TWI434197B - Knowledge camouflage method - Google Patents

Description

Knowledge camouflage method

The invention relates to a knowledge camouflage method, in particular to a knowledge camouflage method for achieving anti-exploration by data camouflage technology.

At present, with the increasing precision of information technology and data processing technology, the ability of enterprises to collect, store and apply data has also been significantly improved. Especially in the trend of rapid development and popularization of the Internet, the knowledge economy has flooded the world, and anyone who has knowledge can create new economies. Therefore, in the era of knowledge economy, if enterprises can efficiently convert large amounts of information into valuable knowledge, they can quickly respond to the future and seize the market pulsation, thereby enhancing the competitiveness of enterprises.

However, in the past, it was no longer appropriate to process large amounts of data by manual and personal experience. Instead, knowledge extraction combined with artificial intelligence and automated tools was used. Therefore, the development of data mining technology has also become popular, and it has been widely valued and applied by academic circles and industry.

Data exploration is a professional information technology. The main purpose is to mine information that is useful or interesting to users in a huge amount of information, and to transform the information that is mined into knowledge by the domain know-how of the enterprise. Operate in corporate decision-making and help managers manage business activities. Based on the type of data and the needs of users, data mining techniques can be divided into the following six categories: Clustering, Classification, Regression, Time Series. (Time-series), association rules, sequence discovery, and so on.

These different data exploration techniques each have their own characteristics and exploration and analysis functions. Users can select different analysis tools according to their needs while mining data, and mine different types of knowledge from a large amount of data. For example, classification uses the knowledge of known categories to study their characteristics, and then predicts the category of new data based on these characteristics. The grouping is based on the characteristic attributes of the data itself to find the similarity between the data (Similarity) and to be clustered according to the degree of similarity.

At present, the technology of data exploration is mainly used in industries such as finance, circulation, manufacturing and bioinformatics, and is also very practical for other fields. Data mining can help these users to extract meaningful knowledge data from the data as a basis for decision-making or analysis. In contrast, if these materials with knowledge value are inadvertently lost or outflowed, those who are interested can use the technology of data exploration to dig out the knowledge contained in the data, thereby achieving improper benefits or malicious damage, causing victims to suffer losses or even loss. Competitive ability. Therefore, while using data exploration technology, it is more important to pay attention to the issue of knowledge security arising from data exploration.

According to Oracle Chief Executive Larry Ellison, at the "Oracle OpenWorld" conference held in San Francisco in 2005, "As more companies put business applications on the Internet, they allow employees to be at home." Or affiliates around the world connect these applications The security risks of the database will increase. To reduce security risks, companies should encrypt their databases. At the same time, Larry Ellison suggested that companies prohibit users from performing data backups without encryption, because if there is no encrypted backup, once lost, it is equivalent to losing important information about the enterprise.

In the past, in the research field of data exploration, more emphasis on improving the technology of data exploration, or focusing on technology integration and innovation, etc., how to protect the knowledge contained in the data, not easily explored by data exploration, Received a clear attention. In today's increasingly developed Internet, companies have linked the database to the Internet, which in turn increases the security risks of data. The security considerations for these data are still based on data. Security and system risk management are the main factors, and how to protect the security of knowledge is not considered. That is to say, if someone is authorized to access the data, the person can dig out the knowledge in the data. But in fact, being authorized to access data and being authorized to know the knowledge of the enterprise is completely different levels of power. Therefore, how to protect the knowledge contained in the data while using data exploration is the problem that is currently being solved.

In view of the above-mentioned problems of the prior art, the object of the present invention is to provide a knowledge camouflage method to solve the problem of knowledge security of data exploration.

According to an object of the present invention, a knowledge camouflage method is proposed. First input a raw data, use at least one data camouflage method to add interference data or disturb the original data, and record a modified information as a basis for removing the interference data and restoring the data, the modified original data is a protection data, when the protection When the data reaches a knowledge camouflage target, the protection data is output.

Among them, the original data is a data set with knowledge content.

