JPWO2016203555A1 - Concealed similarity search system and similarity concealment search method - Google Patents

Abstract

A similarity concealment search system for searching data to be searched similar to search target data that is concealed according to a search request, wherein a user terminal inputs search target data including qualitative data or quantitative data An input unit that accepts data, an encryption processing unit that encrypts qualitative data included in the search target data using an encryption method that can be used to determine whether the data matches or does not match, and quantitative data and search target data included in the search target data Based on the quantitative data included in the data, the similarity determination processing unit that calculates a condition for both to have a certain degree of similarity, the encrypted qualitative data and the condition, and the search target data And a communication unit that transmits to a similar search server for searching. The similar search server receives qualitative data and conditions from a user terminal, or determines that it is similar to search target data. Based on the server communication unit that transmits the search target data to the user terminal, the qualitative data received from the user terminal, and the qualitative data included in the search target data stored in the storage unit in advance If it is determined whether or not they are similar, and if it is determined that they are similar, the search target data including quantitative data that satisfies the conditions from the search target data determined to be similar is searched for A similarity determination calculation processing unit that outputs data similar to the data.

Description

  The present invention relates to a similarity search system and a similarity search method for searching for data similar to the data held by itself among plaintext or encrypted data stored in a server.

  In recent years, big data analysis that draws unknown and effective knowledge from vast amounts of data has attracted attention. In addition, companies are gathering not only their own data, but also collecting data from outside the company and using it for marketing and operational efficiency, various analyzes of information are increasingly recognized as important activities. On the other hand, such data management is inadequate, and there are many incidents and accidents in which a large amount of information including personal information of customers is leaked, which is a social problem.

  In such a leakage accident, the leaked company will bear a large amount of compensation, and in some cases, it will be related to the existence of the company. In addition, even for unrelated companies, there is a concern that an individual who is an information holder may have an adverse effect such as making it difficult to collect information and obtaining an effective analysis result when providing information.

  In these circumstances, legislation has been developed to correctly collect and use information. In addition, since the sensitivity and privacy of information handled in each industry, especially personal information, is different, guidelines in each industry are being developed. However, these are merely legal improvements and are effective in deterring information leakage, but are not information leakage countermeasures themselves.

  Conventionally, techniques for preventing information leakage and privacy infringement caused by such information have been considered such as access control for restricting people who access information and effective encryption for theft of equipment. However, in recent cases of information leakage, there are many cases due to misuse of authority of outsourcing contractors entrusted with managing computer resources. In order for such an outsourcer to manage information, a legitimate authority to access a database or the like is handed over, so that even information unnecessary for management work can be browsed and abused. In contrast, conventional access control is powerless, and when encryption is used to extract data from a recording medium like a storage cipher and decrypt it when used, Again, the administrator is powerless because he can view the decrypted data under a legitimate procedure.

  A decryption key is not placed in the outsourcer or in the vicinity of it, and only a limited and valid user in the company has the decryption key, and when using it, the ciphertext can be pulled back and decrypted. However, in the conventional encryption technology, when it is stored in a database, nothing can be processed in the encrypted state. For example, basic operations such as retrieval and aggregation in the case of numerical data must be given up. This is just a ciphertext storage and wastes the database function.

  Therefore, recently, attention has been focused on research on a technology that enables some processing without decryption while using a highly secure encryption technology. For example, research on searchable ciphers that can determine whether plaintext matches or does not match without decryption has been actively conducted. This is intended for the case where the plain text is a character string (name, address, etc.), and it is difficult to determine the order relationship such as a size comparison for numerical data. The comparison of numerical data, especially the determination of whether or not they are similar under a certain scale, is an important calculation function that appears frequently in normal business as well as in any analysis. At this time, when numerical data composed of a plurality of items is regarded as a vector, in many cases, the “scale” is expressed by a normal distance (Euclidean distance) in the vector space, and thus an inner product of the vectors. Therefore, when determining the similarity along the “scale” in which the data (vector) is kept secret, for example, using the technique disclosed in Non-Patent Document 1, the inner product calculation with the vector kept secret A method to execute is necessary.

