Privacy protection mental health data truth value discovery algorithm supporting outlier data filtering

Abstract

The invention relates to the field of network security, and provides a privacy protection mental health data truth value discovery algorithm supporting outlier data filtering. The algorithm aims to solve the outlier data problem existing in the psychological health data acquisition process, and meanwhile, the privacy of a data provider is protected. The algorithm mainly comprises the steps of parameter initialization, psychological health data ciphertext uploading, range detection proof generation, outlier data filtering, decryption key generation, psychological health data truth value updating and data provider weight updating. Through cooperation of a key generation center, a psychological analysis platform and a psychological health data provider, an algorithm filters outlier data in an encryption state, deduces a normal level of a psychological health index, and updates a weight of the data provider until a convergence condition is reached. According to the algorithm, data privacy is effectively protected, and the accuracy of mental health data analysis is improved.

Description

Technical Field

[0001] The invention belongs to the field of network security and proposes a privacy-preserving mental health data truth value discovery algorithm that supports outlier data filtering. Background Art

[0002] In today's digital age, mental health issues are receiving increasing attention, and the collection and analysis of mental health data is of great significance for understanding the individual's mental state and formulating intervention measures. Mental health data covers many aspects such as individual emotional expression, behavioral patterns, and psychological test results. These data can not only reflect the individual's current mental state, but also provide a basis for predicting potential psychological risks. With the rapid development of artificial intelligence technology, its application in the field of mental health is becoming more and more extensive, such as evaluating the user's mental health by analyzing text data on social media. However, the accuracy of the artificial intelligence model depends on the accuracy of the training data and the accuracy of a certain mental health indicator benchmark. The determination of the benchmark will directly lead to the deviation of the analysis results.

[0003] However, in the actual process of collecting mental health data, outlier data is often generated due to various factors such as individual differences, the complexity of the data collection environment, and interference during data transmission. Outlier data refers to data points that are significantly different from normal data. They may be caused by incorrect data records, abnormal psychological state performance, etc. The existence of outlier data will first make it difficult for the psychological analysis platform to find the baseline of a certain mental health indicator. Secondly, even if it is found, it will deviate from the normal range and obtain a baseline with a large error, which will lead to inaccurate predictions of the artificial intelligence model.

[0004] To this end, experts and scholars in the field of data science have proposed a series of methods such as data cleaning and outlier data filtering, which can effectively solve the problem of low data quality and thus improve the accuracy of model training. However, mental health data is different from other types of data (such as environmental indicators, height, etc.). This type of data will involve the personal privacy of the provider, such as sensitive information such as mental illness history and emotional distress. However, most existing studies have not considered the issue of protecting the privacy of mental health data.

[0005] Therefore, in order to solve the above problems, the present invention proposes a privacy-preserving mental health data truth discovery algorithm that supports outlier data filtering. This method can filter outlier mental health data in an encrypted state without leaking the privacy of the mental health data of the data source, and infer the normal level of a certain mental health indicator based on the filtered data, and this level is only known to the psychological analysis platform. Summary of the invention

[0006] The present invention provides a privacy-preserving mental health data truth value discovery algorithm that supports outlier data filtering. The algorithm does not require the full participation of mental health data providers, and uses function encryption and masking mechanisms to complete the filtering of outlier mental health data and infer the normal level of a certain mental health indicator without leaking the privacy of the original mental health data.

[0007] In order to achieve the above-mentioned purpose, the present invention adopts the following technical means:

[0008] The present invention provides a privacy-preserving mental health data truth value discovery algorithm supporting outlier data filtering, comprising the following steps:

[0009] Step 1: Parameter initialization: The key generation center KGC selects security parameters, generates system public parameters and keys of mental health data providers;

[0010] Step 2: Upload encrypted mental health data: Each mental health data provider encrypts their own mental health data and uploads it to the psychological analysis platform CS;

[0011] Step 3: Generation of range detection proof: The psychological analysis platform CS selects a challenge and sends it to the mental health data provider. After receiving the challenge, the mental health data provider generates a mental health data range proof and sends it to the psychological analysis platform CS.

