KR101285209B1 - Method of enforcing privacy based on All or Nothing Transform - Google Patents

Abstract

AONT-based privacy enhancement method is provided for distributed data storage management that is suitable for large data processing and minimizes database capacity through variance of transform size and XOR operation according to All-Or-Nothing Transform (AONT) characteristics. . Fragmentation is performed on the information to be encrypted according to the fragmentation attribute, and all the attributes except the fragment attribute are divided into a plurality of transform pieces having a variable size in an AONT manner. One of the plurality of translation pieces is stored in the database for the security agent and the other translation pieces are mapped and stored in the server database.

Description

Method of enforcing privacy based on All or Nothing Transform}

The present invention relates to a privacy enhancement method, and more particularly, to an AONT-based privacy enhancement method.

Recently, due to the development of the Internet, the cost of managing and maintaining large amounts of data over tens or hundreds of TB stored and processed is emerging. To solve these problems, a computing environment for realizing data outsourcing services such as SaaS is required. Currently, cloud computing is attracting attention as a technology for managing Internet outsourcing services. Cloud computing is defined as computing that utilizes Internet technologies to provide IT resources in the form of services that provide highly scalable and flexible IT resources to a variety of external customers. Such cloud computing enables the efficient storage and management of large volumes of data based on virtualization and the utilization of IT resources inexpensively and efficiently. In order to store and manage such a large amount of data, a distributed data management system has become indispensable.

Currently, health related information such as internet portal service, personal information, medical history information is stored and managed in a distributed database and shared and utilized by users. However, because the information shared and utilized includes sensitive information of individuals, the infringement of privacy causes the infringement of privacy. In particular, if a data is shared or leaked without the data provider's consent in a cloud computing environment, a privacy breach may occur. For example, the medical history, name, address, telephone number, age, etc. of a patient who leaked medical information stored and managed in an undisclosed medical data repository by an insider are sold to health professionals nationwide. The medical records were leaked from the hospital to the insurance company and used to manipulate the insurance.

Therefore, data encryption may be used as a method for protecting data, and privacy should be ensured so that it is impossible to identify whose data even after the data is leaked. In addition, companies using cloud computing should provide confidentiality to prevent leakage by insiders who manage data in the cloud by controlling access to sensitive data.

As such, a major issue with the introduction of cloud computing is security. Currently, the Cloud Security Alliance (CSA) and Gartner publish seven security guidelines for evaluating the top seven security threats and threats to cloud computing. Loss and leakage threats can occur, and information on access rights and data need to be separated.

On the other hand, all data is associated with a specific individual, causing privacy infringement. In the case of data generated in the hospital, the patient list and the patient's medical records are related, so that it is possible to know who is suffering from what disease. Therefore, if the patient list and the medical record can be prevented from being related, privacy infringement can be prevented, and if the medical information is stored, the data must be securely separated so that sensitive information and identity information are not related. In order to solve this problem, it is necessary to provide anonymity of data, that is, unlinkability.

Among these studies, the fragmentation and encryption method proposed by V. Ciriani et al. Provides confidentiality by disabling the connection between the data and the identification between the data. However, this method suffers from an increase in database capacity due to duplicate encrypted data of attributes due to fragmentation.

In a cloud service environment that stores and manages large amounts of data, fragmentation methods suitable for processing large amounts of data are required to solve confidentiality and privacy problems.

Conventional fragmentation and encryption schemes have been proposed by V. Ciriani et al. As a way to ensure the privacy of data. This approach ensures privacy by using fragmentation and encryption to disable linkage between attributes. First, we set confidentiality constraints according to the association with sensitive data and use encryption method for attributes except fragmentation and fragmented data. The encryption method used here uses the CBC mode.

Confidentiality constraint

Confidentiality constraints establish constraints on associations between attributes in order to achieve fragmentation. For example, in the hospital, if there are attributes such as patient number, patient name, date of birth, address, illness name, and doctor, the information is not sensitive to each attribute, but it can be regarded as sensitive information due to the association of patient name and disease name. have. This violates privacy because the patient's name can infer the person's illness. Therefore, the confidentiality constraint can be defined as follows.