Kana Shimizu, "Privacy preserving search for chemical compound libraries", JSBi-55, CBI / JSBi2011 (2011).

  In Non-Patent Document 1, it is possible to calculate the inner product while keeping it secret using homomorphic encryption. However, due to the slow processing speed of homomorphic encryption, it is difficult to process a large amount of data. Also, when considering the inner product calculation of one vector (A) and the other large number of vectors (B), if the vector (B) can obtain the inner product calculation result, the vector ( B) can restore the vector (A) and cannot keep confidentiality. Therefore, the inner product value itself is not disclosed in the vector (B), and it is necessary to calculate the ciphertext of the inner product value.

  However, in the method of sending ciphertexts corresponding to all of the large number of vectors held by vector (B) to vector (A), the amount of communication increases and processing on the vector (A) side as well as the vector (B) side is performed. It is also unrealistic to become enormous. Therefore, it is necessary to perform transmission / reception and decoding of data after narrowing down the vector (B) to some extent.

  That is, the inner product is used to determine the similarity of data when searching for and retrieving similar data from a large amount of numerical data stored in an external storage or the like while keeping numerical data held by the user confidential. In many cases, a secret inner product calculation protocol as in Non-Patent Document 1 is required. However, in order to determine similarity between a large amount of data, there is a problem that a huge amount of communication and a huge amount of calculation processing are required on both the user and the external device side, and the efficiency is deteriorated. there were.

  The present invention has been made in view of the above, and similarity concealment that can efficiently determine the similarity of data while maintaining secrecy without applying a processing load compared to the prior art. An object is to provide a search system and a similarity search method.

  In order to solve the above-mentioned problems and achieve the object, the similarity concealment search system according to the present invention searches for similarity search data similar to the search target data concealed according to the search request. In the search system, the user terminal can determine whether or not the input unit that receives input of the search target data including qualitative data or quantitative data matches the qualitative data included in the search target data. Based on the encryption processing unit for encrypting with an encryption method and the quantitative data included in the search target data and the quantitative data included in the search target data. A similarity determination processing unit for calculating a condition for having a search condition, a similarity search server for searching the search target data using the encrypted qualitative data and the condition The similarity search server receives the qualitative data and the condition from the user terminal, or the search target data determined to be similar to the search target data. Whether or not they are similar based on the server communication unit to be transmitted to the terminal, the qualitative data received from the user terminal, and the qualitative data included in the search target data stored in advance in the storage unit If the data is determined to be similar to each other, the search target data including quantitative data satisfying the above condition among the search target data determined to be similar is similar to the search target data. And a similarity determination calculation processing unit that outputs as data to be processed.

  Moreover, this invention is grasped | ascertained also as the similarity secret search method performed with the said similarity secret search system.

  According to the present invention, it is possible to efficiently determine the similarity of data while maintaining confidentiality without applying a processing load as compared with the prior art.

It is a figure which shows the structural example of the similarity concealment search system which concerns on this embodiment. It is a figure which shows the structural example of the data storage and similarity search server which concerns on 1st embodiment. It is a figure which shows the structural example of the user terminal which concerns on 1st embodiment. It is a table | surface showing the classification | category of the data which concern on 1st embodiment. It is a flowchart which shows each user's process and data storage / similarity search server process which concern on 1st embodiment. It is a flowchart which shows the data generation procedure used for the user similarity determination which concerns on 1st embodiment. It is a flowchart which shows the procedure of the data storage by the user's data encryption which concerns on 2nd embodiment, its transmission, and a data storage / similarity search server. It is a flowchart which shows a user's data encryption which concerns on 2nd embodiment, determination condition production | generation, its transmission, and the narrowing-down procedure of the similar data by a data storage / similarity search server.