[0012] Step 4: Filtering of outlier mental health data: The psychological analysis platform CS performs range detection on the original mental health data based on the ciphertext of the mental health data provider and the mental health data range proof, filters the outlier data, and sends the detection results to the key generation center KGC to request a decryption key;

[0013] Step 5, Decryption key generation: The key generation center KGC generates a decryption key based on the detection results and sends it to the psychological analysis platform;

[0014] Step 6, update the true value of mental health data: For the same type of mental health data, the psychological analysis platform CS aggregates and decrypts the ciphertext according to the decryption key to update the true value of the mental health data, and updates the auxiliary information of the weight update of the mental health data provider according to the new round of the true value of the mental health data, and sends it to the key generation center KGC;

[0015] Step 7. Update weights of mental health data providers: The key generation center updates the weights of mental health data providers based on the weight update auxiliary information sent by the psychological analysis platform.

[0016] Repeat steps 5 to 7 until convergence is reached.

[0017] In the above algorithm, step 1 includes:

[0018] Step 1.1: Given the security parameter K, the number of mental health data providers K, and the type of mental health data to be collected M, the key generation center KGC randomly selects two secure large prime numbers p, q←SP(κ), calculates N=pq, and then randomly selects Calculate g = (g′) 2N mod N 2 Set the boundary later Hash function H: The system public parameters are pp = (N, g, X, Y, H), where Represents modulo N 2 The integer ring, that is, all integers from 0 to N 2 The set of integers from -1;

[0019] Step 1.2: Based on the system public parameter pp, the key generation center KGC generates a Integer Gaussian distribution with mean 0 Randomly select a K-dimensional vector (s 1 ,s 2 , ..., s K ) as the master private key msk, and each scalar s of msk k As the encryption key of the mental health data provider, the corresponding public key is calculated as And send it to the corresponding mental health data provider through a secure channel. The key generation center KGC is open

[0020] Step 2 specifically includes the following steps:

[0021] Step 2.1. Each mental health data provider p k , where k∈{1,...,K}, and M types of mental health data are Use the pseudo-random number generator R to select a mask for each data Then calculate the randomized perception data where m∈{1,...,M};

[0022] Step 2.2, then generate the ciphertext of each mental health data, record the ciphertext as Send the ciphertext to the psychological analysis platform CS.

[0023] Step 3 specifically includes the following steps:

[0024] Step 3.1: For each mental health data provider p k , psychological analysis platform CS selection challenge C k =(C k，1 , C k,2, ..., C k，m ..., C k，M ) is sent to the corresponding mental health data provider, where C k，m different from each other, k∈{1,...,K}, m∈{1,...,M}, C k，M represents the challenge value of the k-th mental health data provider on the M-th category of mental health data;

[0025] Step 3.2: After receiving the challenge, the mental health data provider calculates the range detection proof (va k , vb k ) and send it to the psychological analysis platform CS.

[0026] Step 4 specifically includes the following steps:

[0027] Step 4.1, the psychological analysis platform CS performs range detection on the original mental health data based on the ciphertext of the mental health data provider and the mental health data range proof, filters outlier data, and the filtered mental health data provider set is recorded as Send it to the key generation center KGC.

[0028] Step 4.1: For each mental health data provider p k , the psychological analysis platform CS is based on the ciphertext of the mental health data provider and the mental health data calculation range proof (va′ k , vb′ k );

[0029] Step 4.2: Perform range detection on the original mental health data to verify va k =va′ k , vb k =vb′ k If all of them are true, then the data of the mental health data provider is discarded, and the filtered mental health data provider set is recorded as That is, the detection result is obtained and then sent to the key generation center KGC to request the decryption key.

[0030] Step 5 specifically includes the following steps:

[0031] Step 5.1: The key generation center KGC generates the decryption key And send it to the psychological analysis platform CS.

[0032] Step 6 specifically includes the following steps:

[0033] Step 6.1: The psychological analysis platform CS aggregates and decrypts the ciphertext according to the decryption key, updates the mental health data and obtains the aggregation result of this round of mental health data.

[0034] Step 6.2: Calculate the auxiliary information UW of the weight update of the mental health data provider and send it to the key generation center KGC.

[0035] Step 7 specifically includes the following steps:

[0036] Step 7.1: The key generation center KGC updates the weight ω of each mental health data provider k .

