KR20230102488A - Fingerprinting apparatus and method for storing and sharing data in the cloud - Google Patents

Abstract

A fingerprinting device and method for storing and sharing data in a cloud are disclosed. The fingerprinting method for storing and sharing data in the cloud involves partitioning a relational data set stored in a relational database into a preset number to create a data partition, and selecting subset data from the created data partition. step, generating fingerprint bits for the selected subset data, performing fingerprint embedding using the generated fingerprint bits, and performing fingerprint decoding on a relational data set stored in a relational database determined to be attacked. includes

Description

Fingerprinting apparatus and method for storing and sharing data in the cloud

The present invention relates to a fingerprinting apparatus and method for storing and sharing data in a cloud.

Along with the growth of digital data in information systems, the challenges of preventing leakage of key knowledge information, protection of ownership, and protection of security and personal information are emerging as important issues. use has become a reality. Digital fingerprinting technology is used as an effective solution to protect the copyright and personal information of digital assets from malicious traitor identification and illegal distribution through the cloud, and is also used for the purpose of tracking users and unauthorized distribution of multimedia contents distributed through the cloud. Used a lot. Accordingly, digital fingerprinting technology is recognized as an essential technology for embedding security data in the cloud or protecting data storage such as a database.

Fingerprints identify and verify information embedded in numeric attributes in relational databases. In addition, fingerprints carry a malicious traitor through the matching of the decoding string and the embedding string. In general, a bit string embedded as a fingerprint is a meaningful sequence such as an image or an audio signal. A meaningless sequence is a random bit or form computed from client information at runtime using some private properties, such as a secret key and a cryptographic hash function. It has robustness against human attacks. The main purpose of malicious attacks is to destroy, make undetectable, or untraceable relational databases. When fingerprints are inserted into digital content, some noise and data distortion may occur in the original content.

Republic of Korea Patent Registration No. 10-1496854 (2015.02.23)

The present invention manages digital privacy through numerical attribute protection in a cloud environment, through which secure data is embedded in the cloud, copyright and personal information of digital assets are protected from illegal distribution, and malicious traitors are identified in the cloud. An object of the present invention is to provide a fingerprinting device and method for storing and sharing data in a cloud that protects a relational database in the cloud.

According to one aspect of the present invention, a fingerprinting method for storing and sharing data in a cloud performed by a fingerprinting device is disclosed.

A fingerprinting method for storing and sharing data in the cloud according to an embodiment of the present invention includes the steps of partitioning a relational data set stored in a relational database into a preset number to create a data partition, the generation Selecting subset data from the generated data partition, generating fingerprint bits for the selected subset data, performing fingerprint embedding using the generated fingerprint bits, and storing the relational data stored in the relational database determined to be attacked. and performing fingerprint decoding on the data set.

)을 산출하고, 상기 산출된 선택 임계값을 이용하여 상기 서브셋 데이터를 선택한다.In the step of selecting the subset data, the selection threshold of the data partition (using the following equation)

) is calculated, and the subset data is selected using the calculated selection threshold.

는 표준편차이고,

는 표준 편차의 백분율 값이고,

는 평균이고, c는 미리 설정된 신뢰 계수이다.here,

is the standard deviation,

is the percentage value of the standard deviation,

is an average, and c is a preset reliability coefficient.

, 날짜-시간 스탬프 협정 세계시(UTC: Universal Time Coordinated)

, 로그인 아이디

및 구매자 아이디

를 이용하여 상기 지문 비트를 생성한다.The step of generating the fingerprint bit includes an IP address and a MAC address.

, date-time stamp Universal Time Coordinated (UTC)

, login ID

and buyer ID

The fingerprint bit is generated using

In the generating of the fingerprint bit, a hexadecimal string is generated using a Message Digest 5 (MD5) hash function, and the fingerprint bit is generated by converting the generated hexadecimal string into a binary bit string.

In the step of performing the fingerprint embedding, the fingerprint embedding is performed using an encoding rule of the following equation.

Here, f _i denotes the same fingerprint coefficient as the original coefficient u _i , v _i denotes the fingerprint length, and fb _i denotes the fingerprint signature part.

The performing of the fingerprint decoding may include generating the data partition, selecting the subset data, generating the fingerprint bit, and embedding the fingerprint for the relational database determined to have been attacked. This is done by applying the steps to

According to another aspect of the present invention, a fingerprinting device for storing and sharing data in the cloud is disclosed.

A fingerprinting apparatus for storing and sharing data in a cloud according to an embodiment of the present invention includes a memory for storing a command and a processor for executing the command, wherein the command sets a relational data set stored in a relational database in advance. Creating a data partition by partitioning into numbers, selecting subset data from the generated data partition, generating fingerprint bits for the selected subset data, A fingerprinting method is performed, which includes performing fingerprint embedding and performing fingerprint decoding on a relational data set stored in a relational database determined to have been attacked using the fingerprint detection method.

A fingerprinting apparatus and method for storing and sharing data in the cloud according to an embodiment of the present invention minimizes distortion caused by embedding numeric attributes of fingerprint bits into numeric attributes of a relational database, traitor attack and fingerprint decryption of a designated database can be used to identify traitors, generate signatures using security and private parameters known only to the data owner to specify user authentication in a cloud environment, collusion attacks, false claims of ownership, injection attacks, It can protect databases in the cloud from attacks in various scenarios such as deletion attacks, modification attacks, and brute force attacks.

1 is an overview of a security model according to an embodiment of the present invention;
2 is a flowchart illustrating a sequential workflow for a fingerprinting method for storing and sharing data in a cloud according to an embodiment of the present invention.
3 is a diagram illustrating a security model according to an embodiment of the present invention;
4 illustrates a framework for fingerprint bit generation;
5 is a diagram schematically illustrating a system structure for performing a fingerprinting method for storing and sharing data in a cloud according to an embodiment of the present invention.
6 is a flowchart schematically illustrating a fingerprinting method for storing and sharing data in a cloud performed by a fingerprinting device according to an embodiment of the present invention;
7 is a diagram schematically illustrating the configuration of a fingerprinting device for storing and sharing data in a cloud according to an embodiment of the present invention.

