function secret sharing key pairs based on the security parameters and the random t-point function; After sending the two pairs of function secret sharing key pairs to the server group, make the pseudo-random function generated by the medical management center public.

[0011] Optionally, the process of sending the two pairs of function secret sharing key pairs to the server group includes:

[0012] Send the first bits of the two pairs of function secret sharing key pairs to the first computing server, send the last bits of the second pair to the second computing server, and send the first bit of the first pair and the last bit of the second pair to the auxiliary server.

[0013] Optionally, the process of the medical management center providing doctor registration services through replicated secret sharing with the server group includes:

[0014] The medical management center obtains the attribute information of each doctor and verifies its authenticity. If the verification passes, convert the attribute information of each doctor into vector form to obtain the doctor attribute vector; Decompose the doctor attribute vector into three sharing shares, combine the three sharing shares two by two to obtain three pairs of shared values, which are held by two computing servers and an auxiliary server respectively; Each server vertically stacks the sharing values of all doctors locally to obtain two vector matrices, completing doctor registration, where the sum of the three sharing shares is the doctor attribute vector.

[0015] Optionally, the process of the patient user registration service includes:

[0016] Based on the corresponding relationship between the patient user and the medical management center, obtain the pseudo-random function; Complete the registration by submitting personal information based on the pseudo-random function and the system prompt.

[0017] Optionally, the process of calculating the similarity between the user demand vector and the doctor attribute vector based on the key evaluation algorithm includes:

[0018] Each server in the server group obtains the corresponding function secret sharing key based on the index vector of the user demand vector; calculates the secret sharing share of the similarity between the doctor attribute vector and the user demand vector based on the function secret sharing key, the order vector, and the sharing share held by the server, and each server sums its respective secret sharing share to obtain the corresponding similarity vector.

[0019] Optionally, the auxiliary server sums the similarity vectors calculated by each server, sorts the obtained similarity values, and sends the doctor information that meets the similarity requirements to the patient user.

[0020] Optionally, the protocol includes: The medical management center generates a random number, and through replicated secret sharing, generates three non-zero additive sharing shares and three multiplicative sharing shares; after pairwise combining the three non-zero additive sharing shares and sending them to two computing servers and an auxiliary server respectively with the corresponding multiplicative sharing shares; after each server calculates the product of the held scoring vector and the random number share, the first computing server and the auxiliary server exchange data and each recover the true value; the computing server and the auxiliary calculator respectively calculate the reciprocals of the two multiplicative sharing shares they hold, the first computing server and the auxiliary server update the reciprocals of the multiplicative sharing shares through the true value, and the computing server and the auxiliary server take the logarithm of the absolute value of the updated reciprocals of the multiplicative sharing shares; where the product of the three multiplicative sharing shares is the random number.

[0021] Optionally, the protocol includes:

[0022] The computing server and the auxiliary server respectively use ∏ mul ( ) to process the sharing shares of the held scoring vector and the sharing shares of all doctor attribute vectors, and then through the protocol for processing; each server in the server group discloses its local share, and each server reconstructs after receiving the missing sharing shares and takes the logarithm of the obtained result; each server respectively calculates the replicated secret sharing share of the weight of each patient's score, and the first computing server and the auxiliary server update the replicated secret sharing share through the logarithm of the sum of the encrypted local shares by replicated secret sharing; each server calculates the sum of the weights of all patient scores locally.

[0023] Optionally, the process of updating the doctor score based on the values of each dimension of the scoring vector and the corresponding weight includes:

[0024] Construct three groups of non-zero shared values. For each group of non-zero shared values, combine two of the three sharing shares pairwise to obtain three pairs of shared values, which are held by two computing servers and one auxiliary server respectively; each server performs the ∏ mul ( ) operation on the replicated secret sharing share of each patient's weight held by it and the replicated secret sharing share of the scoring vector corresponding to the patient, and then sum the results to obtain the replicated secret sharing share of the weighted sum of scores; the server obtains the updated doctor score based on the sum of the replicated secret sharing share of the weighted sum of scores held by it and the replicated secret sharing shares of all patients' weights.

