WO2018152904A1

WO2018152904A1 - Method for processing security outsourcing data of relational database supporting ciphertext data operation

Info

Publication number: WO2018152904A1
Application number: PCT/CN2017/077109
Authority: WO
Inventors: 王晓菲; 张玉清; 伍前红
Original assignee: 中国科学院大学; 北京航空航天大学
Priority date: 2017-02-24
Filing date: 2017-03-17
Publication date: 2018-08-30
Also published as: CN106934301A; CN106934301B

Abstract

Disclosed is a method for processing security outsourcing data of a relational database supporting a ciphertext data operation. The method involves: 1) a database owner utilizing a private key to encrypt a database to be outsourced and then generate an outsourced ciphertext database, and sending same to a cloud service provider; 2) an authorized terminal user submitting, to the database owner, a plaintext data operation request in the form of an SQL statement; 3) the database owner converting a plaintext SQL statement of the data operation request into a ciphertext SQL statement set, and sending same to the cloud service provider; 4) the cloud service provider directly executing, on the outsourced ciphertext database, the ciphertext SQL statement, and if there is a query request, returning an obtained ciphertext query result to the database owner; and 5) the database owner receiving and decrypting the ciphertext query result, and returning a plaintext record to the terminal user. The present invention guarantees the data confidentiality and effective utilization of an outsourced relational database.

Description

Relational database security outsourcing data processing method supporting ciphertext data operation

Technical field

The invention relates to a relational database security outsourcing data processing method for supporting ciphertext data operations, which belongs to the field of cloud computing security and database security.

Background technique

At present, outsourcing private relational databases to cloud service providers can bring huge convenience and benefits to business groups and individual users. At the same time, however, the security and privacy issues of outsourced databases are becoming more and more serious. Sensitive information of cloud computing users (such as financial transaction records, personal medical files, etc.) is facing different levels of abuse and disclosure.

Public cloud environments typically use cloud-driven database encryption to protect against malicious attacks by external attackers, but it is still difficult to prevent internal attacks from untrusted cloud service providers. Therefore, in order to avoid the problem of privacy leakage that cloud encryption may bring, cloud database encryption should be implemented entirely by the database owner, thus truly realizing the service mode of “encrypted outsourcing and post-query decryption”.

Client encryption mainly faces two problems: first, which encryption algorithm is used to encrypt the relational database; and second, how to perform data operations on the encrypted ciphertext relational database. Related research has achieved a series of important results, including full homomorphic encryption technology, partial homomorphic encryption technology. More work is done around one or more of ciphertext domain equivalent queries, range queries, aggregate queries, and fuzzy queries, but the types of data operations it supports are very limited. Some comprehensive ciphertext database query schemes can support some SQL-based query operations, but there are four shortcomings:

(1) Different types of SQL statements need to call different encryption technologies to complete the interpretation, resulting in reduced database service performance;

(2) The cloud service provider cannot directly execute the SQL statement, and the cloud service model is difficult to completely get rid of the dependence on the trusted agent;

(3) The cloud service provider needs to know the decryption key before performing the data operation, and has not completely overcome the security risk of cloud encryption;

(4) The database access mode of the end user and the relational structure of the cloud database are forced to change.

Summary of the invention

In view of the problems existing in the prior art, an object of the present invention is to provide a relational database security outsourcing data processing method that supports ciphertext data operations. This method protects the data confidentiality and effective data utilization of the relational database in the outsourcing process. The client encryption technology adopted fundamentally solves the security problem of cloud encryption. 1 is a block design and overall architecture of the present invention.

The technical solution used by the present invention to solve the technical problem thereof is: a relational database security outsourcing data processing method (including a query request, an insert request, an update request, and a delete request) that supports ciphertext data operations. This approach works between end users, database owners, and cloud service providers. The relational database is encrypted by the database owner before outsourcing, and then the cloud service provider directly executes the SQL-based full-type data operation on the outsourced ciphertext database, and the returned query result is decrypted by the database owner instead of the end user. Specifically, the method flow of the present invention is:

1. The database owner outsources the relational database security to the cloud service provider:

(1) The database owner generates a private key;

(2) The database owner encrypts the outsourced database through the cryptographic module to generate an outsourced ciphertext database and sends it to the cloud service provider;

(3) The cloud service provider receives and stores the outsourced ciphertext database in place;

(4) The database owner creates a user-defined function at the cloud service provider.