Among them, the method of using the data camouflage method to add interference data or disturb the original data is to add a plurality of pieces to modify the information, or to disturb the field of the original data or the position order.

Among them, the interference data is based on the original data and is derived by formula derivation, such as offset, random oscillation, etc., used to add the original data to generate protection data, so that the knowledge of the protection data and the knowledge of the original data are obvious. The difference is to camouflage the knowledge of the original material.

Among them, the disturbance of the original data refers to the field or position of rearranging and combining the original data values, thereby generating protection data, so that the knowledge of the protection data is significantly different from the knowledge of the original data, thereby disguising the knowledge of the original data.

Among them, the process of protecting data generation can be completely recorded or deduced by a modified formula to modify the information, so that the protection data can be restored to the original data according to the modified information.

Wherein, when the protection data needs to be restored, the information or the information such as the field or the position of the original data is removed from the protection data by using the modification information to restore the protection data to the original data.

According to the above, according to the knowledge camouflage method of the present invention, it may have a Or multiple advantages:

(1) Effectively disguise the knowledge in the original data without being correctly analyzed by the data exploration technology.

(2) This knowledge camouflage method can flexibly adjust the effect of knowledge camouflage by the user's demand when performing data camouflage.

(3) This knowledge camouflage method is a reverse-recovery knowledge camouflage technology.

Please refer to FIG. 1 , which is a flow chart of the knowledge camouflage method of the present invention. The method comprises the following steps: (S10) inputting a raw material for data camouflage method for knowledge camouflage; (S11) using a data camouflage method to disturb the original data, and recording a modified information when modifying; (S12) original data After the data is disguised, it becomes a protection material; (S13) determining whether the protection data reaches a termination condition, and if so, to step (S14), otherwise to step (S11); and (S14) outputting the protection data.

In the above steps, the original data is a data set with knowledge content. The data camouflage method obtains the modification information which generates the interference data or rearranges the combined original data values in the original data, so that the original data is modified to protect the data according to the modification information, and then the protection data is judged to be If the termination condition is reached; if it is, the protection data will be output; if not, return to the previous step and continue to modify the original data until the termination condition is reached. Furthermore, the manner in which the data camouflage method adds interference data or disturbs the original data includes adding at least one field of modifying the field information, adding a plurality of pieces of the modified information, or modifying the field or position of the original data value.

Moreover, the camouflage data method is based on a formula to derive a modification information for modifying the original data, and according to the modification information, adding the interference data or modifying the field or the value of the original data to generate the protection data. In addition, the process of protecting data generation can be completely recorded or deduced by a modified formula to modify the information, so that the protection data can be restored to the original data according to the modified information. Moreover, the termination condition is that the user expects to reach an expected interference effect threshold or an execution limit threshold to meet the goal of knowledge camouflage.

The knowledge camouflage method proposed by the present invention is exemplified by an algorithm, but in actual implementation, it is not limited to such a method or a specific data exploration method, and the repeated parts are not described again.

Please refer to FIG. 2, which is a flow chart of an embodiment of data scrambling using a clustering algorithm according to the present invention. The method comprises the following steps: (S20) inputting a raw data; (S21) obtaining a group center of gravity value of the original data and a basic information by using a grouping algorithm; (S22) setting a termination condition, when the original data is subjected to a grouping algorithm After the modification is made and the termination condition is met, the step of generating the protection data is terminated; (S23) generating a modification information by using a random number generator and a sub-value; (S24) adding the modification information to the original data in a random manner to form a protection data; (S25) reusing the grouping algorithm to obtain a group center of gravity value of the protection data (S26) judging whether the protection data reaches the termination condition, if the step (S28) is performed, otherwise the step (S27) is performed; (S27) deleting the modification information in the protection data, and restoring the protection data to the original data, and returning to the step (S23) regenerating the modification information; and (S28) generating the modification information based on the basic information of the original data, and modifying the information to edit the original data to generate and output the protection data.