[0025] Compared with the prior art, the present invention has the following advantages and technical effects:

[0026] Compared with the existing privacy protection medical recommendation scheme, the present invention introduces a doctor attribute scoring vector. In the recommendation stage, by calculating the similarity among the patient's needs, the doctor's attribute vector, and the doctor's attribute scoring vector, and using this similarity as an index to recommend the optimal doctor for the patient; the present invention performs doctor-patient matching through the similarity between the doctor and the patient. Compared with the similarity calculation methods used in other schemes, the communication overhead between the patient and the server in the present invention is significantly reduced; the present invention optimizes the part related to communication calculation in doctor-patient matching, realizes multi-party calculation of data through function secret sharing, and introduces two types of servers, namely computing servers and auxiliary servers, to reduce the online calculation overhead. It can not only effectively protect the data security of patients and doctors, but also more effectively resist malicious scoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] The drawings constituting a part of this application are used to provide a further understanding of this application. The schematic embodiments of this application and their descriptions are used to explain this application and do not constitute an improper limitation to this application. In the drawings:

[0028] Figure 1 is a schematic diagram of realizing data encryption by means of FSS and RSS in an embodiment of the present invention;

[0029] Figure 2 is a schematic diagram of a privacy protection doctor-patient matching model in an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0030] It should be noted that, without conflict, the embodiments in this application and the features in the embodiments can be combined with each other. The following will refer to the drawings and combine the embodiments to detail this application.

[0031] It should be noted that the steps shown in the flowchart of the drawings can be executed in a computer system such as a set of computer executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0032] Example 1

[0033] As Figure 1-2 shown, in this example, a privacy - protected intelligent medical service recommendation method based on weight matching is provided, including:

[0034] Symbols and definitions:

[0035] U i : represents the i - th patient user;

[0036] D i : represents the i - th doctor;

[0037] represents the demand vector of patient user U i ;

[0038] represents the attribute vector of doctor D i ;

[0039] represents the attribute score vector of doctor D i ;

[0040] L: represents the dimension of the attribute / demand vector;

[0041] r1, r2: randomly selected;

[0042] represents the sharing share held by server CSP i ;

[0043] CSP1, CSP2, SP: two computing servers CSP and an auxiliary server SP;

[0044] an ordered vector defined as {1, 2, …, L};

[0045] DMPF.Gen(1 λ , f): For a given security parameter 1 λ and function f, the key generation algorithm DMPF.Gen() can output a pair of keys.

[0046] For a given security parameter 1 λ and function f, the key generation algorithm DMPF.Gen() can output two keys and

[0047] The key evaluation algorithm MPF.Eval() can evaluate the parsing key corresponding to the given participant index b Evaluate with the input string x and obtain the result. Eval() is the execution algorithm of the multi-point function secret sharing protocol.

[0048] For the parsing key corresponding to the given participant index b and the input string x, the key evaluation algorithm MPF.Eval() can output

[0049] ∑ mul : For two replicated secret-shared values x and y, through ∑ mul the sharing of x·y can be obtained.

[0050] (1) In the system initialization stage:

[0051] In this stage, the Healthcare Management Center (HMC) generates the materials required for the security key and sends them to the patient user U i and the server group CSPs. The Healthcare Management Center initializes the system with the security parameter k and performs the following operations:

[0052] Step 1.1.1: The HMC generates a random t-point function where i ∈ [t].

[0053] Step 1.1.2: The HMC locally executes the DMPF.Gen(1 k , f) algorithm twice to generate two pairs of function secret sharing key pairs and

[0054] Step 1.2: The HMC generates a pseudorandom function F: {0, 1} k × {0, 1} k → {0, 1}, which is used to generate zero sharing.

[0055] Step 1.3: The HMC sends Figure 1 as shown to CSP1 respectively, sends to CSP2, and sends to SP. Subsequently, the HMC makes the pseudorandom function F: {0, 1} × {0, 1} k × {0, 1} k → {0, 1} public.