[ Definition 1] Confidentiality Constraints

If A is an attribute set, then the confidentiality constraint is a subset c of A. (c⊆A)

Table 1 describes the medical data.

patient
number patient
name birth year
Month day address Disease charge
doctor One Kwang Yong Park 53/10/07 Seoul High blood pressure AA 2 Do Jeong Min 81/01/03 Ulsan obesity BB 3 Jun Jun 52/02/12 Seoul High blood pressure AA 4 Hyojin Kim 81/01/03 Gyeongju obesity CC

Patient number is considered sensitive (c ₀ ).

The association of patient names with other attributes is considered sensitive (c ₁ , c ₂ , c ₃ , c ₄ ).

-Date of birth and address can be deduced from the name of the illness. That is, it is considered sensitive in relation to other information (c ₅ , c ₆ ).

Using the medical data in Table 1 as an example, the set of confidentiality constraints can be set as follows.

c ₀ = {patient number}, c ₁ = {patient name, date of birth}

c ₂ = {patient name, address}, c ₃ = {patient name, disease name}

c ₄ = {patient name, doctor},

c ₅ = {date of birth, address, illness}

c ₆ = {date of birth, address, doctor}

Fragmentation and Encryption

Fragmentation and encryption are used to meet confidentiality constraints. Here, fragmentation is applied to the entire property according to the constraint, and the property set is divided so that it is not visible to each other. In other words, association inference between attribute values is not possible without access to the encryption key. Encryption applies CBC encryption to all attributes according to constraints.

 [Definition 2] Fragmentation

R is the relational schema, the short-cut sum of R is F = {F ₁ , F ₂ , ..., F _m } and F _i ⊆R. (i = 1, 2, ...., m)

Map all of the attributes in the short-cut F to the physical fragment and encrypt all other attributes except the fragment in R.

[Definition 3] Physical Fragments

에 대한 물리적 단편은 관계스키마

이다.R is the relational schema, and R's short edit sum is F = {F ₁ , F ₂ , ..., F _m }. Each fragment attribute set

Physical fragments for relation schema

to be.

Here, enc means encryption of all attributes of R that do not belong to the fragment and are encrypted by XOR operation with each attribute value and IV. IV is an initialization vector in CBC mode.

는 관계스키마 r∈R이고 F_i에 대한 평문 튜플 t∈r은 F_i ^e의 튜플 t^e∈f_i ^e로 맵핑된다. 그리고 단편 F_i 를 제외한 속성은 t^e[enc]로 맵핑된다. Each fragment attribute set

The relationship schema r∈R a plaintext tuple t∈r for F _i of F _i are mapped to a tuple t ^{e e} _i ^e ∈f. And the attribute except for the fragment F _i is mapped to t ^e [enc].

Here, the physical fragment f _i ^e performs the following operation.

Taking the physical data of Table 1 as an example, the physical fragment f ₁ ^e is described as follows.

R = {Patient number, patient name, date of birth, address, illness name, doctor}

The short edit F of R is as follows.

-F = {{patient name}, {date of birth, address}, {name of doctor, doctor}}

F ₁ = {patient name}

-F ₂ = {date of birth, address}

F ₂ = {name of disease, doctor}

When fragments are performed here, due to confidentiality constraints, the patient number itself is encrypted rather than fragmented because it is considered sensitive information.

In order to calculate the physical fragment t ^e [enc], RF ₁ is a set of attributes minus {patient number, patient name, date of birth, address, disease name, doctor}-{patient name} and the result of XOR operation with t ^e [IV]. Encrypted in CBC mode.

Performing such a physical fragment is generated as shown in FIG. 1 illustrates a conventional cryptographic based physical fragment.

f ₁ ^e (IV, enc, patient name)

-f ₂ ^e (IV, enc, date of birth, address)

f ₃ ^e (IV, enc, disease name, doctor)

Fragmentation and encryption can provide confidentiality due to encrypted data even when fragments f ₁ ^e , f ₂ ^e, and f ₃ ^e are shared due to fragmentation and encryption, and can prevent privacy breaches by disabling connections between attributes. have. However, when fragmenting, the storage of duplicate attributes increases the capacity of the data storage space. Fragments f ₁ ^e , f ₂ ^e and f ₃ ^e are described in Tables 2, 3 and 4, respectively.

enc patient
name patient
number birth year
Month day address Disease charge
doctor cryptogram cryptogram cryptogram cryptogram cryptogram Plain text

enc birth year
Month day address patient
number patient
name Disease charge
doctor cryptogram cryptogram cryptogram cryptogram Plain text Plain text

enc Disease charge
doctor patient
number patient
name birth year
Month day address cryptogram cryptogram cryptogram cryptogram Plain text Plain text

Tables 2 to 4 show the attributes that are actually encrypted in each fragment when the physical fragment is performed as shown in FIG. If the physical fragment is not performed, only the attributes {patient number, date of birth, address, disease name, doctor, patient name} need to be stored. However, when a physical fragment is performed, the same attributes are stored in duplicate depending on the number of fragments.