Singular expressions used herein include plural expressions unless the context clearly dictates otherwise. In this specification, terms such as "consisting of" or "comprising" should not be construed as necessarily including all of the various components or steps described in the specification, and some of the components or some of the steps It should be construed that it may not be included, or may further include additional components or steps. In addition, terms such as "...unit" and "module" described in the specification mean a unit that processes at least one function or operation, which may be implemented as hardware or software or a combination of hardware and software. .

Hereinafter, various embodiments of the present invention will be described in detail with reference to the accompanying drawings.

1 is a diagram showing the outline of a security model according to an embodiment of the present invention, and FIG. 2 is a flowchart showing a sequential workflow for a fingerprinting method for storing and sharing data in the cloud according to an embodiment of the present invention, 3 is a diagram showing a security model according to an embodiment of the present invention, FIG. 4 is a diagram illustrating a framework for generating a fingerprint bit, and FIG. 5 is a diagram showing data storage and sharing in the cloud according to an embodiment of the present invention. It is a diagram schematically illustrating a system structure for performing a fingerprinting method for. Hereinafter, the concept of a fingerprinting method for storing and sharing data in a cloud according to an embodiment of the present invention will be generally described with reference to FIGS. 1 to 5 .

로 표현되는 관계형 데이터베이스가 고려된다. 여기서, PK는 기본키(primary key)를 나타내고,

속성(기본키가 없는 경우, 속성이 추가되어 고유 속성 역할을 하고, 새로운 튜플(tuple)의 증가에 따라 자동증가) 및 A_i(i=0, …, n-1)는 핑거프린팅을 위하여 필요한 후보로 지명된 숫자 속성이다.

에 n개의 튜플이 있다고 가정하면, 그 중 1/r은 지문

의 인코딩에 사용된다. 여기서, r은 하기에 주어진 임계값 공식을 통해 계산되는 사용자 정의 매개변수이다. 강력한 워터마킹 방식은 데이터에 작은 왜곡을 불가피하게 도입하므로, 각 속성값은

로 표시되는 LSB(least significant bit)의 수정이 허용될 수 있다고 가정한다. 본 명세서에 개시된 모든 매개변수는 하기 표 1과 같이 나타낼 수 있다.In terms of schema data,

A relational database, represented by , is considered. Here, PK represents a primary key,

Attributes (if there is no primary key, attributes are added to serve as unique attributes, and automatically increase according to the increase of new tuples) and A _i (i = 0, ..., n-1) are required for fingerprinting. It is a nominated numeric attribute.

Assuming there are n tuples in , 1/r of them are fingerprints

used for encoding. where r is a user-defined parameter calculated through the threshold formula given below. A strong watermarking scheme inevitably introduces small distortions into the data, so each attribute value

Assume that the correction of the least significant bit (LSB), denoted by , is acceptable. All parameters disclosed in this specification can be represented as shown in Table 1 below.

는 상수이고, 속성값에 대하여 독립적으로 사용된다. 그러므로, 이는 속성값의 이진 자릿수에 따라 달라진다.Referring to Figure 1, the block diagram shown in Figure 1 summarizes all the key components of the fingerprinting technology.

is a constant and is used independently of attribute values. Therefore, it depends on the number of binary digits in the attribute value.

Referring to FIG. 2, FIG. 2 shows a sequential workflow of a schema in which a fingerprint signature embedded in a relational database is generated. The fingerprinting method according to an embodiment of the present invention is used in a relational database of a stand-alone system as well as a relational database of a cloud. A traitor attack and fingerprint decoding of a given database are used to identify traitors. This block diagram summarizes all the key components of the encoding, decoding and fingerprinting sequence of a traitor attack.

Cloud computing is an automated shared environment and provides a secure entrance to obtain restricted content with high reliability. Personal content includes authentic information that allows access to desired digital information for safe purchase that prevents the distribution of illegal copies of digital information. For example, restricted content such as student transcripts, coupon numbers for lottery tickets, sales profits, income tax information, and patient medical histories are used for users who successfully pass a security and payment barrier called data-as-a-service. . Moreover, data as a service provides on-demand and support to both data users and data providers. Microsoft Azure, Infochimps, Amazon, and Oracle are among the most popular service providers that support customers at low cost through 24/7 availability and sustainability. All service providers have terms and conditions to provide access to users in need of assistance obtaining personal content.

Referring to FIG. 3 , the security model according to an embodiment of the present invention has thread and traitor identification.

Fingerprints and watermarks can be quickly destroyed or removed from the marked database. Some of the common attacks that Mallory can perform on threads in a cloud environment are shown in FIG. 3 . Mallory has multiple attack vectors such as insertion, deletion, modification and collusion attacks. Mallory obtains multiple copies and compares them to false claim attacks and brute-force attacks. When such an attack is applied to a fingerprinted database, the attack turns the fingerprinted database into a vulnerable database. Mallory then tries to claim that this database is his database. After that, this vulnerable database is passed to the attacker channel, the embedded fingerprint is decoded, and a majority vote is applied. Finally, a malicious traitor is identified.

In a collusion attack, Mallory could access multiple copies of the fingerprinted relational database, compare the differences between the copies, and remove all signature values. The attacker applies a mix-and-match or majority-vote strategy to explode the fingerprint signal. This attack only applies to fingerprints. All other attacks involve watermarks and fingerprints. Since some values do not change after the fingerprint is embedded in the selected row, the fingerprinting scheme according to the embodiment of the present invention relies on mixing and matching. Mallory skips this row and cannot perform collusion attacks. The fingerprinting method according to an embodiment of the present invention overcomes a collusion attack without losing data quality.

In a false claim attack, the traitor tries to provide fake evidence to claim that the owner pirated the false evidence. Mallory adds a new fingerprint and watermark signature, and attempts to claim false ownership in this kind of attack. A spoofing attack occurs when an attacker randomly changes the value of a fingerprinted bit and discovers that the value is being successfully guessed. The fingerprinting scheme according to an embodiment of the present invention is also resilient to this type of thread because the fingerprint is applied to the selected tuple. The selection of these tuples can be made with a calculated formula. Mallory adds fake fingerprints to the database to claim ownership. However, when passing this database in the thread model and decoding the fingerprint from the vulnerable database, the theft can be identified.