[0056] (2) In the user registration stage:

[0057] In this stage, the Healthcare Management Center provides registration services for doctors and users respectively. Specifically as follows:

[0058] 1) Doctor registration service:

[0059] For each doctor D j ∈ D, he needs to submit his attribute information to the HMC. After the HMC verifies the authenticity of his information, if it passes, the attribute information will be converted into a vector form, denoted as the attribute vector where b il = {0, 1, …, 2 n - 1}.

[0060] Then the HMC encrypts and shares the doctor's attribute vector through replicated secret sharing, that is, for each doctor's attribute vector the HMC decomposes it into three sharing shares satisfying Then as Figure 1 shown, each server obtains a pair of shared values: CSP1 holds CSP2 holds SP holds Subsequently, the servers (CSP1 and CSP2) and the auxiliary server SP vertically stack all doctors' attribute vectors locally to obtain two vector matrices, that is, CSP1 holds CSP2 holds SP holds

[0061] 2) Patient user registration service:

[0062] For each patient user U i ∈ U, establish a corresponding relationship with the HMC to obtain the function f. Then provide the required personal information according to the system prompt for user registration.

[0063] (3) In the doctor recommendation stage:

[0064] In this stage, the auxiliary server SP retrieves suitable doctors for each patient according to the requirements submitted by each user, and provides a doctor recommendation list sorted in descending order of the recommendation score to the patients. This stage can be divided into the following three steps:

[0065] 1) User requirement submission:

[0066] Before the user submits his requirements, he needs to verify his identity with the HMC. After the identity verification passes, the HMC will send the multi-point function secret sharing key to the user

[0067] Subsequently, the patient U i preprocesses his requirement vector and converts it into an integer requirement vector, denoted as where a il ∈ {0, 1, …, 2 n - 1}, and ail The value represents the degree of the patient's need for this attribute (or the severity of the symptom). The larger the value, the higher the degree of need (the more severe).

[0068] Then, through the obtained multi-point function secret sharing key the indices and values of the demand vector are permutation encrypted (the index represents the position), and the encrypted index vector is denoted as where a il = f(a' il ). Then the user sends the encrypted vector to the server group CSPs.

[0069] 2) Attribute-demand similarity calculation:

[0070] First, the server group CSPs calculate the similarity of the index vector of the demand vector through the multi-point function secret sharing DMPF and the function secret sharing FSS technology. For each doctor attribute vector the server obtains the corresponding function secret sharing key K with the index vector of the demand vector as a parameter, and then multiplies the calculation result by b jl , and then calculates where s represents the s-th server participating in the calculation.

[0071] CSP i After receiving the patient's demand vector, calculates the similarity between the demand vector of patient U i and the attribute vector of each doctor D j . For each doctor D j , j ∈ {1, 2,..., β}, the server CSP i obtains the secret sharing share of the similarity between this doctor and the patient's demand vector by performing the following operations

[0072] At the CSP1 side: Locally calculate Then calculate to obtain

[0073] At the CSP2 side: Locally calculate Then calculate to obtain

[0074] At the SP side: Locally calculate Then calculate to obtain

[0075] 3) Doctor screening:

[0076] At this stage, the computing server transmits the local similarity vector to the auxiliary server, enabling the auxiliary server to recover the true value of the similarity. Then the auxiliary server sorts the similarities from high to low and sends the information of the top k doctors with the highest similarities to the patient U i . The specific operations are as follows:

[0077] At the CSP1 side: Transmit to the SP

[0078] At the CSP2 side: Transmit to the SP

[0079] At the SP side: After receiving and , calculate and select the top k doctors with the highest similarities and send them to the patient U i .

[0080] (4) Patient scoring stage

[0081] After doctor D j diagnoses user U i and provides medical services, user U i can score the service to evaluate the service quality provided by the doctor. Subsequently, the user sends the scoring vector to the server group CSPs, and the server group updates the attribute scores of the doctor based on this user score. This stage is mainly divided into three steps:

[0082] 1): Submission of user score

[0083] The user generates a scoring vector according to the score and then generates three sharing shares through the replicated secret sharing technology satisfying Send the share to CSP1, send to CSP2, send to SP

[0084] 2) Privacy-preserving calculation of doctor reputation update

[0085] In the process of updating the doctor attribute scores, it is necessary to protect not only the latest feedback scores of the patients but also the weight values of various data. First, it is necessary to calculate the data weight where represents the weighted distance between the updated value and the initial true value, and std represents the standard deviation between (v1, v2,..., v K ).