2. The end user requests data protection for the privacy protection of the relational database:

(1) The terminal user logs into the database system and submits the plaintext SQL statement to the database owner;

(2) The database owner determines whether the query request has been authorized according to the access control policy, and if the access is illegal, the operation is rejected;

(3) The database owner converts the plaintext SQL statement into a ciphertext SQL statement set by the conversion module and sends it to the cloud service provider;

(4) The cloud service provider invokes the user-defined function, executes the ciphertext SQL statement in the ciphertext database, and returns the ciphertext query result to the database owner;

(5) The database owner receives and decrypts the query result set and returns a plaintext record to the end user.

The security outsourcing data processing method of the present invention includes the following five modules:

(1) A cryptographic module (OPEA algorithm) for encrypting or decrypting an outsourced database. The encryption function of the OPEA algorithm has order-preserving and additivity, which can maximize the operability of data. The ciphertext domain of the OPEA algorithm is divided into a plurality of ciphertext partitions arranged in order. FIG. 2 is a mapping structure of the cryptographic module, and upper and lower boundaries of the ith ciphertext partition are respectively defined as U _i and L _i . Specifically, the one-to-many mapping structure used in the figure maps a single plaintext value into multiple ciphertext values in a ciphertext partition, which is used to construct an encryption function of the OPEA algorithm, thereby improving the security of the cryptographic module. Sex to defend against ciphertext attacks, statistical attacks, and weakly selected plaintext attacks.

(2) The conversion module (SQL-Translator interpreter) is used to convert the plaintext SQL statement to be requested, and at the same time ensure that the ciphertext SQL statement obtained after the conversion can be executed normally in the ciphertext database. FIG. 3 is a processing flow of the conversion module.

(3) User-defined functions, deployed in cloud service providers, are mainly used for operations such as equivalence comparison, aggregation comparison, and value calculation and string segmentation in the cloud.

(4) Encryption rules, the integer data is directly encrypted using the OPEA algorithm; when the floating point data is encrypted, the ciphertext space is divided according to the minimum precision of the plaintext space; the character data is converted to the whole matching unit according to the minimum matching unit selected by the end user. After the type, it is encrypted, filled and spliced. Specific to the outsourced relational database, the plaintext space is a finite set of all the data in the database.

(5) Access control, using different keys to encrypt data tables or data fields with different security levels; terminal users who have obtained legal access authorization can log in to the database system through the shared user password; the database owner is based on the data of the currently logged-in user. Access rights get the corresponding encryption and decryption key.

The beneficial effects of the invention are:

(1) The present invention solves two main problems of client encryption, and ensures data confidentiality and effective utilization of the outsourced relational database.

(2) The overall architecture of the present invention is applicable to an unmodified commercial cloud database platform such as Microsoft Windows Azure SQL Database.

(3) The present invention only needs to encrypt the database once, so the storage overhead is lower, and the relationship between the data access mode and the cloud database is preserved.

(4) The OPEA algorithm of the cryptographic module of the present invention can resist ciphertext attacks, statistical attacks, and weakly selected plaintext attacks. The average time complexity of the simplified boundary generation sub-algorithm, the encryption sub-algorithm and the decryption sub-algorithm reaches O(T), O(1) and O(log T), respectively, and T is the maximum value of the plaintext domain, which is basically superior to the mainstream. Symmetric cryptographic algorithm. In addition, the comparison operation time and network communication overhead of the OPEA algorithm are neglected. The noise growth problem in the ciphertext domain can be effectively controlled in a constant time by adjusting the key.

(5) The conversion module of the present invention supports the current most types of SQL data operations, and includes data operation statements such as equivalence query, range query, aggregate query, fuzzy query, insert, update, and delete. The conversion time of the SQL-Translator interpreter is in the order of microseconds and satisfies the data conversion requirements of any type and length.

(6) The present invention directly executes the ciphertext query without decrypting the outsourced data, and the query result is accurate, without error judgment, and no miss judgment. With the same query precision, the client post-processing overhead is low, and the overall query efficiency is high.

DRAWINGS

Figure 1 shows the overall architecture of the system. The module design, data flow direction and range of roles of the present invention are given.

Figure 2 is a cryptographic module mapping structure diagram. The division of the OPEA algorithm, the division method of the interval, and the plaintext domain are given. The mapping relationship of the ciphertext domain.

Figure 3 is a flow chart of the conversion module processing. The workflow of the SQL-Translator interpreter is given, including the core conversion module and the sub-conversion module.