  Next, modes for carrying out the present invention (referred to as “embodiments”) will be described in detail with reference to the drawings as appropriate. Note that, in the embodiment described below, the same components are denoted by the same reference symbols in principle, and repeated description is omitted.

"System configuration"
FIG. 1 is a diagram showing a configuration example of a similarity concealment search system 1000 to which the similarity concealment search system and the similarity concealment search method according to the present invention are applied. As shown in FIG. 1, the similarity confidential search system 1000 has a configuration in which a plurality of user terminals 300 to 500 and a similarity search server 200 are connected to each other via a network 100. For example, in a certain service provision form, the similar search server 200 is installed in a cloud provider and is configured from a server device that is a general computer. In addition, the user terminal 300 operated by a user who uses the service of the cloud provider is configured by an information processing apparatus that is a general computer such as a PC, a mobile phone, or a smartphone. In FIG. 1, the present system is described as a configuration including three user terminals 300 to 500, but the number is arbitrary.

<< first embodiment >>
Next, a first embodiment of the similarity concealment search system 1000 will be described with reference to FIGS.

(Similar search server)
FIG. 2 is a diagram illustrating a configuration example of the similar search server 200 according to the first embodiment. As illustrated, the similarity search server 200 includes a control unit 210 that processes data, a storage unit 220 that stores data, an input unit 201 that receives input of information, an output unit 202 that outputs information, and an external device. And a communication unit 203 that transmits and receives data. For example, the control unit 210 includes an arithmetic device such as a CPU (Central Processing Unit), the input unit 210 includes an input device such as a keyboard, and the output unit 202 includes a display device such as an LCD (Liquid Crystal Display). The communication unit 203 includes a communication device such as a NIC (Network Interface Controller). The control unit 210 includes an overall processing unit 211 and a similarity determination calculation processing unit 212.

  The overall processing unit 211 performs overall control of processing in the similarity search server 200, stores information received via the input unit 201 in the storage unit 220, and transmits and receives information to and from the user terminals 300 to 500 To control. Further, the overall processing unit 211 performs processing for displaying data on the output unit 202, reads data stored in the storage unit 220, and transmits the data to the user terminals 300 to 500 via the communication unit 203.

  The similarity determination calculation processing unit 212 determines the similarity between the search data received from the user terminals 300 to 500 and the search target data stored in the similar search server 200 in advance, and becomes equal to or greater than a certain threshold. Data is extracted from the search target data and output. Specifically, the similarity determination calculation processing unit 212 determines the similarity between an item indicating qualitative data included in the search data and an item indicating qualitative data included in the search target data. For example, when the item indicating the qualitative data included in the search data and the item indicating the qualitative data included in the search target data match 70% or more, it is determined that the two are similar. This similarity determination is performed while being encrypted with an encryption key for encrypting an item indicating qualitative data.

  Further, the similarity determination calculation processing unit 212 narrows down the search target data determined to be similar, and includes items indicating quantitative data included in the narrowed search target data and the search data. Calculate the inner product value with the item indicating the quantitative data encrypted with the encryption key for encrypting the quantitative data, and within the narrowed search target data, the inner product value is a certain threshold value or more Extract the data. For example, when the inner product value of an item indicating quantitative data included in the search data and an item indicating quantitative data included in the narrowed search target data is 0.7 or more, the data is narrowed down. From the search target data. This similarity determination is performed while being encrypted with an encryption key for encrypting quantitative data. The extracted data is transmitted as search result data 3303 by the overall processing unit 211 via the communication unit 203.

  The data storage unit 230 of the storage unit 220 stores the data 231 extracted from the search target data by the similarity determination calculation processing unit 212. The temporary information storage unit 240 stores information that is temporarily required for processing executed by the control unit 210. Processing performed by each unit of the similarity search server 200 will be described later using a flowchart.