In an injection and append attack, an attacker assumes that the injected signature is compromised, and adds an extra tuple to the data set, resulting in a synchronization error. Furthermore, Mallory considers the degradation of data quality and condition. However, the fingerprint value embedded in the relational database does not change. Fingerprints can be decoded with 100% decoding accuracy, and traitors can be quickly identified.

In a delete attack, Mallory first aims to delete a subset of data in a relational database. It then attempts to remove the fingerprint from the embedded data in order to destroy data quality and condition. When passing a vulnerable database in the threat layer, if Mallory removes 95% of the data subset, the traitor is identified on decoding. However, the fingerprinting method according to the embodiment of the present invention is still strong against a 95% record deletion attack.

In a modification attack, some relational database processing is unintentionally performed on fingerprinted and walkmarked data. So, Mallory adds some marked data alternately with deletion, insertion, which can completely or partially destroy the watermark string. In the fingerprinting scheme according to an embodiment of the present invention, a security key with secret parameters generates a combination of static parameters and dynamic parameters. These parameters are IP/MAC address, Universal Time Coordinated (UTC), Login ID and Buyer ID. Also, these parameters create various scenarios in real life. So, Mallory does not guess the secret key, but finds the exact tuple with the fingerprint embedded. However, the fingerprinting scheme according to an embodiment of the present invention is flexible for all kinds of modifications and updates of up to 95% of fingerprinted data.

In brute-force attacks, Mallory tries all possible combinations to guess the secure signature and think about secret parameters such as the secret key. The attack is prevented by assuming that the size of the secret parameter is sufficiently long. For secret key generation, static and dynamic combinations of parameters are used and generated in real time. Whenever a user logs in to generate a fingerprint, some parameters are changed according to circumstances. Eight real-time situations can occur while a user is fingerprinting. These parameters are the user's IP address, the system's MAC address, Coordinated Universal Time, the Buyer ID generated at the time of purchase request, and a login ID that is unique for every user in the cloud. When a customer creates a dataset access request, user-specific parameters are used to create a unique fingerprint string. A 35-bit unique identification signature modified to a string is obtained for each customer. Signature generation through the fingerprinting method according to an embodiment of the present invention has a function in which an attacker cannot generate and decrypt an existing signature. There is a 35-bit string length with 34,359,738,368 possible combinations. It takes too much time for Mallory to guess the correct bit or execute a brute force attack. Another strong motivation for robustness is the inclusion of the entire signature string in each specified attribute.

비밀번호, IP/MAC 주소

, UTC

및 구매자 아이디

또는 주문 아이디

가 할당되어, 성공적으로 주문된다. 데이터 소유자만 알고 있는 보안 및 개인 매개변수가 사용된다. 16진수 보안키 k_s와 암호화 해시 함수

이다. 도 4를 참조하면, 도 4는 지문 비트 생성을 위한 프레임워크를 나타낸다.Various scenarios are observed for real-time use of a cloud environment for signature generation using multiple parameters specifying user authentication. These parameters include some private and public parameters. Private parameters are only for legitimate users, while public parameters are considered information known by other users, such as email addresses for login accounts in the cloud environment. login account

Password, IP/MAC address

, UTC

and buyer ID

or order id

is allocated and ordered successfully. Security and privacy parameters known only to the data owner are used. Hexadecimal secret key k _s and cryptographic hash function

am. Referring to FIG. 4, FIG. 4 shows a framework for generating fingerprint bits.

Table 2 below lists parameter changes for fingerprint generation.

또는 기능이 다른

이다.As shown in Table 2, various parameters participate in fingerprint generation. Five parameters belong directly to cloud users. These parameters have two required states,

or different function

am.

Scenario 1: If the client uses a unique buyer ID, uses the same IP and MAC address, but logs in at a different time and uses the login ID, it can be formed as follows.

를 사용하여 고유한 클라우드 로그인 계정

를 통해 클라우드 환경에 접속한 다음, 이 클라우드에서 특정 정보(데이터베이스)를 가져오도록 주문할 수 있다. 개별 구매자 아이디

는 데이터베이스를 확보하기 위해 사용하여 생성된 각 요청에 대하여 할당된다. 이 시나리오는 다음과 같이 나타낼 수 있다.Scenario 2: Users have their own MAC and system's static IP address

Use your own cloud login account

You can connect to the cloud environment via , and then order specific information (databases) to be fetched from this cloud. Individual Buyer ID

is allocated for each request created using it to secure the database. This scenario can be represented as follows.

Scenario 3: There are buyer IDs with different IP/MAC addresses, but the same login time and ID. This can be represented by the following scenario.

Scenario 4: Some buyer IDs have different IP/MAC addresses and login IDs, but the duration varies, which can be represented by the following scenarios.

Scenario 5: There are buyer IDs that open with different IP/MAC addresses and login IDs, but log in at the same time, which can be represented by the following scenario.

Scenario 6: There are buyer IDs with different IP/MAC addresses, login IDs, and login times, which can be represented by the following scenario.

Scenario 7: There is a buyer ID with the same IP/MAC address, login ID, and login time, which can be represented by the following scenario.

Scenario 8: Some buyer IDs have the same IP/MAC address and login time, but different login IDs, which can be represented by the following scenario.

Dataset D is divided into m non-overlapping partitions. This data partition does not change the physical location of data. Nonetheless, these data partitions are based on logical grouping. Initially, the total number of subgroups is equal to the value of m, kept secret, and known only to the data owner to make the partitioning approach secure and resilient to attacks. Finally, the subgroups are divided into two main partitioning bins called even and odd partitioning bins. Even the m' partition contains only candidates with a remainder of zero. Finally, the odd split bin contains the nominated candidates belonging to the remainder of split number 1.

A threshold is computed to select a candidate attribute from the selected usability limited to the attribute. A threshold is calculated to select candidate attributes for marking a digital signature. Noise is reduced in the data, and high security is provided to the end user. This makes it more difficult for an attacker to guess the value of the specified property. Moreover, an important advantage of using an appropriate threshold is that desired data distortion is minimized and reliability of the data is maintained. It is considered to have minor functionality in that it allows minor data changes to be made that do not affect the originality of the data. Only tuples with candidate attribute values greater than the threshold are selected to be inserted into the final sub-data set as the original sub-data set. This division is used logically, not physically. If the candidate attribute values satisfy the threshold, all tuples are inserted into this database.