Figure 4 is an example of a database to be outsourced;

Figure 5 is an example of an outsourced ciphertext database;

Figure 6 is an example of a ciphertext query result;

FIG. 7 is an example of the result of the plaintext query after decryption.

detailed description

The invention will now be further described with reference to the accompanying drawings.

Figure 1 is the overall architecture of the system, mainly related to the three core service modules of the cryptographic module, the conversion module and the user-defined function. The specific method includes the following steps:

1. Figure 2 is a cryptographic module mapping structure diagram. The implementation method of the cryptographic module includes:

(1) Design and implement an addendum encryption algorithm OPEA, the discrete ciphertext domain is composed of sequential ciphertext partitions, and its encryption function E:X→Y satisfies the following two conditions (X and Y are respectively OPEA) Plaintext and ciphertext space; specific to outsourced database applications, plaintext space is a collection of all data in the text database, and ciphertext space refers to a collection of all data in the ciphertext database):

a)

a<b if and only if E(a) <E(b);

b)

If a+b<c, then E(a)+E(b)<E(c).

The algorithm consists of three sub-algorithms:

a) Boundary generator algorithm BoundaryGen, input random non-negative integer set R={R _i } and random positive integer σ>max _1≤i≤T {R _i }-R ₁ as key, calculate and output according to key The lower boundary set of the ciphertext partition of the ciphertext domain L={L _i }={L]}(1≤i≤T) and the upper boundary set U={U _i }={U[i]}(1≤ i ≤ T). Wherein, R _i is a set of random numbers in the i th element, L _i and L [i] is the i-th lower boundary ciphertext partitions, U _i, and the U-[i] is the i-th upper boundary ciphertext partition . The sub-algorithm first sets the first ciphertext partition [L ₁ , U ₁ ], with L[1]=σ and U[1]=L[1]+R ₁ , and then iteratively delimits the subsequent ciphertext partitions. The boundary value is required to satisfy L[t]=max _1≤i<t {U[i]+U[ti]} and U[t]=L[t]+R _t (2≤t≤T). Here T is the maximum value of the plaintext domain, that is, the maximum possible value of all the data in the plaintext database.

b) Encryption sub-algorithm Enc, input boundary set and positive integer plaintext m, output random set of {L[m], L[m]+1,...,U[m]} as ciphertext c=E(m) . The encryption sub-algorithm is based on the ciphertext partition when encrypting the plaintext value m. For the boundary sets L and U, the encryption function E(·) will select from the set {L[m], L[m]+1,...,U[m]} and return a random number as the ciphertext value c.

c) Decryption sub-algorithm Dec, input boundary set and ciphertext c, and output the ciphertext partition number of ciphertext c as a plaintext value by means of a halved search algorithm.

(2) Simplify the boundary generation sub-algorithm of the OPEA algorithm, and arrange the keys R _i (1 ≤ _i ≤ T) in non-descending order to obtain a linear boundary function:

(3) The OPEA algorithm is extended to determine the ciphertext relationship E'(a) + E'(b) > E'(c), where E': X → Y' is an encryption function of the spreading algorithm.

a) The extended boundary generation sub-algorithm BoundaryGen' requires the key σ>3·max _1≤i≤T {R _i } to calculate and output the extended upper boundary set U'={U'[ according to the key. i]}(1≤i≤T) and the extended lower boundary set L'={L'[i]}(1≤i≤T). The extended sub-algorithm delimits the first ciphertext partition U'[1]=σ and L'[1]=U'[1]-R ₁ , and the subsequent ciphertext partition should satisfy the upper boundary U'[t]=min _{1≤ i<t} {L'[i]+L'[ti]} and the lower boundary L'[t]=U'[t]-R _t (R _t <U'[t]-U'[t-1] , 2 ≤ t ≤ T), the linear boundary function is:

b) The extended cipher sub-algorithm and the extended decryption sub-algorithm are similar to Enc and Dec.

(4) The database owner encrypts all data items using the OPEA algorithm, and encrypts the data table name, column name, and the like using an anti-collision hash function or a symmetric cryptography technique.

2. User-defined function implementation methods include:

(1) The database owner selects a random integer x∈[max{R ₁ , R ₂ }, σ+R ₁ ) according to the key and the plaintext to be compared in the query request, and sends it to the cloud service provider, where R ₁ = U[value ₁ ]-L[value ₁ ] and R ₂ =U[value ₂ ]-L[value ₂ ] are the ciphertext partition lengths corresponding to the plaintext values value ₁ and value ₂ to be compared, respectively.