(User terminal)
FIG. 3 is a diagram illustrating a configuration example of the user terminal 300 according to the first embodiment. In addition, since the structure of each user terminal is the same structure, below, the structural example of the user terminal 300 is shown. When these user terminals do not perform encryption, the encryption processing unit 312, the decryption processing unit 313, the encryption / decryption key storage unit 340, the similarity determination processing unit 314, the encryption / decryption key Each unit necessary for encryption / decryption such as the storage unit 340 may not be provided.

  As illustrated, the user terminal 300 includes a control unit 310 that processes data, a storage unit 320 that stores data, an input unit 301 that receives input of information, an output unit 302 that outputs information, an external device, and the like. And a communication unit 303 that transmits and receives the data. Each of the above units is configured by a device having the same function as the similar search server 200.

  The control unit 310 includes an overall processing unit 311, an encryption processing unit 312, a decryption processing unit 313, and a similarity determination processing unit 314.

  The overall processing unit 311 performs overall control of processing in the user terminal 300, stores received information in the storage unit 320 via the input unit 301, and transmits / receives information to / from the similarity search server 200. Control. In addition, the overall processing unit 311 performs processing for displaying data on the output unit 303, reads data stored in the storage unit 320, and causes the encryption processing unit 312 to encrypt data or decrypt data, respectively. The encryption processing unit 313 decrypts the encrypted data. The overall processing unit 311 generates data similar to the search target data selected by the user in advance as search data for searching the search target data based on the information received by the input unit 301. . This search data becomes the pre-encryption data 3301. In addition, the overall processing unit 311 transmits the encrypted data 3302 after the pre-encrypted data 3301 is encrypted by the encryption processing unit 312 to the similarity search server 200 via the communication unit 303.

  The encryption processing unit 312 encrypts the item indicating the qualitative data with the encryption key 341 for encrypting the item indicating the qualitative data in the pre-encryption data 3301 and generates the encrypted data 3302 To do. Further, the encryption processing unit 312 vectorizes the target vector obtained by vectorizing the item indicating the quantitative data and the item indicating the quantitative data included in the search target data, which is obtained by the similarity determination processing unit 314. A set of base vectors constituting a target vector that satisfies a threshold for determining that the inner product value with the target vector is equal to or greater than a certain value is obtained by an encryption key 341 for encrypting an item indicating quantitative data. Encrypt. Satisfying this threshold is a condition for both to have a certain similarity or more.

  The decryption processing unit 313 decrypts the item indicating the qualitative data and the item indicating the quantitative data of the search result data 3303 received from the similar search server 200 with the decryption key 342 to generate decrypted data 3304.

  The similarity determination processing unit 314 performs a linear decomposition process that is a process for obtaining a set of base vectors that constitute a target vector that satisfies the above threshold, and sets a set of base vectors and a scalar value that is a coefficient of the base vector Are output.

  The storage unit 320 includes a data storage unit 330, an encryption / decryption key storage unit 340, and a temporary information storage unit 350. The data storage unit 330 includes the pre-encryption data 3301, the post-encryption data 3302, search result data 3303 that is similar to the pre-encryption data 3301 searched by the similar search server 200, and search result data. 3303 stores the decrypted data 3304 decrypted by the decryption key 342.

  The encryption / decryption key storage unit 340 stores an encryption key for encrypting the pre-encryption data 3301 and a decryption key 342 for decrypting the search result data 3303. The encryption key may store an encryption key for encrypting qualitative data and an encryption key for encrypting quantitative data, or may store the same encryption key. . Similarly to the encryption key, the decryption key may be stored for each qualitative data and quantitative data, or the same decryption key may be stored. The temporary information storage unit 350 stores information temporarily necessary for processing executed by the control unit 310 as in the case of the similar search server 200.

  Hereinafter, the user terminals 300 to 500 transmit the search target data to the similar search server 200 in advance, and the similar search server 200 stores the search target data as data 231 in the data storage unit 230. Suppose you are. In the first embodiment, the data 231 will be described on the assumption that qualitative data is recorded as plain text or cipher text, and quantitative data is recorded as plain text.