In the above steps, the original data is first defined as D , as shown below: D = { d _i , i =1, 2, 3... n }

After analysis by the clustering algorithm, the first group center of gravity value is obtained, as shown below: C = { c ₁ , c ₂ ,..., c _k }

And the number of data with each group: cn ₁ , cn ₂ ,..., cn _k . Where k is the number of clusters, and each center of gravity is as follows: c _j = ( c _{j 1} , c _{j 2} ,..., c _jm )

The value of j in the above formula is greater than or equal to 1 and less than or equal to k . Since the clustering algorithm represents all the data points with the center of gravity of the cluster, it can reduce a lot of calculations.

The basic information o _j is the data point of the jth group closest to the position of its center of gravity c _j , which is the following formula: Let o _j = d _i , then d _i must satisfy .

The basic information o _{j is} used as the basic material for generating the j-th group modification data, and the j- group offset e _j is calculated by the offset ratio r and the basic information o _j previously defined by a user, which are listed below. :

Set the calculation number t =1 and the group center of gravity threshold T , and set the increment multiple b _j of the number of modification information to be equal to t . If the j group has reached the modification target of the termination condition, b _j no longer changes, and each group is calculated. The amount of information required to be modified is as follows:

The above formula is also the information of each group modification to perform the tth time.

In an embodiment of the invention, the modification information is represented by Y , and the modification information Y is added to the original data D in a random manner to generate the protection data D 'and D '=( D Y ). Then, using the grouping algorithm, the second group center of gravity value of D ' is obtained as C '={ c ' ₁ , c ' ₂ , c ' ₃ ,..., c ' _k }. Determine whether the termination condition is reached, ie, whether c ' _j is very close ( c _j + c _j × r ), such as ∥ c ' _j -( c _j + c _j × r )∥ T and [1, k ]. If the termination condition has not been reached, the modification information Y generated this time is discarded, and the modification number t is set to t +1 and the modification information Y is regenerated; if the termination condition is reached, the protection data D ' is output.

In the data camouflage design of the clustering algorithm, a user-defined offset ratio r is included , and r ≠0. The main purpose is to calculate the position of the gravity point offset of each group. As the target of the disguise knowledge of the grouping algorithm, the value of r can be adjusted by the user according to the demand.

The data camouflage design of the clustering algorithm allows the user to flexibly adjust the offset ratio r , and according to the adjustment offset ratio r and the basic information o _j to generate the modification information Y which is similar to the original data D. Then, by evaluating the effect, it is checked whether the position of the newly generated cluster center of gravity point reaches the offset target. If the offset target is not reached, the position of the center of gravity of each group is biased to the offset target set by the user by adding the method of modifying the information Y , so as to complete the data camouflage purpose of the grouping algorithm and ensure the group in the original data D. The focus of the center of gravity, that is, its group knowledge, will not be extracted by the correct analysis due to the outflow of data.

Please refer to FIG. 3, which is a flow chart of an embodiment of the present invention for performing data de-protection using a clustering algorithm. The method comprises the following steps: (S30) generating the modification information by using the original random number generator and the original seed value Seed ; (S31) deleting the modification information in the protection data; and (S32) deleting the modification information according to the modification information. The protected data is restored to the original data.

In the above steps, using the seed value Seed , the protection data D ', the offset ratio r , and setting the random number generator, the modification information Y of the protection data D ' is generated, and then the restored original data D is output.

Please refer to Figure 4, which is a table diagram of an embodiment of the original data. In the figure, it is the Iris Plants Database group, the center of gravity value and the number of clustered data. In this example, the grouping parameter k is 3, and the Iris grouping center point is C = { c ₁ , c ₂ , c ₃ }, first set this data as the original data D , the number of data n = 150, and analyze the clustering algorithm to know the correct center of gravity C. Next, the parameter setting is made, the offset ratio r is set to 0.1, and the total number of calculations is 100 times, and the seed of the random number generator is set by the user.