(2) Design and implement the equivalence comparison function EqualityCom(x, E(value ₁ ), E(value ₂ )), and the decision step is:

a) If the distance between E(value ₁ ) and E(value ₂ ) is not greater than x, the plaintext values are equal, and the function outputs 0, otherwise step b) is continued;

b) If E(value ₁ ) is greater than E(value ₂ ), then value _{1 is} greater than value ₂ and the function outputs 1; if E(value ₁ ) is less than E(value ₂ ), then value _{2 is} greater than value ₁ and the function outputs - 1.

(3) Designing an aggregate comparison function

L[value], U'[value]),

Is a collection of data items in a database field, SUM is a summation function, the plaintext value corresponds to a constant value in the query condition, L[value] and U'[value] are respectively the lower boundary of the ciphertext partition corresponding to the value Extend the upper boundary. Request here

And R is arranged in non-descending order, and the decision step is:

a) if

Not greater than L[value] and

Not less than U'[value], then

Equal to value, the function outputs 0, otherwise continue to perform step b);

b) if

Greater than U'[value], then

Greater than value, the function outputs 1;

Less than L[value], value is greater than

The function outputs -1.

(4) Deploy security and value calculation protocols between the database owner and the cloud service provider with the user-defined function SumEqualityCom. Assuming that Att is a collection of data items for a database field, the protocol steps are:

a) The cloud service provider calculates SUM(E(Att)) and SUM(E'(Att)) and sends the ciphertext accumulation result to the database owner;

b) The database owner decrypts d=Dec(SUM(E(Att)), L) and d'=Dec(SUM(E'(Att)), U');

c) If d is equal to d', the database owner knows SUM(Att)=d and returns E(SUM(Att))=SUM(E(Att)) to the cloud service provider. The agreement is over.

d) Otherwise, the database owner will continue to try to find SumEqualityCom(SUM(E(Att)), SUM(E'(Att)), L[d+i], U'[d+i])=0 The established integer i (0 ≤ i ≤ d'-d), so that SUM (Att) = d + i is known, and E (d + i) is returned to the cloud service provider. The agreement is over.

(5) Design and implement the string splitting function Split(str, delimiter), str is the ciphertext search pattern string of the LIKE operator to be split, delimiter is the separator, and the split result is stored and returned by the Table type.

3. Figure 3 is a flow chart of the conversion module processing. The implementation method of the conversion module includes:

(1) Design and implement the SQL-Translator interpreter for the outsourced database encrypted by the OPEA algorithm, using standard Transact-SQL data manipulation statements as input. Here, it is assumed that tbl is a data table name of Ming Library, and Att, Att ₁ and Att ₂ are data item sets of a field of Ming Library, val, val ₁ and val ₂ are constant values, and * ^C is a database object corresponding to the ciphertext field.

(2) Convert the database entity in the plaintext data operation statement, the conversion rule is:

a) The data table name and column name are anonymized using an anti-collision hash function or a symmetric encryption technique;

b) Constant values are converted using the OPEA algorithm.

(3) Dividing the converted data operation statement into a minimum conversion unit, which includes a complete predicate expression or clause structure.

(4) The corresponding sub-conversion module is called in turn for the minimum conversion unit. The sub-conversion module i corresponds to the minimum conversion unit i, and the conversion rule of the minimum conversion unit i is provided, and is converted into a corresponding ciphertext SQL clause. For the four types of data operation statements of query, insert, update, and delete, Table 1 details the conversion rules of the minimum conversion unit.

Table 1 shows the conversion rules for the minimum conversion unit.

The conversion rules in Table 1 are described as follows:

■ For query operation statements: Query operation statements of the form SELECT<Att ₁ , Att ₂ ,...>FROM<tbl> are directly converted to

SELECT<Att ₁ ^C ,Att ₂ ^C ,...>FROM<tbl ^C >

The conversion rules of the remaining clauses are as follows, where the integer x ∈ [max 1 _≤ _i _{≤ T} {R _i }, σ + R ₁ ) is specified:

WHERE clause

A comparison operator expression of the form Att ₁ =Att ₂ is directly converted to EqualityCom(x,Att ₁ ^C ,Att ₂ ^C )=0.

The comparison operator expression of the form Att ₁ >Att ₂ is directly converted to EqualityCom(x,Att ₁ ^C ,Att ₂ ^C )>0.

A comparison operator expression of the form Att ₁ <Att ₂ is directly converted to EqualityCom(x, Att ₁ ^C , Att ₂ ^C )<0.

Similarly, the comparison operator <>,! =,! >,<=,! The conversion rule of <, >= can be obtained by combining the above rules.