The fingerprinting scheme according to an embodiment of the present invention embeds a strong multi-bit fingerprint string identifier for each client in a numeric attribute of a relational database-specific tuple. Robust fingerprinting schemes aim to protect databases from illicit distribution with 100% decoding accuracy, proof-of-ownership and resilience to a wide range of attacks. At this time, after the attack, only one fingerprint tuple remains. Fingerprinting is used to insert a unique bit code for each buyer, which creates the least distortion or noise in the relational data. There is an assumption that database clients can tolerate some noise. The fingerprinting database provides multiple copies to multiple buyers.

문자열을 디코딩하는데 사용된다. 데이터 소유자인 앨리스(Alice)가 일부 데이터 세트가 자신의 관계 데이터 세트

에서 불법적으로 복사, 재배포 또는 변조되었다고 비난할 때, 가능한 시나리오에서 지문 감지가 발생한다. 앨리스 또는 제3자는 지문 감지 알고리즘을 통해 의심스러운 데이터베이스의 소유권을 확인한다. 지문 감지 알고리즘의 지문 디코딩 효율성은 일반적으로, 작은 대역폭에 따라 다르다. 대역폭이 클수록 디코딩 정확도가 높아진다. 즉, 다중 공격에 대하여 더 많은 지문 기술이 필요하며, 그 반대의 경우도 마찬가지이다.Fingerprint decoding is the fingerprint bits embedded in the suspect data.

Used to decode a string. Alice, the data owner, wants some data sets to be her relational data sets.

Fingerprint detection occurs in a possible scenario when it is accused of being illegally copied, redistributed or tampered with. Alice or a third party verifies ownership of the suspect database through a fingerprint detection algorithm. The fingerprint decoding efficiency of a fingerprint detection algorithm is generally dependent on a small bandwidth. The larger the bandwidth, the higher the decoding accuracy. That is, more fingerprinting techniques are needed for multiple attacks and vice versa.

A majority voting procedure is performed, which aims to improve decoding accuracy and security against a wide range of malicious attacks.

Strong fingerprinting methods aim to protect digital content through piracy detection, proof of ownership, or copyright. The structure of a system in which a fingerprinting method according to an embodiment of the present invention is performed may be shown as shown in FIG. 5 .

Referring to Figure 5, a strong fingerprint identifies and verifies buyer information embedded in numeric attributes in a relational database. Convey the malicious traitor by matching the decoding string to the array of embedding strings. A bit string embedded as a fingerprint is a meaningful sequence generated through an appropriate sequence. A meaningful pattern is an image, owner information or audio signal. Meaningless sequences can be random bits or computed from client information at runtime using some private properties, such as the secret key k _s and cryptographic hash function MD5. These nonsensical string sequences are powerful and reinforce the properties of the desired fingerprint. Meaningless bit strings are unpredictable and robust against a wide range of malicious attacks, such as insert, delete, and update attacks. The main purpose of a malicious attack is to destroy, make undetectable, untraceable, or alter relational databases. When a fingerprint is embedded in digital content, some noise and data distortion may occur in the original content. Before embedding the fingerprint bits and after completing the signature embedding procedure in the database, the mean and standard deviation are used to calculate the overall change in relational database calculations.

The fingerprinting method according to an embodiment of the present invention focuses on minimizing noise generated by embedding fingerprinting and overall data distortion. Relational databases are assumed to be able to tolerate small amounts of noise that cause data distortion.

Meanwhile, the fingerprinting method according to an embodiment of the present invention can minimize distortion by embedding a fingerprint bit into a numeric attribute of a relational database. Another drawback of existing techniques is their reliance on the original content values of selected attributes of the original data. In the fingerprinting method according to an embodiment of the present invention, an array including some values that are independent of original data and known only to the data owner or data protection authority is introduced. This file observed size is very small compared to the original database. Another aspect related to the robustness of fingerprinting technology is the amount of total bandwidth available for marking fingerprints. Greater bandwidth means more possibilities and space to embed the fingerprint string. The smallest bandwidth means you have limited options for fingerprinting, and it's not much stronger against malicious attacks from attackers. A high level of robustness can be achieved by increasing the size of the usability constraints and relational database bandwidth.

6 is a flowchart schematically illustrating a fingerprinting method for storing and sharing data in a cloud performed by a fingerprinting device according to an embodiment of the present invention.

를 입력으로 받아 데이터 세트 D를 지문 데이터 세트 DFP로 변환한다. 워터마킹에 의한 수정(왜곡)은 사용성 제약 행렬

에 의하여 제한된다. 본 발명의 실시예에 따른 클라우드에서 데이터 저장 및 공유를 위한 핑거프린팅 방법은 결국 데이터 세트를 사용할 모든 가능한 타입의 어플리케이션에 대하여 한번만 정의된다. 이하, 도 6을 참조하여, 본 발명의 실시예에 따른 클라우드에서 데이터 저장 및 공유를 위한 핑거프린팅 방법에 대하여 설명하기로 한다.A strong fingerprint algorithm is used to embed the fingerprint bits into the data set. The fingerprint embedding algorithm uses a secret key k _s and fingerprint bits

as input and converts the data set D into the fingerprint data set DFP. Modification (distortion) by watermarking is a usability constraint matrix

limited by The fingerprinting method for storing and sharing data in the cloud according to an embodiment of the present invention is defined only once for all possible types of applications that will eventually use the data set. Hereinafter, referring to FIG. 6 , a fingerprinting method for storing and sharing data in a cloud according to an embodiment of the present invention will be described.

In step S610, the fingerprinting apparatus creates a data partition by partitioning the relational data set stored in the relational database into a preset number.