For example, when performing a physical fragment for Table 1 is stored as shown in Figure 1, each storage space is shown in Table 5, Table 6, Table 7. In this case, the total storage space is required 720KB. (For the sake of understanding, the size of ciphertext and plaintext should be 10KB).

<Table 5>: Ciphertext size (200KB) + Plaintext size (40KB) = 240KB

<Table 6>: Ciphertext size (160KB) + Plaintext size (80KB) = 240KB

<Table 7>: Ciphertext size (160KB) + Plaintext size (80KB) = 240KB

enc patient
name patient
number birth year
Month day address Disease charge
doctor 10 KB 10 KB 10 KB 10 KB 10 KB 10 KB 10 KB 10 KB 10 KB 10 KB 10 KB 10 KB 10 KB 10 KB 10 KB 10 KB 10 KB 10 KB 10 KB 10 KB 10 KB 10 KB 10 KB 10 KB

enc date of birth address patient
number patient
name Disease charge
doctor 10 KB 10 KB 10 KB 10 KB 10 KB 10 KB 10 KB 10 KB 10 KB 10 KB 10 KB 10 KB 10 KB 10 KB 10 KB 10 KB 10 KB 10 KB 10 KB 10 KB 10 KB 10 KB 10 KB 10 KB

enc Disease charge
doctor patient
number patient
name birth year
Month day address 10 KB 10 KB 10 KB 10 KB 10 KB 10 KB 10 KB 10 KB 10 KB 10 KB 10 KB 10 KB 10 KB 10 KB 10 KB 10 KB 10 KB 10 KB 10 KB 10 KB 10 KB 10 KB 10 KB 10 KB

As a result, fragmentation results in duplicate attributes and increases the storage capacity to store them.

The present invention was created to improve the above-mentioned conventional problems, and is suitable for processing large data through variability of transform piece size according to AONT characteristics and XOR operation, and to manage distributed data storage that can minimize database capacity increase. The purpose is to provide an AONT-based privacy enhancement method.

The AONT-based privacy enhancement method according to the present invention comprises the steps of performing fragmentation on information to be encrypted according to fragmentation attributes and dividing all the attributes except the fragment attribute into a plurality of transform pieces having variable sizes in an AONT manner; And storing one of the plurality of transform pieces in a first database and the remaining transform pieces are mapped and stored in a second database.

The present invention proposes a privacy enhancement method using the AONT method to solve the problem of the increase of storage space due to the overlapping of attribute data. Through the proposed technique, we can minimize the data storage space while enhancing the privacy through the existing fragmentation. The present invention is suitable for processing a large amount of data through the variability of the size of the transform piece according to the AONT characteristics and the XOR operation can minimize the increase of the database capacity.

1 is a diagram illustrating a conventional encryption-based physical fragment.
2 is a diagram illustrating an AONT method according to an embodiment of the present invention.
3 is a flowchart of treatment information to which the present invention is applied.
4 illustrates an AONT based physical fragment according to an embodiment of the present invention.
5 is a diagram illustrating a configuration of a transform piece according to an embodiment of the present invention.
6 and 7 are diagrams illustrating a state in which the remaining transform fragments are stored in a DB for a server except for one transform fragment according to an embodiment of the present invention.
8 is a graph showing a result of comparing the present invention with the conventional method.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. In the following description of the present invention, a detailed description of known functions and configurations incorporated herein will be omitted when it may make the subject matter of the present invention rather unclear. The following terms are defined in consideration of the functions of the present invention, and may be changed according to the intentions or customs of the user, the operator, and the like. Therefore, the definition should be based on the contents throughout this specification.

The All-Or-Nothing Transform (AONT) scheme was proposed by Rivest, the inventor of RSA cryptography, and was originally used in OAEP to increase the strength of RSA cryptography. If AONT operation is performed on the plain text data, output data having the same size as the original data can be obtained. When all bits of the output data are collected, the original data can be restored.