[0086] It can be seen that the log() function is involved in the process of calculating data weights. A protocol is designed, and the implementation steps are as follows:

[0087] Initially: CSP1 holds CSP2 holds SP holds

[0088] Step 1.1: HMC generates a random number c and generates three non-zero additive sharing shares through replicated secret sharing and three multiplicative sharing shares <c> i , satisfying c = <c> 1· <c> 2· <c>3. Send to CSP1, send to CSP2, and send to SP.

[0089] Step 1.2: Using and as inputs, CSP1 calculates mul through the multiplication component ∑ and obtains Similarly, CSP2 obtains SP obtains

[0090] Step 1.3: CSP1 and SP exchange the values of and and each restores the true values

[0091] Step 1.4: CSP1 locally calculates Similarly, CSP2 locally calculates and obtains <x>2 and <x>3. Locally calculated by SP <x>3 and <x>1.

[0092] Step 1.5: CSP1 and SP each perform local updates <x> 1＝ <x>1·xc。

[0093] Step 1.6: Local calculation by CSP1 Similarly, local calculation by CSP2 yields and Local calculation by SP yields and

[0094] Finally: CSP1 obtains and CSP2 obtains and SP obtains and

[0095] Through the above protocol, the difficult-to-process log() function in privacy-preserving reputation update calculation can be implemented. Then construct a protocol to achieve multi-party secure data weight calculation, so that each participating party can finally obtain the replicated secret sharing share of the data weight ω k . The specific implementation steps are as follows:

[0096] Initially: CSP1 holds the score x k of patient U k and the sharing share of the doctor attribute vector

[0097] Step 2.1: CSP1 calculates Similarly, CSP2 calculates to obtain and SP calculates to obtain and ∏ mul The calculation rule of ( ) is explained as follows: ∏ mul (x1, x2, y1, y2) = x1 * y1 + x1 * y2 + x2 * y1 + x2 * x2.

[0098] Step 2.2: CSP1 calculates Similarly, CSP2 calculates to obtain SP calculates to obtain

[0099] Step 2.3.1: Each participating party CSPs publicly disclose their local shares and

[0100] Step 2.3.2: After each participating party CSPs receive the missing shares, each party reconstructs to obtain The true value of and locally calculated by each party

[0101] Step 2.4: CSP1 calculates Similarly, CSP2 obtains SP obtains

[0102] Step 2.5: CSP1 and SP perform local calculations respectively

[0103] Step 2.6: Each participating party CSPs performs local calculations n represents the total number of patients who score the doctor.

[0104] Finally: CSP1 obtains CSP2 obtains SP obtains where k represents the k-th patient who scores, and 1 represents the calculation result of the first server: represents the weight of the k-th patient's score, represents a replicated secret sharing share, owned / calculated by the first server.

[0105] 3) True value evaluation session

[0106] After each server obtains the value of ω(x i ) in the second step, it is necessary to further calculate and update the evaluation. For the new evaluation values (v1, v2,..., v K ) given by the user and the corresponding weights (ω1, ω2,..., ω K ), the updated evaluation can be obtained through the formula We designed protocol to achieve the secure calculation of the new evaluation, so as to achieve the update of the doctor's score. The implementation steps are as follows: Initially: Each server holds And the server constructs three pairs of non-zero shares {a1, a2, a3}, {b1, b2, b3} and {c1, c2, c3}, satisfying a1 + a2 + a3 = 0, b1 + b2 + b3 = 0, c1 + c2 + c3 = 0, and each party holds {a i , a i+1 , b i , b i+1 , c i , c i+1}.