The BETWEEN operator expression of the form Att BETWEEN Att ₁ AND Att ₂ is directly converted to

EqualityCom(x,Att ^C ,Att ₁ ^C )>=0∧EqualityCom(x,Att ^C ,Att ₂ ^C )<=0

The NOT BETWEEN operator expression of the form AttNOT BETWEEN Att ₁ AND Att ₂ is directly converted to

EqualityCom(x,Att ^C ,Att ₁ ^C )<0∨EqualityCom(x,Att ^C ,Att ₂ ^C )>0

The IN operator expression of the form Att IN (Att ₁ , Att ₂ , ...) is directly converted to

EqualityCom(x,Att ^C ,Att ₁ ^C )=0∨EqualityCom(x,Att ^C ,Att ₂ ^C )=0∨...

The NOT IN operator expression of the form AttNOT IN (Att ₁ , Att ₂ , ...) is directly converted to

EqualityCom(x,Att ^C ,Att ₁ ^C )! =0∧EqualityCom(x,Att ^C ,Att ₂ ^C )! =0∧...

An IS operator expression of the form Att IS NULL is directly converted to

EqualityCom(x,Att ^C ,NULL ^C )=0

An IS NOT operator expression of the form Att IS NOT NULL is directly converted to

EqualityCom(x,Att ^C ,NULL ^C )! =0

For the LIKE operator of the form Att[NOT]LIKE pat[ESCAPE esch], the following conversion methods are gradually adopted: add the temporary column Att_Match for Att; escape the wildcard according to the escape character esch and encrypt the ordinary characters with the OPEA algorithm; The layer cursor matches the length of the data item; the search mode string pat is divided into 3 parts by the Split function, and the matching conditions of the start, the middle, and the end part are respectively generated, and each matching condition is spliced using AND or OR; the inner layer cursor is declared to match the current data. The ciphertext string of the item; updates the Att_Match column and the query criteria. The conversion result of the LIKE operator consists of all the SQL statements required for the above conversion steps.

For a subquery of the form s ₁ WHERE EXISTS (s ₂ ), where s ₁ and s ₂ are SELECT statements, the following conversion methods are gradually adopted: converting the inner query s ₂ , and the query result is stored in the temporary table #INTER_TABLE2; Convert the outer query to s ₁ WHERE EXISTS(SELECT*FROM #INTER_TABLE2). The conversion result of the subquery statement is composed of all the SQL statements required for the above conversion step. Similarly, the conversion rules for NOT EXISTS subqueries and operator subqueries are the same.

For the query conditions con ₁ and con ₂ , there are the following conversion rules:

The combined query condition of the form con ₁ AND con ₂ is directly converted to con ₁ ^C ∧con ₂ ^C .

The combined query condition of the form con ₁ OR con ₂ is directly converted to con ₁ ^C ∨con ₂ ^C .

ORDER BY clause

The ORDER BY clause of the form ORDER BY Att ₁ , Att ₂ ,...[ASC|DESC] is directly converted to

ORDER BY Att ₁ ^C ,Att ₂ ^C ,...[ASC|DESC]

During the execution of the ciphertext SQL statement, the cloud service provider first determines the records with the same attribute values in Att ₁ ^C using the EqualityCom function, and then sorts them according to Att ₂ ^C.

GROUP BY clause

For the GROUP BY clause of the form GROUP BY Att, the following methods are used to convert: the self-join query is performed on the Att column, and the result of satisfying the query condition is stored in the temporary table #TEM; the query is converted into the temporary table query GROUP BY Att ^C. The result of the conversion of the GROUP BY clause consists of all the SQL statements required for the above conversion steps.

HAVING clause

An aggregate function of the form MIN(Att) is directly converted to MIN(Att ^C ).

An aggregate function of the form MAX(Att) is directly converted to MAX(Att ^C ).

An aggregate function of the form COUNT(Att) is directly converted to COUNT(Att ^C ).

For an aggregate function of the form SUM(Att), it is converted into a security and value calculation protocol between the cloud service provider and the database owner, by which the E(SUM(Att)) is finally calculated.

An aggregate function of the form AVG(Att) is directly converted to E(SUM(Att))/COUNT(Att ^C ).

■ For insert operation statements: insert operation statements of the form INSERT INTO<tbl>(<Att ₁ ,Att ₂ ,...>)VALUES(val ₁ ,val ₂ ,...) are directly converted to

INSERT INTO<tbl ^C >(<Att ₁ ^C ,Att ₂ ^C ,...>)VALUES(E(val ₁ ),E(val ₂ ),...)