로 분할하여 h 함수 및 g 함수를 이용하여 XOR 연산을 통해 출력데이터

를 얻을 수 있다. AONT 방식은 다음과 같이 계산되는데, 여기서 s는 평문 블록수이다. 먼저, 분할된

에 대해

를 계산하고

를 계산한다. The AONT method has a hash function h: {0, 1} ^{ℓ (s-1)} → {0, 1} ^ℓ and pseudo random number generator g: {0, 1} ^ℓ → {0, 1} ^{ℓ (s-1)} It consists of an operation. First, the plain text data M

Output data through XOR operation using h and g functions

Can be obtained. The AONT method is calculated as follows, where s is the number of plaintext blocks. First, divided

About

Calculate

.

를 계산한다.And,

.

과

를 분산하여 저장하는 것이다.Here, the calculated output data

and

It is to distribute and store.

를 분할하여 변환조각

로 변환하는 경우, 각 변환조각 크기의 합과 평문의 크기는 동일하고 변환 후 변환조각에 대해 다시 AONT를 수행하여 분할된 변환조각의 개수나 크기를 다시 설정할 수 있다. 도 2의 AONT 방식은 다음과 같은 특성이 있다.Plain text in Figure 2

Convert to and fragment

In the case of conversion to, the sum of the transform pieces and the size of the plain text are the same, and after conversion, AONT is performed on the converted pieces to reset the number or size of the divided pieces. The AONT method of FIG. 2 has the following characteristics.

The total sum of the divided pieces of data is equal to the size of the original data.

-Compared to the conventional secret distribution method, the fragments can be made very small.

-You can freely set the number or size of transform fragments.

-The processing speed is fast because the total capacity after large data conversion is small.

These characteristics make it suitable for large data processing, and the existing secret dispersion method increases the total capacity according to the number of distributed information, whereas in the case of AONT method, the processing speed is fast because the partition information and the original data are the same size. In addition, the capacity can be optimized by variable partitioning, thereby lowering the system cost.

Conventional fragmentation and encryption schemes protect privacy between each attribute due to fragmentation, and protect confidentiality by encrypting attributes excluding fragment attributes. However, as the number of fragmentation causes duplicate attributes, the storage space increases. In the proposed method, when encrypting the duplicate attributes, the AONT structure is used for fragments of various sizes to reduce the storage space.

The present invention proposes a method of reducing storage space by using AONT-based variable fragments.

When physical fragments are performed in the fragmentation and encryption schemes, the storage schema increases due to the fact that each fragment's relational schema duplicates the same attributes. Therefore, when a physical fragment is performed, the problem of capacity increase is solved by using AONT rather than encryption.

Figure 3 is a flow chart of the medical information on the hospital information system for applying the present invention when storing the attribute information of {patient number, patient name, date of birth, address, disease name, doctor} occurs when the patient receives medical care in the hospital In addition, privacy can be enhanced by separating the associations between attributes in a fragmented and AONT fashion.

3 is a block diagram of fragmentation and AONT performed using a security agent (SA) 320 and managing a share DB when the patient's medical information is stored, and the flow of medical information is as follows. .

① When a patient visits a hospital and receives medical treatment, medical information is created.

② The generated treatment information is transmitted to the SA 320 through the server 310.

③ The SA 320 fragments the medical care information according to the fragmentation attribute and performs AONT on the remaining attributes except the fragment attribute. That is, the AONT method divides the data into a plurality of transform pieces having variable sizes.

④ Conversion fragments generated after AONT are transmitted to the server 310 and stored in the share DB (322).

⑤ The server 310 stores the transform fragment in the DB 312 according to the fragment.

The method of the present invention proposes a fragmentation and AONT based privacy enhancement method focusing mainly on the operation of the procedure ③.

Fragmentation and AONT

In the existing fragmentation and encryption method, the storage space is increased due to the duplication of attributes. However, the use of the present invention reduces the storage space even if the attributes are duplicated.

To use AONT in the fragmentation and encryption scheme mentioned in the prior art, [Definition 3] is redefined as follows.

이다. 각 단편속성 집합

에 대한 물리적 단편은 관계 스키마

및

이다. [Definition 3 ''] R Relation Schema, R's Short Calculation

to be. Each fragment attribute set

Physical Fragment for Relationship Schema

And

to be.

Here, AONT means splitting all the attributes of R which do not belong to the fragment by AONT method, one of the divided transform fragments is stored in f _i ^s , and the other transform fragments are stored in Share database s _i ^s .