[0037] Because the present invention adopts the above technical solution, it has the following beneficial effects:

[0038] 1. The problem of outlier data interference in the process of mental health data collection is solved by means of encrypted upload and range detection proof generation. The mental health data provider encrypts the data and uploads it. The psychological analysis platform generates a range detection proof through a challenge-response mechanism to ensure that the data is range-checked in an encrypted state and outlier data that does not meet the normal range is filtered out, thereby improving the accuracy and reliability of the data.

[0039] 2. The problem of privacy leakage of mental health data is solved by filtering outlier data and generating decryption keys. The psychological analysis platform filters outlier data in an encrypted state and requests decryption keys from the key generation center to ensure that the data is always encrypted before decryption, protecting the privacy of data providers while ensuring the transparency and security of data processing.

[0040] 3. The problem of inaccurate mental health data benchmarks has been solved by updating the true value and weight of mental health data. The psychological analysis platform aggregates and decrypts ciphertexts according to the decryption key, updates the true value of mental health data, and updates the weight of data providers based on the true value, ensuring that the data benchmark is more accurate and improving the reliability of mental health data analysis.

[0041] 4. The problem of dynamic adjustment of data provider weights is solved by means of iterative updates and convergence conditions. The algorithm updates the true value of the data and the weight of the data provider through multiple iterations until the convergence conditions are reached, ensuring that the weight of the data provider can dynamically reflect the reliability of its data, further improving the accuracy and stability of mental health data analysis.

[0042] 5. Through the collaboration between the key generation center and the psychological analysis platform, the balance problem between data privacy protection and data analysis efficiency is solved. The key generation center is responsible for key generation and distribution, and the psychological analysis platform is responsible for data processing and analysis. The collaboration between the two ensures that data is processed efficiently under the premise of privacy protection, which improves the overall efficiency of mental health data analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 It is a flow chart of the present invention.

[0044] 1-The key generation center initializes system parameters and completes key distribution.

[0045] 2-Mental health data providers encrypt their own data and complete the mental health data report upload.

[0046] 3-The psychological analysis platform and the mental health data provider use a challenge-response approach to generate range detection proofs.

[0047] 4-The psychological analysis platform performs outlier psychological health data filtering and sends the detection results to the key generation center to request a decryption key.

[0048] 5-The key generation center generates a decryption key and sends it to the psychological analysis platform.

[0049] 6-The psychological analysis platform completes the update of each type of mental health data indicator, calculates and sends weight update auxiliary information.

[0050] 7-The key generation center completes the weight update of the mental health data provider. DETAILED DESCRIPTION

[0051] The following will provide a detailed description of the implementation of the present invention. Although the present invention will be described and illustrated in conjunction with some specific embodiments, it should be noted that the present invention is not limited to these embodiments. On the contrary, modifications or equivalent substitutions made to the present invention should all be included in the scope of the claims of the present invention.

[0052] In addition, in order to better illustrate the present invention, numerous specific details are given in the following specific embodiments. It will be understood by those skilled in the art that the present invention can also be implemented without these specific details.

[0053] The detailed algorithm flow is:

[0054] Step 1 parameter initialization:

[0055] Step 1.1: Given the security parameter K, the number of mental health data providers K, and the type of mental health data to be collected M, the key generation center KGC randomly selects two secure large prime numbers p, q←SP(κ), calculates N=pq, and then randomly selects Calculate g = ( g ′) 2N mod N 2 Set the boundary later Hash function H: The system public parameters are pp = (N, g, X, Y, H);

[0056] Step 1.2: Based on the system public parameter pp, the key generation center KGC generates a Integer Gaussian distribution with mean 0 Randomly select a K-dimensional vector (s 1 ,s 2 , ..., s K ) as the master private key msk, and each scalar s of msk k As the encryption key of the mental health data provider, the corresponding public key is calculated as And send it to the corresponding mental health data provider through a secure channel. The key generation center KGC is open

[0057] Step 2 Upload the encrypted mental health data:

[0058] Step 2.1 Each mental health data provider p k All hold M-type mental health data Where k∈{1,...,K}. First, for each mental health data Use the pseudo-random number generator R to select the mask Then calculate the randomized perception data where m∈{1,...,M}, we get

[0059] Step 2.2 is to encrypt each mental health data Mental health data provider k Random Selection ,calculate where u m =H(l m ), l m A label for a certain type of mental health indicator, such as sleep duration. And upload it to the psychological analysis platform CS.