Similarly, for a subquery statement contained in an insert operation statement, the conversion rule is the same as the query operation statement.

■ For the update operation statement: the update operation statement of the form UPDATE<tbl>SET<Att>=val is directly converted to

UPDATE<tbl ^C >SET<Att ^C >=E(val)

Similarly, for the update condition specified by the FROM clause and the WHERE clause in the update operation statement, the conversion rule is the same as the query condition of the query operation statement.

■ For the delete operation statement: the delete operation statement of the form DELETE FROM <tbl> is directly converted to

DELETE FROM<tbl ^C >

Similarly, for the delete condition specified by the WHERE clause in the delete operation statement, the conversion rule is the same as the query condition in the query operation statement.

(5) Splicing and arranging the converted ciphertext SQL clauses to obtain a complete ciphertext SQL statement and retaining the same semantics. For example, the following plaintext query:

SELECT Att ₁ FROM tbl WHERE Att ₁ =Att ₂ ORDER BY Att ₁

GROUP BY Att ₁ HAVING SUM(Att ₂ )>val

After the SQL-Translator conversion, the following ciphertext query statements are obtained by splicing and arranging:

SELECT Att ₁ ^C FROM tbl ^C WHERE EqualityCom(x,Att ₁ ^C ,Att ₂ ^C )=0 ORDER BY Att ₁ ^C GROUP

BY Att ₁ ^C HAVING SumEqualityCom(SUM(E(Att ₂ )), SUM(E'(Att ₂ )), L[val], U'[val])>0

(6) Output ciphertext SQL statements that can be executed directly on the outsourced ciphertext database.

A specific embodiment of the present invention is given below.

1. Assume that the database owner's plaintext database consists of a data table named EXAMPLE. The table contains two plaintext fields (or plaintext attribute columns) with column names C_CUSTKEY and C_NATIONKEY. There are 5 plaintext records (ie 5 rows of data) in the table, as shown in Figure 4.

2. The database owner encrypts the plaintext database by means of the cryptographic module, and sends the encrypted ciphertext database to the cloud service provider. The plaintext data table named EXAMPLE in the plaintext database corresponds to the ciphertext data table named [8e3b72508e05135569ace4ed9b96d137] in the ciphertext database, and there are 5 ciphertext records in the table, as shown in Fig. 5; wherein d0d0a493e28066e951fa7a980e81cb05 and 4eed023b3ef2a4fd3b3fae34f828ba5e are respectively used for anti-collision The ciphertext column name obtained by encrypting the plaintext column names C_CUSTKEY and C_NATIONKEY by the hash function. The

ciphertext data items

121 and 4423 of the first line are respectively ciphertext values obtained by encrypting the

plaintext data items

1 and 15 of the first line of the plaintext database one by one using the OPEA algorithm.

3. Assume that the end user proposes the following query operation request (ie plaintext SQL statement):

SELECT*FROM EXAMPLE WHERE C_NATIONKEY>10

The end user requests to query the data record with a C_NATIONKEY value greater than 10 from the EXAMPLE table. Where C_NATIONKEY>10 is the query condition, and the query result should return two plaintext records (1, 15) and (2, 13).

4. The database owner converts the plaintext SQL statement by means of the conversion module to obtain the ciphertext SQL statement:

SELECT*FROM[8e3b72508e05135569ace4ed9b96d137]

WHERE EqualityCom(100,[4eed023b3ef2a4fd3b3fae34f828ba5e], 2891)>0

Among them EqualityCom (100, [4eed023b3ef2a4fd3b3fae34f828ba5e], 2891) is a user-defined function. EXAMPLE ^C = [8e3b72508e05135569ace4ed9b96d137], x=100, C_NATIONKEY ^C = [4eed023b3ef2a4fd3b3fae34f828ba5e], E(10)=2891.

5. The database owner sends the ciphertext SQL statement to the cloud service provider. Responsible for ciphertext data by cloud service provider The query is executed in the library, and the user-defined function EqualityCom needs to be called during execution. After the query is completed, the ciphertext query result that meets the query conditions is obtained, as shown in FIG. 6.