는 관계스키마 r∈R이고, F_i에 대한 평문 튜플

는 F_i ^e의 튜플

로 맵핑된다. 그리고 상기 각 단편 속성 집합 F_i를 제외한 속성은

및

로 맵핑된다. 여기서, 물리적 단편 f_i ^s는 다음과 같은 연산을 수행한다.Each fragment attribute set

Is the relational schema r∈R, and the plaintext tuple for F _i

Is a tuple of F _i ^e

Is mapped to. And the property except for each fragment property set F _i is

And

Is mapped to. Here, the physical fragment f _i ^s performs the following operation.

For example, the physical fragment f ₁ ^s is described as follows.

R = {Patient number, patient name, date of birth, address, illness name, doctor}

Fragment set F of R is

-F = {{Patient name}, {date of birth, address}, {name of doctor, doctor}}

F ₁ = {patient name}

-F ₂ = {date of birth, address}

F ₃ = {name of disease, doctor}.

를 계산하기 위해 R-F₁는 {환자번호, 환자이름, 생년월일, 주소, 병명, 담당의사} - {환자이름}을 뺀 속성들의 집합에 대해 AONT 방식이 수행된다. 여기서, 분할된 변환조각은 3개인 경우, 하나는

, 나머지 2개는

에 저장한다.Physical fragments

In order to calculate RF ₁ , AONT is performed on a set of attributes minus {patient number, patient name, date of birth, address, disease name, doctor}-{patient name}. Here, if there are three divided fragments, one is

, The other two

.

(t[{환자번호, 환자이름, 생년월일, 주소, 병명, 담당의사}-{환자번호}])

(t [{patient number, patient name, date of birth, address, name of illness, doctor}-{patient number}])

As described above, when the physical fragment for each attribute is performed as shown in FIG. 4.

Composition of Conversion Share

When the transform fragment is stored in each relation schema except for the fragment attribute, only one transform fragment is stored in f _i ^s and the remaining transform fragment is stored in Share DB s _i ^s (FIG. 5).

Fragmentation is performed and stored in Share DB as follows.

① The fragment attribute is mapped to f _i ^s and the remaining attributes are distributed by AONT method.

(2) The divided transform fragments are mapped to each f _i ^s according to the fragment attribute. At this time, only one of the divided transform pieces is stored in f _i ^s of the DB 312.

③ The remaining conversion pieces are mapped and stored in the Share DB 322 and can be restored by a combination of f _i ^s and Share DB 322 at the time of restoration. Here, the share DB 322 is composed of only the conversion pieces, so the search and restoration is impossible. Therefore, DB search can be done through f _i ^s and restored.

By the AONT feature, each plaintext is variably divided by the user to generate a transform fragment. For example, when the user performs the AONT method for a plain text having a size of 10KB, the user may generate three transform pieces each having sizes of 2KB, 4KB, and 4KB. At this time, the 2KB conversion pieces (share 1) are used for fragmentation and the remaining 4KB conversion pieces (share 2, share 3) are used when stored in the Share DB 322. If the conversion piece size is 4KB and the storage space is 24, a storage space capacity of 96KB is required (FIGS. 6 and 7).

For example, when the AONT method of the patient's medical information is performed by SA, only one plaintext attribute and one transform fragment are stored in f _i ^e according to the fragmentation attribute as shown in FIG. 4, and the other transform fragments are stored as shown in FIGS. 6 and 7. It is stored in the Share DB (322).

After redefining [Definition 3], if the physical fragment is made as shown in FIG. 4, the actual storage space is reduced as shown in Table 8, Table 9, and Table 10. FIG. As a result, when fragmentation is performed using the AONT method, storage space is required as much as 304KB. If the combined size of the transform pieces (192KB) is added, a total of 496KB of storage space is required.