[0060] Step 3 Range detection proof generation:

[0061] Step 3.1: For each mental health data provider p k , psychological analysis platform CS selection challenge C k =(C k，1 , C k，2 , ..., C k，m , ..., C k，M ) and the normal range of M-type mental health indicators Sent to the corresponding mental health data provider, where C k，mare different from each other, k ∈ {1,..., K}, m ∈ {1,..., M}, a m ≤ b m ;

[0062] Step 3.2. After receiving the challenge, the mental health data provider p k calculates the range detection proof (va k , vb k ), as shown in formula (1), and sends it to the psychological analysis platform CS.

[0063]

[0064] where h k is the public key of the mental health data provider p k .

[0065] Step 4. Outlier mental health data filtering:

[0066] Step 4.1. For each mental health data provider p k (k ∈ {1,..., K}), the psychological analysis platform CS calculates (va′ , vb′ k , vb′ k ) according to the existing ciphertexts according to formula (2).

[0067]

[0068] The correctness of formula (2) is as follows:

[0069]

[0070] Then check whether va k = va′ k , vb k = vb′ k both hold. If there is a non - conforming situation, discard the data of this mental health data provider, and denote the set of finally - passed - inspection mental health providers as Send it to the key generation center KGC to request the decryption key.

[0071] Step 5. Decryption key generation:

[0072] Step 5.1. The key generation center KGC calculates Calculate for each type of mental health indicator where m ∈ {1,..., M}, and send to the psychological analysis platform CS, where represents y k the relative weight of the k - th mental health data provider, ω krepresents the weight of the k-th mental health data provider.

[0073] Step 6: Update the true value of mental health data:

[0074] Step 6.1: For each type of mental health indicator l m , Psychological Analysis Platform CS Computing Then calculate u m =H(l m ), and finally get the inferred true value of each type of mental health data The true value results of all categories of mental health data are recorded as

[0075] Step 6.2: Then, in order to update the weight of each mental health data provider for the next round, the psychological analysis platform CS calculates the auxiliary information for weight update And send it to the key generation center KGC.

[0076] Step 7 Mental Health Data Provider Weight Update:

[0077] Step 7.1: After receiving UW, the key generation center KGC calculates

[0078] Then for each mental health data provider pk, Update its reliability information to in express elements.

[0079] Repeat steps 5 to 7 until convergence conditions are met.

Claims

1. A privacy-preserving mental health data truth discovery algorithm that supports outlier data filtering, characterized by: Including the key generation center KGC, the psychological analysis platform CS and K mental health data providers P k , where k∈{1, ..., K}, the psychological analysis platform CS is responsible for issuing M different mental health data collection tasks, and the mental health data provider is responsible for providing the mental health data held by the individual. The psychological analysis platform CS analyzes the normal level of the mental health indicators (such as the number of meals, sleep duration, number of pessimistic vocabulary expressions, etc.) of the mental health data provider, including the following steps: Step 1: Parameter initialization: The key generation center KGC selects security parameters, generates system public parameters and keys of mental health data providers; Step 2: Upload encrypted mental health data: Each mental health data provider encrypts their own mental health data and uploads it to the psychological analysis platform CS; Step 3: Generate range detection proof: The psychological analysis platform CS selects a challenge and sends it to the mental health data provider. After receiving the challenge, the mental health data provider generates a mental health data range proof and sends it to the psychological analysis platform CS. Step 4: Filtering of outlier mental health data: The psychological analysis platform CS performs range detection on the original mental health data based on the ciphertext of the mental health data provider and the mental health data range proof, filters the outlier data, and sends the detection results to the key generation center KGC to request a decryption key; Step 5, Decryption key generation: The key generation center KGC generates a decryption key based on the detection results and sends it to the psychological analysis platform; Step 6, update the true value of mental health data: For the same type of mental health data, the psychological analysis platform CS aggregates and decrypts the ciphertext according to the decryption key to update the true value of the mental health data, and updates the auxiliary information of the weight update of the mental health data provider according to the new round of the true value of the mental health data, and sends it to the key generation center KGC; Step 7: Update weight of mental health data provider: The key generation center updates the weight of the mental health data provider according to the weight update auxiliary information sent by the psychological analysis platform; Repeat steps 5 to 7 until convergence is reached.