가 된다. 이 논리적 분할은 집합 이론의 기본 연산(예를 들어, 합집합, 교집합 및 공집합)을 수행한다. 각 파티션에 공통 요소가 없기 때문에, 다른 세트의 교집합이 비어 있는 것으로 관찰된다. 각 튜플은

로 하나의 파티션에만 속한다. 또한, 각 튜플은 각 파티션이 원본 데이터셋으로 소진되는 교집합의 속성이

로 만족된다. 이 논리적 데이터 파티션은 모든 각 하위 그룹이 최소한 하나의 튜플을 포함하는 속성을 가진다.

이므로, 데이터 파티션 그룹은 비어 있을 수 없다. 각 파티션의 여러 튜플이 동일할 필요는 없다. 데이터 파티션은 두 레벨의 파티션으로 완료된다. 데이터는 레벨 1에서 m개의 하위 카테고리로 데이터를 분류하도록 파티션된다. 그런 다음, 최종 파티션이 수행되고, 모든 하위 그룹은 m' = 0 및 m' = 1에 대하여 2개의 주요 카테고리로 분류된다. 관계형 데이터베이스에서, 모든 튜플의 기본키는 비밀키 k_s 및 암호화 해시 함수와 연결된다. 비밀키는 개인 매개변수이며, 데이터 소유자만 알고 있다. 해시 함수 타입에 따라 16진수 문자열이 생성된다. 광범위한 보안을 제공하는 MD5(Message Digest 5) 해시 알고리즘을 이용하여 1a 28비트의 해시값이 생성된다. 16진수 값이 획득된 후, m의 메소드 값을 고려한다. m 값도 비밀로 유지된다. 모든 행을 m개의 하위 그룹으로 분류한다. 첫번째 레벨의 하위 그룹화를 완료한 후, 모든 하위 그룹 요소를 두 레벨의 추가 파티션으로 나눈다. 이러한 데이터 파티션을 수행하는 알고리즘은 하기 표와 같이 나타낼 수 있다.To initiate the embedding process, we first divide the relational data set into several subgroups. If m is a private constraint number known to the data owner, these partitions do not overlap and partitioning is logically performed. The physical data location remains similar and does not affect the shape of the original data set. This data set is divided into m (e.g. m=100) partitions.

becomes This logical division performs the basic operations of set theory (eg, union, intersection, and empty set). Since each partition has no elements in common, the intersection of the other sets is observed to be empty. each tuple is

belongs to only one partition. In addition, each tuple has the property of intersection where each partition is exhausted into the original dataset.

is satisfied with This logical data partition has an attribute where every subgroup contains at least one tuple.

, the data partition group cannot be empty. Multiple tuples in each partition need not be identical. Data partitioning is done with two levels of partitioning. The data is partitioned to classify the data into m subcategories at level 1. Then, a final partition is performed, and all subgroups are classified into two main categories for m' = 0 and m' = 1. In a relational database, the primary key of every tuple is associated with a secret key k _s and a cryptographic hash function. The private key is a private parameter, known only to the owner of the data. A hexadecimal string is generated according to the hash function type. The 1a 28-bit hash value is generated using the MD5 (Message Digest 5) hash algorithm, which provides extensive security. After the hexadecimal value is obtained, consider the method value of m. The value of m is also kept secret. Classify all rows into m subgroups. After completing the first level of subgrouping, divide all subgroup elements into two levels of additional partitions. An algorithm for performing such data partitioning can be represented as shown in the table below.

In step S620, the fingerprinting device selects subset data from the created data partition.

In order to minimize distortion and noise of original data, only selected tuples are designated for fingerprinting. Attribute selection is based on usability constraints defined by data owners. Here, the usability constraint controls the usability of data loss, is defined by the vendor, and is used to identify within the available bandwidth for fingerprint embedding. Here, bandwidth is the maximum capacity of the relational database where the data resides. The fingerprint decoding efficiency of a fingerprinting algorithm generally depends on this bandwidth. The larger the bandwidth, the higher the decoding accuracy. That is, the more fingerprinting techniques, the stronger they are against multiple attacks and vice versa. Data receivers want ideal-free original data with minimized distortion. Vendors define usability constraints to bind changes.

To preserve information, vendors define usability constraints in relational databases. Selected attribute constraints are selected based on the included values. The resulting value performs an increment or decrement operation on limited attribute values. The mean and standard deviation of the final values are not changed. For attribute selection for selecting data to be fingerprinted, a data selection threshold is calculated using the following equation.

은 데이터 선택을 위한 입력으로 주어지고, 데이터베이스

는 결과 출력으로 주어진다. 데이터베이스

는 값이 임계값 제한을 충족하는 튜플만 포함한다. 지문 인코딩 프로세스의 경우, 속성 A_i가 임계값

을 초과하는 튜플만 데이터베이스

에 삽입된다.

에서 각 파티션 튜플에 대하여, 평균(

), 사용자에 의하여 선택된 일부 백분율 값(

), 표준 편차(

) 및 표준 편차의 백분율 값(

)이 있다. 백분율 값은 0에서 100 사이이며, 데이터 소유자에 의하여 정의된다. c는 0에서 1 사이의 값을 갖는 신뢰 계수이다. 표준 편차의 백분율 값

과 c는 비밀이 유지된다. 보안 매커니즘을 강화하면, 공격자는 지문이 삽입된 선택된 튜플을 추측하기 매우 어렵다. 이 값은 데이터 소유자에 의하여 임의로 정의된다. 두 개의 비밀 매개변수를 사용하는 목적은 표준 편차의 백분율 값

과 신뢰 계수 c의 두 매개변수 값을 모두 정확하게 추측할 수 없기 때문이다. 공격자가 임계값을 계산하는 것은 모든 면에서 불가능해 진다.

를 표준 편차의 백분율 값

과 곱하고, 결과 평균(

)을 신뢰 계수 c의 값과 곱한 값으로 나누어 임계값 계산이 수행된다. 하기 수학식을 통해 획득되는 임계값 및 임계값

보다 큰 모든 속성 값은 데이터베이스

에 저장된다.data partition

is given as an input for data selection, and

is given as the resulting output. database

contains only tuples whose values satisfy the threshold constraint. For the fingerprint encoding process, the property A _i is the threshold

Only tuples that exceed the database

is inserted into

For each partition tuple in , the average (

), some percentage value selected by the user (

), Standard Deviation(

) and the percentage value of the standard deviation (

) is there. Percentage values range from 0 to 100 and are defined by the data owner. c is a confidence coefficient with a value between 0 and 1. Percentage value of standard deviation

and c are kept secret. By strengthening the security mechanism, it is very difficult for an attacker to guess the selected tuple that has been fingerprinted. This value is arbitrarily defined by the data owner. The purpose of using the two secret parameters is the percentage value of the standard deviation.