(Data classification)
FIG. 4 is a table in which search target data to be searched from the search target data by the user terminal 300 is classified. Since the search target data handled by this system includes items indicating qualitative data and items indicating quantitative data, the classification of both will be described below. The qualitative data is, for example, data only for distinguishing classifications and types such as gender, address, occupation, etc., and is data that cannot be calculated numerically. Further, the quantitative data is data that can be calculated numerically, for example, in terms of numerical values such as numerical data, number of sheets, amount of money, etc. of examination results in a hospital.

  In this system, as shown in FIG. 4, in both cases of items indicating qualitative data and items indicating quantitative data included in the search target data, these data are not sent to others (for example, patients receiving treatment or If it can be disclosed to a person other than the family, it is used as search data for narrowing down the search target data without encryption (S401, S402).

  On the other hand, if these data cannot be disclosed to others, the items indicating qualitative data are encrypted using an encryption method capable of determining whether they match or not, and used as search data (S403). . As such an encryption method, for example, a searchable encryption technique (Japanese Patent No. 5412414) can be used. The item indicating the quantitative data is encrypted using an encryption method capable of determining the similarity between the two while being kept secret and used as search data (S404). As such an encryption method, for example, various conventionally known homomorphic encryption techniques can be used. Whether or not each item included in the search data can be disclosed depends on the disclosure criteria, for example, when the consent of the patient or their family cannot be obtained, or when confidentiality is high for research purposes. Can be determined.

(Example of narrowing down)
FIG. 5 is a flowchart showing a processing procedure of a process in which the user terminal 300 searches for similar data from search target data stored in advance using the search target data, and the similar search server 200 narrows down. . In the following, as an example of narrowing down and searching for similar data from the search target data using the search target data, the search data of a patient is set as search target data, and the test data of a patient similar to that patient is searched Target data. The examination target data and the search target data include items indicating one or more qualitative data (for example, patient name, address, sex, disease name, etc.) and items indicating one or more quantitative data (for example, Patient age, laboratory values, etc.).

  First, in the user terminal 300, the overall processing unit 311 generates items of search data by extracting items indicating qualitative data and quantitative data from the search target data, and the generated search data is encrypted. It is stored as data 3301 (S501). At this time, if it is not necessary to encrypt the extracted item, the process proceeds to S505.

  Subsequently, the encryption processing unit 312 encrypts items indicating qualitative data in the pre-encryption data 3301 using the encryption key 341, and generates post-encryption data 3302 (S502). At this time, if an item indicating quantitative data is not included in the search data, the process proceeds to S505.

  Subsequently, the similarity determination processing unit 314 executes the linear decomposition process, and outputs a set of basis vectors and a set of scalar values that are coefficients of the basis vectors (S503). Specific processing of the linear decomposition processing will be described later with reference to FIG.

  The encryption processing unit 312 encrypts the set of basis vectors and the set of scalar values output by executing the above linear decomposition process using the encryption key 341 (S504).

  The overall processing unit 311 transmits the data generated in S501 to S504 to the similarity search server 200 (S505). Note that the steps S501 to S504 do not necessarily have to be executed. If the search data is plaintext only, if only items indicating qualitative data are included, only items indicating quantitative data are included. In this case, it can be selected and executed as described above. Either S502 or S503 may be processed first.

  In the similar search server 200, the similarity determination calculation processing unit 212 narrows down the search target data using the data received from the user terminal 300 as a key (S506). For example, when the data received from the user terminal 300 includes an item indicating qualitative data, the item and an item indicating qualitative data included in the search target data match 70% or more. In this case, it is determined that both are similar, and the search target data is narrowed down. In the above-mentioned narrowing down, for example, narrowing down is performed by complete matching, partial matching, or synonyms for character string data. In this case, a dictionary for searching for synonyms is stored in advance in the similar search server 200, and the degree of coincidence between the dictionary and the data and the degree of coincidence between the dictionary and the data to be searched are similar. Alternatively, it may be determined that both are similar.