2. The privacy-preserving mental health data truth value discovery algorithm supporting outlier data filtering as claimed in claim 1, characterized in that: Step 1 specifically includes the following steps: Step 1.1: Given the security parameter K, the number of mental health data providers K, and the type of mental health data to be collected M, the key generation center KGC randomly selects two secure large prime numbers p,q←-SP(κ), calculates N=pq, and then randomly selects Calculate g = (g′) 2N mod N 2 Set the boundary later Hash functions The public parameters are pp = (N, g, X, Y, H), where Represents modulo N 2 The integer ring, that is, all integers from 0 to N 2 The set of integers from -1; Step 1.2: Based on the public parameter pp, the key generation center KGC generates Integer Gaussian distribution with mean 0 Randomly select a K-dimensional vector (s1, s2, ..., s K ) as the master private key msk, and each scalar s of msk k As the encryption key of the mental health data provider, the corresponding public key is calculated as And send it to the corresponding mental health data provider through a secure channel, and the key generation center KGC will publicly 3. The privacy-preserving mental health data truth value discovery algorithm supporting outlier data filtering as claimed in claim 1, characterized in that: Step 2 specifically includes the following steps: Step 2.

1. Each mental health data provider p k , where k∈{1,...,K}, and M types of mental health data are Use the pseudo-random number generator R to select a mask for each data Then calculate the randomized perception data where m∈{1,...,M}; Step 2.2, then generate the ciphertext of each mental health data, record the ciphertext as Send the ciphertext to the psychological analysis platform CS.

4. The privacy-preserving mental health data truth value discovery algorithm supporting outlier data filtering as claimed in claim 1, characterized in that: Step 3 specifically includes the following steps: Step 3.1: For each mental health data provider p k , psychological analysis platform CS selection challenge C k =(C k,1 , C k,2 , ..., C k，m , ..., C k，M ) is sent to the corresponding mental health data provider, where C k，m different from each other, k∈{1,...,K}, m∈{1,...,M}, C k，M represents the challenge value of the k-th mental health data provider on the M-th category of mental health data; Step 3.2: After receiving the challenge, the mental health data provider calculates the range detection proof (va k , vb k ) and send it to the psychological analysis platform CS.

5. The privacy-preserving mental health data truth value discovery algorithm supporting outlier data filtering as claimed in claim 1, characterized in that: Step 4 specifically includes the following steps: Step 4.1: For each mental health data provider p k , the psychological analysis platform CS is based on the ciphertext of the mental health data provider and the mental health data calculation range proof (va′ k , vb′ k ); Step 4.2: Perform range detection on the original mental health data to verify va k =va′ k , vb k =vb′ k If all of them are true, then the data of the mental health data provider is discarded, and the filtered mental health data provider set is recorded as That is, the detection result is obtained and then sent to the key generation center KGC to request the decryption key.

6. The privacy-preserving mental health data truth value discovery algorithm supporting outlier data filtering as claimed in claim 1, characterized in that: Step 5 specifically includes the following steps: Step 5.1: The key generation center KGC generates the decryption key And send it to the psychological analysis platform CS.

7. The privacy-preserving mental health data truth discovery algorithm supporting outlier data filtering as claimed in claim 1, characterized in that: Step 6 specifically includes the following steps: Step 6.1: For each type of mental health indicator l m , the psychological analysis platform CS aggregates and decrypts the ciphertext according to the decryption key to obtain the inferred truth value of each type of psychological health data The true value results of all categories of mental health data are recorded as Step 6.2: Calculate auxiliary information for weight update of mental health data provider And send it to the key generation center KGC.

8. The privacy-preserving mental health data truth value discovery algorithm supporting outlier data filtering as claimed in claim 1, characterized in that: Step 7 specifically includes the following steps: Step 7.1: The key generation center KGC updates the weight ω of each mental health data provider k .

Images