This is because both parameter values of , and the confidence coefficient, c, cannot be accurately guessed. It becomes impossible for an attacker to calculate the threshold in all respects.

the percentage value of the standard deviation

multiplied by , and the resulting average (

) by the value multiplied by the value of the confidence coefficient c, the threshold calculation is performed. Threshold value and threshold value obtained through the following equation

All attribute values greater than

is stored in

에서 선택된 튜플의 지정 속성에 대하여 수행되며, rA_i는 선택된 튜플의 속성이다. 선택한 튜플의 지정된 속성은 A_i, 0에서 n-1의 튜플 및 속성 열의 속성 값으로 표시될 수 있다. 여기서, v는 하기 수학식과 같이, 속성

의 값에 적용된다.Fingerprint embedding is database

It is performed on the specified attribute of the tuple selected from , and rA _i is the attribute of the selected tuple. The specified attribute of the selected tuple can be represented by A _i , a tuple from 0 to n-1, and the attribute value of the attribute column. Here, v is a property as shown in the following equation

is applied to the value of

where R _i is the selected tuple.

An algorithm for performing such data selection can be represented as shown in the table below.

In step S630, the fingerprinting device generates fingerprint bits for the selected subset data.

, 날짜-시간 스탬프 협정 세계시(UTC: Universal Time Coordinated)

, 로그인 아이디

및 구매자 아이디

를 기반으로 고객 표시 정보로부터 생성된다. 그런 다음, 지문 문자열 생성은 하기 수학식의 로잉 함수(lowing function)로 계산된다.In this step, F fingerprint bits are generated for each client when an access request is submitted in the cloud environment. Binary fingerprint string is IP address, MAC address

, date-time stamp Universal Time Coordinated (UTC)

, login ID

and buyer ID

It is generated from customer display information based on. Then, generation of the fingerprint string is calculated with a lowering function of the following equation.

, IP MAC 주소, 날짜-시간 스탬프

, 비밀키 k_s를 가지는 로그인 아이디

및 지문 비트

를 생성하는 MD5 해시 함수 사이에서 하기 수학식과 같은 연결 연산이 수행된다.Buyer ID here

, IP MAC address, date-time stamp

, login ID with secret key k _s

and fingerprint bit

A concatenation operation such as the following equation is performed between the MD5 hash functions that generate .

은 MD5 해시 함수에 의하여 생성되는

값을 가지는 고정길이 128비트이다. These fingerprint bits are given as input to the fingerprint encoding function. fingerprint length

is generated by the MD5 hash function.

It is a fixed-length 128-bit value.

는 16진 문자열을 2진 비트 문자열

로 변환한다.fingerprint generation function

converts a hexadecimal string to a binary bit string

convert to

는 디코딩 지속을 위하여 데이터베이스

에서 모든 구매자 지문 코드워드 문자열

을 유지하는 것을 수반한다. 충돌 공격을 무력화하기 위하여 지문 코드 데이터베이스를 비밀로 유지함으로써, 보안 오버헤드에 이른다. 일부 보안 조치는 데이터베이스 데이터 접근 제어, 비밀번호 및 암호화를 보호하기 위하여 필수적이다. 보안 문제를 극복하기 위하여, 데이터 소유자는 암호화 해시 함수

를 이용하여 공개 및 비공개될 수 있는 판매자의 비밀키 k_s 및 구매자의 일련번호

로부터 고유한 구매자 지문

를 생성한다. 함수

는 임의 사이즈 문자열의 데이터 S를 고정 사이즈 문자열

로 매핑하고, 해시 함수에 의하여 반환된 값을 해시라고 한다. 이러한 지문 생성을 수행하는 알고리즘은 하기 표와 같이 나타낼 수 있다.however,

is the database for decoding persistence

All buyer fingerprint codeword strings from

entails maintaining By keeping the database of fingerprint codes secret to defeat collision attacks, the security overhead comes into play. Some security measures are essential to protect database data access control, passwords and encryption. To overcome security issues, data owners use a cryptographic hash function

Seller's private key k _s and buyer's serial number, which can be made public and private by using

unique buyer fingerprint from

generate function

converts the data S of an arbitrary size string into a fixed size string

, and the value returned by the hash function is called a hash. An algorithm for performing such fingerprint generation can be represented as shown in the table below.

In step S640, the fingerprinting device performs fingerprint embedding using the generated fingerprint bits.

의 지정된 속성의 선택된 튜플에 이 비트 문자열이 임베딩된다는 것을 의미한다. 단순화를 위하여, 데이터베이스

의 파티션 세트 P_i에 신호 속성

이 포함되어 있다고 가정한다. 본 발명의 실시예에 따른 핑거프린팅 방법은 관계형 데이터베이스의 지정된 튜플의 숫자 속성에 각 고객에 대한 강력한 다중 비트 지문 서명 식별자를 임베딩한다. 서명 임베딩의 경우, 지문 서명 문자열과 m 간에 XOR 연산이 수행된다. 지문 서명 문자열은 전술한 알고리즘 3을 통해 생성되며, m은 레벨 2의 논리 파티션 번호이다. 결과 비트가 계산되며, 결과는 부호 비트 중 하나이며, m 비트는 동일하다. 그렇지 않으면, 0이 된다. 이 단계는 지문 서명의 실제 길이인 L-1번 수행된다. 모든 집합 비트는 결합되어, 결과 어레이

에 저장된다. 이 어레이의 길이는 지문 서명 문자열과도 동일하다. 결과 배열은 각 구매자에 대한 비트 문자열을 보존하고, 이 배열은 데이터 소유자 또는 저작권 모니터링 기관에 의하여 유지된다.A fingerprinting method according to an embodiment of the present invention uses a multi-bit algorithm. Multiple bits are for each client, the length of the fingerprint is from 0 to L-1, and the database