  The similarity determination calculation processing unit 212 calculates the inner product value of the item indicating the quantitative data included in the search target data narrowed down in S506 and the item indicating the quantitative data included in the data, and is narrowed down. Data for which the inner product value is equal to or greater than a threshold value is extracted from the search target data (S507). For example, all data for which the inner product value of both is equal to or greater than the threshold value 0.7 is extracted from the narrowed search target data.

  Then, the overall processing unit 211 transmits the extracted data as search result data 3303 via the communication unit 203 (S508).

  The decryption processing unit 313 of the user terminal 300 decrypts the item indicating the qualitative data and the item indicating the quantitative data of the search result data 3303 received from the similar search server 200 with the decryption key 342, and decrypts the decrypted data 3304. Is generated (S509). Thereafter, the overall control unit 311 of the user terminal 300 outputs the generated decoded data 3304 to the output unit 302, and the result is evaluated by the user (S510).

(Linear decomposition)
FIG. 6 is a flowchart showing the processing procedure of the linear decomposition process in S503 shown in FIG. The user terminal 300 has a vector X as quantitative data to be concealed (S404). The similarity determination processing unit 314 of the user terminal 300 first determines the natural numbers m and t (S601), and randomly generates t types of linear decompositions using m basis vectors constituting the vector X (S602). ). Here, some scalars aij may be zero.

  Subsequently, the similarity determination processing unit 314 calculates a basis vector Gij and a scalar value that constitute each vector X derived from a threshold value that the inner product value with the vector X should satisfy, that is, a condition of the inner product value that satisfies the threshold value. (S603).

  The similarity determination processing unit 314 sets the vector set {Gij} generated in S602 and the scalar value derived in S603 as data to be output in S503 illustrated in FIG. 5 and each scalar value {aij}. Is stored in the encryption / decryption key storage unit 340 in a secret state without being disclosed to others (S604). When the process of S604 ends, the linear decomposition process ends.

(Summary of first embodiment)
According to the first embodiment, similarity determination is performed while encrypting qualitative data included in data for determining similarity and qualitative data included in data for which similarity determination is performed, and data that is equal to or greater than a predetermined threshold value Furthermore, data whose inner product value of the quantitative data included in the extracted data and the quantitative data included in the similarity determination data is equal to or larger than a predetermined threshold is narrowed down, and the result is returned to the client. Therefore, the user (s) can leave the plaintext or encrypted data in the similar search server, and a certain user (user terminal 300) can disclose part of the data held by himself / herself and keep the rest secret. Similar search. That is, it becomes possible to efficiently search for similar data while concealing the data owned by the user from a large amount of data deposited in a third party such as the cloud.

  For the parts that can be disclosed in plain text, the conventional similarity determination is performed as it is to narrow down, and for the parts that cannot be disclosed, an encryption method that can determine the match / mismatch while encrypting the qualitative data is applied. In addition, the quantitative data is narrowed down by sending additional data necessary for similarity determination. In the first embodiment, a method using random linear division of data to be concealed is described as the additional data. However, other methods may be used.

<< Second Embodiment >>
Next, a second embodiment of the similarity concealment search system 1000 will be described with reference to FIGS.

  In the second embodiment, in contrast to the first embodiment, the user terminal 300 (which may include the user terminals 400 and 500; the same shall apply hereinafter) encrypts the quantitative data portion that is to be concealed among the data deposited in advance. A procedure and a procedure in which the vector X which is the data of the similar search target is encrypted and sent, and the similar search server 200 executes the similar search processing are added.