Table 8: Share Size (40KB) + Plain Text Size (40KB) = 80KB

Table 9: Share Size (32KB) + Plain Text Size (80KB) = 112KB

Table 10: Share Size (32KB) + Plain Text Size (80KB) = 112KB

AONT patient
name patient
number birth year
Month day address Disease charge
doctor 2 KB 2 KB 2 KB 2 KB 2 KB 10 KB 2 KB 2 KB 2 KB 2 KB 2 KB 10 KB 2 KB 2 KB 2 KB 2 KB 2 KB 10 KB 2 KB 2 KB 2 KB 2 KB 2 KB 10 KB

AONT date of birth address patient
number patient
name Disease charge
doctor 2 KB 2 KB 2 KB 2 KB 10 KB 10 KB 2 KB 2 KB 2 KB 2 KB 10 KB 10 KB 2 KB 2 KB 2 KB 2 KB 10 KB 10 KB 2 KB 2 KB 2 KB 2 KB 10 KB 10 KB

AONT Disease doctor patient
number patient
name birth year
Month day address 2 KB 2 KB 2 KB 2 KB 10 KB 10 KB 2 KB 2 KB 2 KB 2 KB 10 KB 10 KB 2 KB 2 KB 2 KB 2 KB 10 KB 10 KB 2 KB 2 KB 2 KB 2 KB 10 KB 10 KB

Comparison with the old way

(1) efficiency problems of traditional methods

In the conventional method, when fragmentation is performed, duplicate attributes exist in each fragment, and the rest is encrypted except the plaintext attribute due to fragmentation. As such, the existing scheme has a problem in that the database capacity increases due to encrypted redundant data through fragmentation.

The following example is described.

(Example 1) Assuming that the storage size of each attribute is 10 KB in a table of 6 rows by 6 columns, the storage space of this table is 240 KB. In this case, if the existing table is fragmented into three tables, the total storage space is increased to 720 KB due to duplicate attributes. As a result, three fragments should use three times as much storage space as existing storage.

As shown in (Example 1), it can be seen that the data storage space increases according to the number of fragments. Here, in an environment where a large amount of data is used for cloud computing services, efficient storage and management of data is required. However, the existing method secures confidentiality by separating the connection of each property, but there is a problem of efficiency of storage space for managing large data.

In order to solve this problem, the present invention minimizes database capacity by using the AONT method. The proposed method of the present invention protects confidentiality and enables efficient management of data storage space through data separation.

If AONT is used instead of the encryption method stored for duplicate attributes, the size can be variably stored for each attribute according to AONT characteristics. In addition, even when the AONT method is performed, the size of the attribute is the same, which is more efficient than the encryption method.

(Example 2) Assuming that the storage size for each attribute is 10KB in a four-row × six-column table with six attributes, the storage space for the table is 240KB. In case of the proposed method, if AONT is performed according to each property, there are 3 shares divided, and each share size is set to 2KB, 4KB, 4KB, 80KB in the first fragment, 112KB in the second fragment, and 112KB in the third fragment. The rest of the Share DBs require 96 KB of storage. As a result, the total storage space is 496 KB, which is 224 KB smaller than that of the conventional method, and thus the storage space is reduced in the case of the method of the present invention in the same number of fragments.

8 shows a storage space according to an increase in the number of fragments and a storage space of the proposed method.

As shown in FIG. 8, when one fragment is stored, the storage space is the same, but as the number of fragments increases, the storage space increases. However, in the case of the proposed method, even if the number of fragments increases, the storage space is reduced compared to the existing method.

Although the present invention has been described as a specific preferred embodiment, the present invention is not limited to the above-described embodiments, and the present invention is not limited to the above-described embodiments without departing from the gist of the present invention as claimed in the claims. Anyone with a variety of variations will be possible.

310: server 312: DB
320: SA 322: Share DB

Claims (3)

i) performing fragmentation on the information to be encrypted according to the fragmentation property through the security agent and dividing all the properties except the fragment property into a plurality of transform pieces having variable sizes in the All-Or-Nothing Transform (AONT) method; ; And
ii) one of the plurality of translation pieces is stored in a first database via a security agent and the remaining translation pieces are mapped and stored in a second database via a server,
In step (i), AONT means splitting all the attributes of the relational schema R which does not belong to the fragment in the AONT manner, and the short edit sum of the relational schema R is

Each fragment attribute set

Physical Fragment for Relationship Schema

And

Is,
Each fragment attribute set

Is the relational schema r∈R, and the plaintext tuple for F _i

Is a tuple of F _i ^e

Mapped to, except for each fragment attribute set F _i ,

And

And the physical fragment f _i ^s is

,

, And

Calculated by
One of the fragments divided in the step (ii) is stored in f _i ^s of the first database through the security agent, the remaining conversion fragments are stored in s _i ^s of the second database through the server How to strengthen your base privacy.
The method according to claim 1, wherein the first database and the second database is a database for a security agent and a database for a server, respectively. delete

Images