Indicates that this bit string is embedded in the selected tuple of the specified attribute of . For simplicity, the database

A signal attribute to the partition set P _i of

Assume that it contains A fingerprinting method according to an embodiment of the present invention embeds a strong multi-bit fingerprint signature identifier for each customer in a numeric attribute of a designated tuple of a relational database. For signature embedding, an XOR operation is performed between the fingerprint signature string and m. The fingerprint signature string is generated through Algorithm 3 described above, and m is a level 2 logical partition number. The result bits are computed, the result is one of the sign bits, m bits are equal. otherwise, it is 0. This step is performed L-1 times, which is the actual length of the fingerprint signature. All set bits are combined, resulting in an array

is stored in The length of this array is also the same as the fingerprint signature string. The resulting array holds a string of bits for each purchaser, and this array is maintained by the data owner or copyright monitoring authority.

A general encoding rule is defined by the following equation.

Here, f _i denotes the same fingerprint coefficient as the original coefficient u _i , v _i denotes the fingerprint length, and fb _i denotes the fingerprint signature part. The total data distortion created during the fingerprinting procedure can be expressed by the following equation.

는 그대로 유지되고, 그렇지 않으면 뒤집는다. 인코딩 후 LSB를 제거하고, 지문 서명 문자열의 나머지 비트를 다시 확인한다. 왼쪽이 0이면 임베딩을 중단한다. 이 단계를 L-1회 반복하고, 속성 값을 업데이트한다. 다중 비트 지문 서명 길이는 0에서 L-1이고 알고리즘 2를 사용하는 실시간 시나리오에서 계산된 공모없는 고정 길이 서명 문자열이다. 비트 m은 선택한 속성의 기본키를 기반으로 지정된 속성에 대하여 생성되므로, 전체 지문 서명에 대하여 동일하게 유지된다. 선택된 튜플의 지정된 모든 속성에 동일한 지문 서명이 적용된다. 모든 데이터 파티션 P_i에 대하여 동일한 적용 절차를 반복한다. 공격자가 핑거프린팅된 속성 값을 성공적으로 삭제하거나 업데이트하고, 핑거프린팅된 속성을 하나만 남길 경우, 보호 수준이 향상된다. 본 발명의 실시예에 따른 핑거프린팅 방법은 100%의 디코딩 정확도를 제공할 수 있다. 때때로, 데이터 통계에 이상이 포함되지 않는 것이 실생활에서 매우 중요하다. 또한, 평균 및 표준 편차 결과와 같은 통계가 지문 서명 임베딩 전후에 동일하거나 대략적으로 같아야 하는 민감한 데이터가 필요할 수 있다.Assume that the resulting array is equal to the least significant bit (LSB) of the attribute specified for multi-bit fingerprint embedding. In this case, LSB

is kept as it is, otherwise reversed. After encoding, the LSB is removed, and the remaining bits of the fingerprint signature string are checked again. If the left side is 0, the embedding is stopped. Repeat this step L-1 times and update the attribute value. The multi-bit fingerprint signature length is 0 to L-1 and is a collusion-free, fixed-length signature string calculated in a real-time scenario using Algorithm 2. Since bit m is generated for the designated attribute based on the primary key of the selected attribute, it remains the same for the entire fingerprint signature. The same fingerprint signature is applied to all specified attributes of the selected tuple. The same application procedure is repeated for all data partitions P _i . The level of protection is improved if an attacker successfully deletes or updates the fingerprinted attribute value, leaving only one fingerprinted attribute. The fingerprinting method according to an embodiment of the present invention can provide 100% decoding accuracy. Sometimes, it is very important in real life that data statistics do not include anomalies. Additionally, sensitive data may require statistics such as average and standard deviation results to be the same or approximately the same before and after fingerprint signature embedding.

An algorithm for performing such fingerprint embedding can be represented as shown in the table below.

In step S650, the fingerprinting device performs fingerprint decoding on the relational data set stored in the relational database determined to have been attacked.

Fingerprint decoding decodes the fingerprint bit string embedded in the illegally copied suspicious data set. The data partition is created using the same steps as performing fingerprint encoding to create the data partition algorithm, as in the fingerprint encoding step.

의 길이의 반복과 동일하다. 지문 비트는 길이 0에서 L-1로 동일하고, 이 단계는 L번 반복된다. 결과 비트와 m비트가 동일하고, 생성된 지문 비트가 속성 값의 LSB와 동일하고 가정한다. 이 경우, 지정된 속성의 가장 중요하지 않은 값은 변화가 없다. 그렇지 않으면, 지정된 속성의 LSB를 뒤집는다. 어레이

에서 지문 비트

를 저장한다. 여기서,

값이 0일 때, 0의 숫자를 1씩 증가시킨다.Fingerprint decoding can be public or private depending on security requirements. The blind decoding algorithm does not require embedded original fingerprint data. Marked attributes and tuples are identified, and fingerprint bits are decoded. The fingerprint decoding step detects the fingerprint string from the suspect database to identify the traitor. For detection, the same procedures and formulas are performed to initiate the decoding process. This procedure consists of an embedding step, logical data partitioning, threshold computation, selection of specified attributes, computation of values in m bidirectional partitions, selection of strings from result array, and result array.

equal to the repetition of the length of Fingerprint bits are equal in length 0 to L-1, and this step is repeated L times. It is assumed that the result bits and m bits are the same, and the generated fingerprint bits are the same as the LSB of the attribute value. In this case, the least significant value of the specified attribute remains unchanged. Otherwise, reverse the LSB of the specified attribute. array

fingerprint bits from

Save the here,

When the value is 0, the number of 0 is incremented by 1.