  In the user terminal 300, the similarity determination processing unit 314 randomly generates orthogonal transformation T for a plurality of quantitative data deposited in the similarity search server 200 (S701), and randomly generates a scalar r (S702). Further, a vector W is randomly generated (S703). The order of generation may be changed. W may be a 0 vector.

  Next, the similarity determination processing unit 314 regards the quantitative data portion {Yi} to be concealed of the search target data to be deposited as a vector, performs affine transformation Y′i = rTYi + W on the vector, {Y In order to deposit 'i} together with other data, these data are transmitted to the similarity search server 200 (S705). The similarity determination calculation processing unit 212 of the similarity search server 200 stores the data received from the user terminal 300 in the storage unit 220 (S706).

  And when concealing the quantitative data to be concealed of the search target data to be searched for by the user terminal 300 as the vector X, in the second embodiment, first, the similarity determination processing unit 314 of the user terminal 300 A scalar s is randomly generated (S801), orthogonal transformation and affine transformation are performed to calculate X ′ = sTX (S802), and each Y′i and scalar value derived from a threshold value that the inner product value with the vector X should satisfy That is, a condition for the inner product value satisfying the threshold is calculated (S803). The user terminal 300 transmits the above X ′ and the condition obtained in S803 to the similar search server 200 (S804). That is, the similarity determination processing unit 314 of the user terminal 300 regards the quantitative data included in the search target data as a vector, and performs orthogonal transformation and affine transformation on the vector as in S802, and after these transformations In order to determine that the vector is similar to the search target data and the search target data, the inner product value to be satisfied by the converted vector is calculated as a condition.

  As will be described later, the similarity determination calculation processing unit 212 of the similarity search server 200 includes quantitative data that satisfies the inner product value among the search target data determined to be similar based on the qualitative data. The search target data is output as the data.

  The similarity determination calculation processing unit 212 of the similar search server 200 calculates the inner product of the recorded {Y′i} and X ′, extracts the one that matches the condition calculated in S803, and is searched for The data is narrowed down and the list is generated (S805). Thereafter, the similarity determination calculation processing unit 212 executes the same processing as the processing after S507 shown in FIG. 5 and transmits the extracted data (S806, S507 and S508 in FIG. 5).

(Summary of second embodiment)
According to the second embodiment, the user can perform narrowing down by performing a prior encryption deposit and a narrowing-down procedure while keeping confidential portions of quantitative data to be concealed. Similar search is possible.

100: network 200: similarity search server 201: input unit 202: output unit 203: communication unit 210: control unit 211: overall processing unit 212: similarity determination calculation processing unit 220: storage unit 230: data storage unit 231: data 240 : Temporary information storage unit 300, 400, 500: User terminal 301: Input unit 302: Output unit 303: Communication unit 310: Control unit 311: Overall processing unit 312: Encryption processing unit 313: Decryption processing unit 314: Similarity Determination processing unit 320: storage unit 330: data storage unit 340: encryption / decryption key storage unit 341: encryption key 342: decryption key 350: temporary information storage unit.

Claims (8)