[0107] Step 3.1: CSP1 performs local calculation Obtain the result where k = 1, …, K. Similarly, CSP2 calculates SP calculates

[0108] Step 3.2: CSP1 locally calculates the replicated secret sharing shares of the scoring weighted sum, Similarly, CSP2 calculates and SP calculates and

[0109] Step 3.3: CSP1 calculates n represents the number of patients. Similarly, CSP2 calculates and SP calculates and

[0110] Finally: CSP1 obtains CSP2 obtains SP obtains

[0111] Different from the traditional method that uses simple boolean vectors to represent whether a doctor has a certain skill or a patient has a relevant appeal, the present invention introduces a generalized weight description mechanism. The professional characteristics of doctors and the demand information of users are respectively represented by more refined weights. Among them, the professional levels of doctors in different fields are distinguished by multiple levels (such as level 0 to level N), and patients assign weights to different attributes of doctors according to their own needs. For example, if a patient pays more attention to the gender of a doctor or the expertise in psychotherapy, a higher weight can be assigned; while if the demand for gastrointestinal diseases is lower, the weight is correspondingly reduced. Through this mechanism, more accurate and personalized matching between doctors and patients can be achieved, effectively improving the recommendation quality of intelligent medical services.

[0112] In addition, to ensure the security of users' sensitive information, during the online recommendation stage, the present invention performs privacy protection processing on both the attribute information of doctors and the pathological information of patients. Specifically, by adopting secret sharing and distributed multi-point function technologies, the privacy of sensitive data is effectively protected. At the same time, based on the Secure Multi-Party Computation (SMPC) technology, online matching between doctors and patients under privacy protection and dynamic update evaluation of doctors' reputation values are realized. While ensuring data security, this method significantly reduces the computational and communication overheads in the online doctor-patient matching process, and further improves the accuracy and efficiency of matching, providing an efficient solution for privacy-protected intelligent medical services.

[0113] The above are only the preferred specific embodiments of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present application should be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.< / x> < / x> < / x> < / x> < / x> < / x> < / c> < / c> < / c> < / c>

Claims

1. A privacy-preserving smart medical service recommendation method based on weight matching, characterized in that: The following steps are involved: The medical management center generates the materials required for the security key and sends them to the patient user and the server group; The medical management center provides doctor registration services and patient user registration services through replicated secret sharing with the server group; After registration is completed, each server in the server group obtains a pair of shared values; The patient user passes the identity authentication through the medical management center. After the authentication is passed, the medical management center sends the multi-point function secret sharing key to the patient user; an encrypted index vector is obtained based on the multi-point function secret sharing key and the user demand vector; after the server group encrypts the index of the encrypted demand vector, the similarity vector between the user demand vector and the doctor attribute vector is calculated based on the secret key evaluation algorithm, and the doctor is selected based on the similarity vector and the corresponding information is sent to the patient user; Get the scoring vector and decompose it. Combine the three shared shares obtained by the decomposition into three pairs of shared values, which are held by two computing servers and one auxiliary server respectively. Agreement and The protocol calculates the weights of each dimension of the user rating vector and updates the doctor's rating based on the values ​​of each dimension of the rating vector and the corresponding weights.

2. The privacy-preserving intelligent medical service recommendation method based on weight matching according to claim 1 is characterized in that: The process by which the medical management center generates the materials required for the security key and sends them to the patient user and the server group includes: The medical management center initializes the system through security parameters and generates a random t-point function; generates two pairs of function secret sharing key pairs based on the security parameters and the random t-point function; after sending the two pairs of function secret sharing key pairs to the server group, the pseudo-random function generated by the medical management center is made public.

3. The privacy protection intelligent medical service recommendation method based on weight matching according to claim 2 is characterized in that: The process of sending the two pairs of function secret shared key pairs to the server farm includes: The first digit of the two pairs of function secret shared key pairs is sent to the first computing server, the last digit of the second pair is sent to the second computing server, and the last digit of the first pair and the first digit of the second pair are sent to the auxiliary server.