6. The cloud service provider returns the ciphertext query result to the database owner. The database owner decrypts it by means of the cryptographic module, and finally obtains the plaintext query result, as shown in Figure 7:

7. The database owner returns the plaintext query result to the end user. The outsourced database query process ends.

Claims

A relational database security outsourcing data processing method supporting ciphertext data operation, the steps of which are:

(1) The database owner encrypts the outsourced database with the private key to generate an outsourced ciphertext database and sends it to the cloud service provider;

(2) The end user logs in to the database system through its password, and submits a plaintext data operation request to the database owner in the form of a SQL statement;

(3) The owner of the database determines whether the data operation request has been authorized according to the access control policy, and if the access is illegal, the operation is rejected; if the access is legal, the plaintext SQL statement of the data operation request is converted into the ciphertext SQL. a collection of statements and sent to the cloud service provider;

(4) the cloud service provider directly executes the ciphertext SQL statement on the outsourced ciphertext database; if it is a query request, returns the obtained ciphertext query result to the database owner;

(5) The database owner receives and decrypts the ciphertext query result, and returns a plaintext record to the end user.
The method according to claim 1, wherein in the step (1), the method for encrypting the outsourced database is: designing and implementing an addendaic encryption algorithm OPEA, and treating all data items in the outsourced database. Encryption is performed, and the data table name and column name in the outsourced database are encrypted by using an anti-collision hash function or a symmetric cryptographic technique; wherein the sequence encryption algorithm OPEA uses a sequential ciphertext partition to form a discrete ciphertext domain, and the encryption is performed. The function E:X→Y satisfies the conditions a) and b); X and Y are the plaintext and ciphertext spaces of the OPEA algorithm, respectively;

a)
b∈X, a<b if and only if E(a) <E(b);

b)
b, c∈X, if a+b<c, then E(a)+E(b)<E(c).
The method according to claim 2, wherein the order-preserving encryption algorithm OPEA comprises a boundary generation sub-algorithm, an encryption sub-algorithm and a decryption sub-algorithm, wherein:

a) The boundary generation sub-algorithm is: input a random non-negative integer set R={R i } and a random positive integer σ>max 1≤i≤T {R i }-R 1 as a key, and set the first ciphertext partition [L 1 , U 1 ], satisfying L[1]=σ and U[1]=L[1]+R 1 ; then iteratively delineating the boundary value of the subsequent ciphertext partition, satisfying L[t]=max 1 ≤ i <t {U[i]+U[ti]} and U[t]=L[t]+R t , the lower boundary set of the ciphertext partition of the output ciphertext domain L={L i }={L[ i]} and the upper boundary set U={U i }={U[i]}; where 1≤i≤T, 2≤t≤T, T is the maximum value of the plaintext domain; R i is the set R The i-th element, L i and L[i] are the lower boundary of the i-th ciphertext partition, U i and U[i] are the upper boundaries of the i-th ciphertext partition;

b) The encryption sub-algorithm is: input boundary set L, U and positive integer plaintext m, output ciphertext c=E(m); in encrypted plaintext value m, using a one-to-many mapping structure, according to the boundary set L and U of the ciphertext partition, the encryption function E(·) will be from the set {L[m], L[m]+1,...,U[m ]} selects and returns a random integer as the ciphertext value c;

c) The decryption sub-algorithm is: input boundary set L, U and ciphertext c, and output the ciphertext partition number of the ciphertext c as a plaintext value by means of a binary search algorithm.
The method according to claim 3, wherein the boundary generation sub-algorithm of the simplified OPEA algorithm is such that the keys R i (1 ≤ i ≤ T) are arranged in non-descending order to obtain a linear boundary function thereof.
U[t]=L[t]+R t , 1≤t≤T; then the boundary set L and U are generated using the linear boundary function.
The method according to claim 3, wherein the boundary generation sub-algorithm of the extended OPEA algorithm is: setting a key σ>3·max 1≤i≤T {R i }, and setting a first ciphertext partition U' [1]=σ and L'[1]=U'[1]-R 1 , the subsequent ciphertext partition should satisfy the upper boundary U'[t]=min 1≤i<t {L'[i]+L' [ti]} and the lower boundary L'[t]=U'[t]-R t (R t <U'[t]-U'[t-1]); its linear boundary function is
L'[t]=U'[t]-R t ,1≤t≤T; obtain the extended upper boundary set U'={U'[i]} and the extended lower boundary set L'={L'[i] }, where 2≤t≤T,1≤i≤T; the encryption function of the OPEA extension algorithm is E':X→Y', which satisfies the conditions a) and b); X and Y' are the plaintext of the OPEA extension algorithm, respectively And ciphertext space;

a)
b∈X, a<b if and only if E'(a)<E'(b);

b)
b, c∈X, if a+b>c, then E'(a)+E'(b)>E'(c).
The method according to claim 2 or 3, wherein the encryption rule of the sequence-preserving encryption algorithm OPEA is: direct encryption of integer data; when encrypting floating-point data, the ciphertext space is divided according to the minimum precision of the plaintext space. The character data is encrypted, filled, and spliced in turn by the smallest matching unit.
The method according to claim 1, wherein in the step (3), the method for converting the plaintext SQL statement of the data operation request into the ciphertext SQL statement set is:

(31) designing and implementing an SQL-Translator interpreter for the outsourced ciphertext database;

(32) Using the SQL-Translator interpreter to convert the database entities in the plaintext SQL statement, the conversion rules are:

a) The data table name and column name are anonymized using an anti-collision hash function or a symmetric encryption technique;

b) the constant value is converted using the OPEA algorithm;

(33) dividing the converted SQL statement into a minimum conversion unit, which includes a complete predicate expression or a clause structure;

(34) sequentially invoking the corresponding sub-conversion module for the minimum conversion unit, and splicing and arranging the converted ciphertext SQL clause to obtain a complete ciphertext SQL statement; wherein the sub-conversion module i corresponds to the minimum conversion unit i The conversion rule of the minimum conversion unit i is provided and converted into a corresponding ciphertext SQL clause.
The method of claim 1 wherein the database owner sets a user-defined function at the cloud service provider; then in step (4) the cloud service provider invokes the user-defined function in the outsourcing The ciphertext SQL statement is executed on the ciphertext database; wherein the user-defined function includes:

a) Equivalent comparison function EqualityCom(x, E(value 1 ), E(value 2 )), the decision step is: if the distance between E(value 1 ) and E(value 2 ) is not greater than x, the plaintext values are equal , the function outputs 0; otherwise, if E(value 1 ) is greater than E(value 2 ), then value 1 is greater than value 2 and the function outputs 1; if E(value 1 ) is less than E(value 2 ), then value 2 is greater than value 1 The function outputs -1; E(·) is the encryption function of the OPEA algorithm, and E(value 1 ) represents the ciphertext value after the plaintext value 1 is encrypted by the OPEA algorithm; wherein the database owner operates according to the key and the data The plaintext to be compared in the request (value 1 and value 2 ) is selected as a random integer x∈[max{R 1 , R 2 }, σ+R 1 ) and sent to the cloud service provider, R 1 =U[value 1 ] -L[value 1 ] and R 2 =U[value 2 ]-L[value 2 ] are the ciphertext partition lengths corresponding to value 1 and value 2 , respectively;

b) String splitting function Split(str, delimiter), str is the ciphertext search pattern string of the LIKE operator to be split, delimiter is the delimiter, and the split result is stored and returned by the Table type.
The method of claim 5, wherein the database owner sets a user-defined function at the cloud service provider; and then in step (4) the cloud service provider invokes the user-defined function in the outsourcing The ciphertext SQL statement is executed on the ciphertext database; wherein the user-defined function includes: an aggregate comparison function SumEqualityCom

Is a data item set of a database field, SUM is a summation function, L[value] and U'[value] are respectively a lower boundary of the ciphertext partition corresponding to a constant value value and an extended upper boundary;
And R is arranged in non-descending order, and the determining step is:
Not greater than L[value] and
Not less than U'[value], then
Equal to value, the function outputs 0; otherwise, if
Greater than U'[value], then
Greater than value, the function outputs 1;
Less than L[value], value is greater than
The function outputs -1; E(·) is the encryption function of the OPEA algorithm, and E'(·) is the encryption function of the OPEA extension algorithm.
The method of claim 9, wherein the security and value calculation protocol is deployed between the database owner and the cloud service provider according to the function SumEqualityCom, the protocol step is: the cloud service provider calculates SUM (E(Att) )) and SUM(E'(Att)), and send the ciphertext accumulation result to the database owner; the database owner decrypts d=Dec(SUM(E(Att)), L) and d'=Dec(SUM) (E'(Att)), U'); if d and d' are equal at this time, the database owner knows SUM(Att)=d and E(SUM(Att))=SUM(E(Att)) Return to the cloud service provider, the agreement ends; otherwise, the database owner looks for SumEqualityCom(SUM(E(Att)), SUM(E'(Att)), L[d+i], U'[d+i] ) = 0 holds the integer i (0 ≤ i ≤ d'-d), so that SUM(Att)=d+i is obtained, and E(d+i) is returned to the cloud service provider, the agreement End; where Att is a collection of data items for a database field.