Please refer to FIG. 5, which is a table diagram of an embodiment of the knowledge camouflage effect. In the figure, the group center of gravity C' after the protection data D' is grouped and modified by the grouping algorithm. A total of 115 revisions were added, which is the goal of reaching the offset center of gravity. The first modified group center of gravity c ₁ ' is compared with the group center of gravity c ₁ before the modification. The number of grouped data of the original c ₁ is 50 pieces of data. After the data is disguised, the protection data is added to 40 pieces. Modify the information so that the value of each field of c ₁ ' is more than 0.1 times higher than the original cluster center point c ₁ , as shown in the following calculation: (5.71-5.00) / 5.00 = 0.142 > 0.1

(3.84-3.41)/3.41=0.126>0.1

(1.74-1.46)/1.46=0.191>0.1

(0.31-0.24)/0.24=0.291>0.1

The same is true for the other c ₂ ' , c ₃ ' two-group center of gravity points, thus confirming that the present invention achieves the objective of the estimated offset ratio r = 0.1. By setting the offset ratio r , the user can set the result of the cluster center of gravity after the knowledge is disguised. This can effectively meet the actual needs of the user, and can also be convenient for the user to customize the reasonable offset direction to achieve the purpose of protecting the camouflage and misleading the illegal user.

Please refer to FIG. 6 , which is a modified information example table diagram of an embodiment of the knowledge camouflage effect. In the figure, in terms of the similarity between the modified information and the original data, since it is based on the data of the original data which is closest to the center of gravity of the group, plus the random number in the specified range, it can be similar to the original data. It is difficult to distinguish the modification information. Next, in the original data and the added modification information, three data are listed in the figure to observe the difference between the two, and observe the data of each group. The values of the same fields are quite close to each other. The modification information cannot be easily filtered by the filter.

The above is intended to be illustrative only and not limiting. Any equivalent modifications or alterations to the spirit and scope of the invention are intended to be included in the scope of the appended claims.

S10~S14‧‧‧Steps

S20~S28‧‧‧Steps

S30~S32‧‧‧Steps

1 is a flow chart of the knowledge camouflage method of the present invention; FIG. 2 is a flowchart of a method for data camouflage using a grouping algorithm according to the present invention; and FIG. 3 is a data release method using the grouping algorithm of the present invention. a flowchart of the method of disguising; FIG. 4 is a table diagram of an embodiment of the original data set; and FIG. 5 is a table diagram of an embodiment of the knowledge camouflage effect; Fig. 6 is a table diagram showing an example of interference data of an embodiment of the knowledge camouflage effect.

S10~S14‧‧‧Steps

Claims (9)

A knowledge camouflage method comprises the steps of: inputting a raw material; using a data camouflage method to add interference data or disturbing the original data, and recording a modified information as a basis for restoring the data; the data after the data is disguised is a protection The data; and when the protected material needs to be restored, the modified information of the record can be used to restore the protected data to the original data. The method of disguising the knowledge as described in claim 1, wherein the original data is a data set having knowledge content. The method for disguising the knowledge as described in claim 1 , wherein the method of using the data masquerading method to add interference data or disturb the original data comprises adding at least one modification field data, adding a plurality of interference data, or modifying the original The field or position order of the data values. For example, the knowledge camouflage method described in claim 3, wherein the interference data is based on the original data, and the formula is used to generate interference data that is similar to the original data, thereby adding the original data to generate the protection data. The method of disguising the knowledge as described in claim 3, wherein the method of disturbing the original data is to create the protection data by modifying a field or a position of the value of the original data. For example, the knowledge camouflage method described in items 4 and 5 of the patent application scope, The process of generating the protection data may be completely recorded or deduced by a modification formula as the modification information, so that the protection data can be restored to the original data according to the modification information. The method of disguising the knowledge as described in claim 4, wherein the protection data refers to a data set containing the original data and the interference data; wherein the original data has not been modified or increased. The method of disguising the knowledge as described in claim 5, wherein the protected data refers to a field or position in which the value of the original data is rearranged or combined; wherein the value of the original material is not modified. The method for disguising the knowledge as described in claim 1, wherein when the protection data is restored, the modification information is used to remove the interference data from the protection data or restore the modification information to a column of the value of the original data. Bit or position.