4. The privacy protection intelligent medical service recommendation method based on weight matching according to claim 1 is characterized in that: The process of the medical management center providing doctor registration services through replicated secret sharing with the server cluster includes: The medical management center obtains the attribute information of each doctor and verifies its authenticity. If the verification is successful, the attribute information of each doctor is converted into a vector form to obtain a doctor attribute vector; the doctor attribute vector is decomposed into three sharing shares, and the three sharing shares are combined in pairs to obtain three pairs of sharing values, which are respectively held by two computing servers and one auxiliary server; each server locally vertically superimposes the sharing values ​​of all doctors to obtain two vector matrices to complete the doctor registration, where the sum of the three sharing shares is the doctor attribute vector.

5. The privacy-preserving intelligent medical service recommendation method based on weight matching according to claim 1 is characterized in that: The process of patient user registration service includes: Based on the correspondence between the patient user and the medical management center, a pseudo-random function is obtained; based on the pseudo-random function and the system prompt, personal information is submitted to complete the registration.

6. The privacy-preserving intelligent medical service recommendation method based on weight matching according to claim 4 is characterized in that: The process of calculating the similarity between the user demand vector and the doctor attribute vector based on the secret key evaluation algorithm includes: Each server in the server group obtains the corresponding function secret sharing key based on the index vector of the user demand vector; the secret sharing share of the similarity between the doctor attribute vector and the user demand vector is calculated based on the function secret sharing key, the sequence vector and the sharing share held by the server, and each server sums its own secret sharing share to obtain the corresponding similarity vector.

7. The privacy-preserving intelligent medical service recommendation method based on weight matching according to claim 1 is characterized in that: The auxiliary server sums up the similarity vectors calculated by each server, sorts the similarity values ​​obtained by the sum, and selects the doctor information with similarity that meets the requirements to send to the patient user.

8. The privacy-preserving intelligent medical service recommendation method based on weight matching according to claim 1 is characterized in that: Said The protocol includes: the medical management center generates a random number, generates three non-zero addition sharing shares and three multiplication sharing shares by replicating secret sharing; the three non-zero addition sharing shares are combined in pairs and sent to two computing servers and an auxiliary server with the corresponding multiplication sharing shares respectively; after each server calculates the product of the score vector held and the random number share, the first computing server and the auxiliary server exchange data and restore the true value respectively; the computing server and the auxiliary calculator respectively calculate the reciprocal of the two multiplication sharing shares held by themselves, the first computing server and the auxiliary server update the reciprocal of the multiplication sharing share by the true value, and the computing server and the auxiliary server take the logarithm of the absolute value of the reciprocal of the updated multiplication sharing share; wherein, the product of the three multiplication sharing shares is the random number.

9. The privacy-preserving intelligent medical service recommendation method based on weight matching according to claim 1 is characterized in that: Said The agreement includes: The computing server and auxiliary server use Π mul () After processing the share of the score vector held and the share of all doctor attribute vectors, Protocol processing; each server in the server group discloses its local share, and each server reconstructs the missing shared share after receiving it, taking the logarithm of the reconstructed result; each server calculates the replicated secret sharing share of the weight of each patient's score, and the first calculation server and the auxiliary server update the replicated secret sharing share by the logarithm of the sum of the local shares encrypted by the replicated secret sharing; each server locally calculates the sum of the weights of all patient scores.

10. The privacy-preserving intelligent medical service recommendation method based on weight matching according to claim 1 is characterized in that: The process of updating the doctor's score based on the values ​​of each dimension of the score vector and the corresponding weights includes: Construct three groups of non-zero shared values, and combine the three shared shares of each group of non-zero shared values ​​in pairs to obtain three pairs of shared values, which are held by two computing servers and one auxiliary server respectively; each server will combine the replicated secret shared share of each patient weight it holds with the replicated secret shared share of the score vector corresponding to the patient mul () The sum is calculated after the operation to obtain the replicated secret sharing share of the weighted sum of the scores; the server obtains the updated doctor score based on the sum of the replicated secret sharing share of the weighted sum of the scores it holds and the replicated secret sharing shares of all patient weights.

Images