A similarity concealment search system that searches for search target data similar to search target data that is concealed according to a search request,
User terminal
An input unit for receiving input of the search target data including qualitative data or quantitative data;
An encryption processing unit that encrypts the qualitative data included in the search target data using an encryption method capable of determining coincidence or mismatch;
A similarity determination processing unit that calculates a condition for both to have a certain degree of similarity based on the quantitative data included in the search target data and the quantitative data included in the search target data; ,
A communication unit that transmits the encrypted qualitative data and the condition to a similar search server for searching the search target data; and
The similar search server
A server communication unit that receives the qualitative data and the condition from the user terminal, or transmits the search target data determined to be similar to the search target data, to the user terminal;
It is determined whether or not both are similar based on the qualitative data received from the user terminal and the qualitative data included in the search target data stored in advance in the storage unit. Similarity determination calculation that outputs search target data including quantitative data satisfying the condition among the search target data determined to be similar as data similar to the search target data A processing unit;
A similarity concealment search system comprising: The similarity concealment search system according to claim 1,
The similarity determination processing unit of the user terminal regards the quantitative data as a vector, generates a set of one or a plurality of base vectors and a scalar value constituting the vector, generates the base vector, and In order to determine that the search target data and the search target data are similar, the inner product value to be satisfied by the basis vector is calculated as the condition,
The similarity determination calculation processing unit of the similar search server includes search target data including the quantitative data satisfying the inner product value among the search target data determined to be similar based on qualitative data. Output as the data,
The similarity concealment search system characterized by this. The similarity concealment search system according to claim 2,
The similarity determination processing unit of the user terminal regards the quantitative data included in the search target data as a vector, and performs orthogonal transform and affine transform on the vector, the vector after each transform, and the search In order to determine that the target data and the search target data are similar, the inner product value to be satisfied by the converted vector is calculated as the condition,
The similarity determination calculation processing unit of the similar search server includes search target data including the quantitative data satisfying the inner product value among the search target data determined to be similar based on qualitative data. Output as the data,
The similarity concealment search system characterized by this. The similarity concealment search system according to claim 1,
The input unit accepts input of the qualitative data or the quantitative data including plaintext that is not concealed;
The encryption processing unit does not perform processing when the qualitative data is the plaintext,
The similarity determination processing unit is based on the quantitative data included in the search target data and the quantitative data included in the search target data. To calculate the conditions of
The similarity concealment search system characterized by this. A similarity concealment search method for searching data to be searched similar to search target data concealed according to a search request,
An input step of receiving input of the search target data including qualitative data or quantitative data from a user terminal;
An encryption processing step of encrypting the qualitative data included in the search target data by an encryption method capable of determining coincidence or mismatch;
A similarity determination processing step for calculating a condition for both to have a certain similarity or more based on the quantitative data included in the search target data and the quantitative data included in the search target data; ,
Transmitting the encrypted qualitative data and the condition to a similar search server for searching the search target data;
Receiving the qualitative data and the condition from the user terminal;
A determination step of determining whether or not both are similar based on the qualitative data received from the user terminal and the qualitative data included in the search target data stored in advance in a storage unit;
When it is determined that they are similar, the search target data including quantitative data that satisfies the above condition among the search target data determined to be similar is output as data similar to the search target data Similarity determination calculation processing step,
A server transmission step of transmitting the search target data determined to be similar to the search target data to the user terminal;
The similarity concealment search method characterized by including this. The similarity concealment search method according to claim 5,
In the similarity determination processing step, the quantitative data is regarded as a vector, a set of one or a plurality of base vectors and a scalar value constituting the vector is generated, the generated base vector, the search target data, In order to determine that the search target data is similar, the inner product value to be satisfied by the basis vector is calculated as the condition,
In the similarity determination calculation processing step, search target data including the quantitative data satisfying the inner product value among the search target data determined to be similar based on qualitative data is output as the data. ,
The similarity concealment search method characterized by this. The similarity concealment search method according to claim 6,
In the similarity determination processing step, the quantitative data included in the search target data is regarded as a vector, orthogonal transformation and affine transformation are performed on the vector, the vector after each transformation, the search target data, and the search target data In order to determine that the search target data is similar, the inner product value to be satisfied by the converted vector is calculated as the condition,
In the similarity determination calculation processing step, search target data including the quantitative data satisfying the inner product value among the search target data determined to be similar based on qualitative data is output as the data. ,
The similarity concealment search method characterized by this. The similarity concealment search method according to claim 5,
In the input step, the input of the qualitative data or the quantitative data including plaintext that is not concealed is accepted,
In the encryption processing step, when the qualitative data is the plaintext, the processing is not performed,
In the similarity determination processing step, both have a certain similarity or more based on the quantitative data of the plaintext included in the search target data and the quantitative data included in the search target data. To calculate the conditions of
The similarity concealment search method characterized by this.

Images