에서 0의 수가 1보다 크다는 것이 확인된 다음,

에 0과 동일한 값을 설정하고, 그렇지 않으면, 1로 설정한다. 감지된 지문 문자열을 저장된 지문과 일치시키고, 반역자 정보, 유죄 사용자를 반환하고, 데이터베이스의 불법 복사본을 배포한다. 지문 디코딩은 의심스러운 데이터로부터 임베딩된 지문 비트 문자열을 디코딩하는데 사용된다. 데이터 소유자가 자신의 관계 데이터 R에서, 일부 데이터 세트가 불법적으로 복사, 재배포 또는 변조되었다는 혐의를 받았을 때, 가능한 시나리오에 따라 지문 검출이 발생한다. 핑거프린팅 알고리즘의 지문 디코딩 효율은 일반적으로, 대역폭에 따라 달라진다. 대역폭이 클수록, 디코딩 정확도가 높아진다. 즉, 지문 기술이 많을수록, 다중 공격에 대하여 더 강력해진다. 이러한 지문 디코딩을 수행하는 알고리즘은 하기 표와 같이 나타낼 수 있다.To properly and efficiently decode the final embedded fingerprint information, a majority voting mechanism is performed. It aims to create better security and decoding accuracy against a wide range of malicious attacks. Eventually, a majority vote is performed and the array

After confirming that the number of 0 is greater than 1 in ,

set to a value equal to 0, otherwise set to 1. It matches the detected fingerprint string with the stored fingerprint, returns traitor information, guilty users, and distributes illicit copies of the database. Fingerprint decoding is used to decode a fingerprint bit string embedded from questionable data. When a data owner suspects that some data sets have been illegally copied, redistributed or tampered with in his relational data R, fingerprint detection occurs according to a possible scenario. The fingerprint decoding efficiency of a fingerprinting algorithm generally depends on bandwidth. The larger the bandwidth, the higher the decoding accuracy. That is, the more fingerprint techniques, the more powerful it is against multiple attacks. An algorithm for performing such fingerprint decoding can be represented as shown in the table below.

7 is a diagram schematically illustrating the configuration of a fingerprinting device for storing and sharing data in a cloud according to an embodiment of the present invention.

Referring to FIG. 7 , a fingerprinting device for storing and sharing data in a cloud according to an embodiment of the present invention includes a processor 10, a memory 20, a communication unit 30, and an interface unit 40.

The processor 10 may be a CPU or a semiconductor device that executes processing instructions stored in the memory 20 .

Memory 20 may include various types of volatile or non-volatile storage media. For example, memory 20 may include ROM, RAM, and the like.

For example, the memory 20 may store instructions for performing a fingerprinting method for storing and sharing data in a cloud according to an embodiment of the present invention.

The communication unit 30 is a means for transmitting and receiving data with other devices through a communication network.

The interface unit 40 may include a network interface and a user interface for accessing a network.

On the other hand, the components of the above-described embodiment can be easily grasped from a process point of view. That is, each component can be identified as each process. In addition, the process of the above-described embodiment can be easily grasped from the viewpoint of components of the device.

In addition, the technical contents described above may be implemented in the form of program instructions that can be executed through various computer means and recorded in a computer readable medium. The computer readable medium may include program instructions, data files, data structures, etc. alone or in combination. Program commands recorded on the medium may be specially designed and configured for the embodiments or may be known and usable to those skilled in computer software. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks and magnetic tapes, optical media such as CD-ROMs and DVDs, and magnetic media such as floptical disks. - includes hardware devices specially configured to store and execute program instructions, such as magneto-optical media, and ROM, RAM, flash memory, and the like. Examples of program instructions include high-level language codes that can be executed by a computer using an interpreter, as well as machine language codes such as those produced by a compiler. A hardware device may be configured to act as one or more software modules to perform the operations of the embodiments and vice versa.

The embodiments of the present invention described above have been disclosed for illustrative purposes, and those skilled in the art having ordinary knowledge of the present invention will be able to make various modifications, changes, and additions within the spirit and scope of the present invention, and such modifications, changes, and additions will be considered to fall within the scope of the following claims.

Claims (7)

In the fingerprinting method for storing and sharing data in the cloud performed by a fingerprinting device,
Partitioning a relational data set stored in a relational database into a preset number to create a data partition;
selecting subset data from the generated data partition;
generating fingerprint bits for the selected subset data;
performing fingerprint embedding using the generated fingerprint bits; and
A fingerprinting method for storing and sharing data in the cloud, comprising performing fingerprint decoding on a relational data set stored in a relational database determined to be attacked.
According to claim 1,
In the step of selecting the subset data,
The selection threshold of the data partition using the following equation (

), and selecting the subset data using the calculated selection threshold.

here,

is the standard deviation,

is the percentage value of the standard deviation,

is the mean, and c is the preset confidence coefficient
According to claim 1,
Generating the fingerprint bit,
IP address, MAC address

, date-time stamp Universal Time Coordinated (UTC)

, login ID

and buyer ID

Fingerprinting method for storing and sharing data in the cloud, characterized in that for generating the fingerprint bit using.
According to claim 1,
Generating the fingerprint bit,
For data storage and sharing in the cloud, characterized in that a hexadecimal string is generated using a Message Digest 5 (MD5) hash function, and the fingerprint bit is generated by converting the generated hexadecimal string into a binary bit string. Fingerprinting method.
According to claim 1,
The step of performing the fingerprint embedding,
Fingerprinting method for storing and sharing data in the cloud, characterized in that the fingerprint embedding is performed using the encoding rule of the following equation.

Here, f _i denotes the same fingerprint coefficient as the original coefficient u _i , v _i denotes the fingerprint length, and fb _i denotes the fingerprint signature part.
According to claim 1,
The step of performing the fingerprint decoding,
Characterized in that the step of generating the data partition, the step of selecting the subset data, the step of generating the fingerprint bit, and the step of performing the fingerprint embedding are applied to the relational database determined to have been attacked. A fingerprinting method for storing and sharing data in the cloud.
In the fingerprinting device for storing and sharing data in the cloud,
memory for storing instructions; and
Including a processor that executes the instructions,
The command is
Partitioning a relational data set stored in a relational database into a preset number to create a data partition;
selecting subset data from the generated data partition;
generating fingerprint bits for the selected subset data;
performing fingerprint embedding using the generated fingerprint bit; and
A fingerprinting device for storing and sharing data in the cloud, characterized in that performing a fingerprinting method comprising performing fingerprint decoding on a relational data set stored in a relational database determined to be attacked.

Images