Abstract
To solve the problem of library privacy protection, given the defects in the existing service model algorithms based on standard RFID currentlysuch as low time efficiency and poor reversibility, a library privacy protection algorithm based on the RFID technology service model is proposed in this paper. The original library privacy protection database is converted into the form of RFID technology service model for transaction reduction according to the library privacy of each transaction vector after the conversion. The matrix column vector is used to conduct the “AND” operation and calculate the support degree of candidate set and obtain the frequent itemset. The experimental results show that compared with the original algorithm, the algorithm proposed in this paper can improve the time efficiency while ensuring the false positive rate, with excellent reversibility, security andstronger practicability.
Introduction
Privacy protection refers to the extraction or “privacy protection” of the knowledge from massive data, to ensure considerable benefits for enterprises and organizations in their business decision making and potential law analysis, etc. However, the privacy protection itself is a task that requires many soft and hardware resources and professional privacy protection personnel (Li et al., 2012; Yuan & Ebrahimi, 2018). For enterprises or organizations with relatively scarce resources, it is an excellent choice to analyze the privacy protection work and commission it to professional privacy protection institutions (Listed, 2013; Zhang et al., 2016). However, for the analysis work, it is inevitable to involve privacy and information security issues. How to ensure no leakage of user privacy will be an essential topic for privacy protection analysis (Papetti et al., 2012; Digiampaolo & Martinelli, 2012). Some of the existing privacy protection technologiescurrently, such as the random response technology, multi-party security computing and sensitive knowledge hiding, etc. do not apply to privacy protection analysis (Yin, 2015). The reason is that privacy protection analysisboth the raw data information and privacy protection results shall be protected in privacy protection analysis. There are even higher requirements for the privacy protection technology in privacy protection analysis, which cannot be achieved by general privacy protection technology (Zheng & Li, 2012; Bertoncini et al., 2012; Chen, 2013).
Privacy of user activities refers to the free use of library resources and enjoyment of library services by users according to their needs and interests without being monitored and checked by others. Such privacy is mainly related to the right to use the library, i.e., everyone has the right to use the library, should be inviolable in the library to complete their own specific tasks. This is the focus of library user privacy, and also the focus of this paper. Library user information refers to the user’s personal information and various kinds of data generated by the user in the process of using the library. Specifically, it includes library user registration records, library circulation records, interlibrary loan records, reference and consultation records, computer database search records, network usage records, etc (Rios et al., 2015). With the advent of the information age, the interaction between readers and libraries is widening, deepening and accelerating day by day in breadth, depth and frequency. Specifically, the development of network technology has dramatically strengthened the capacity to collect, retrieve, reorganize and dusseminate information in the information systems, which also makes libraries facing considerable risks while serving the readers conveniently. The management and protection of reader privacyreader privacy and security. The research and solution of this problem are in line with the legislative requirement of respecting and protecting human rights in China and is also an aspect of the process of the rule of law in our country. The management and protection of reader privacy is also a perfect explanation of the concept of “reader first, service first” in Library circles. The effective way to address this problem is to punish the acts in violation of reader privacy and security and to institutionalize the ways of protection (Gong et al., 2014).
In theoretical research, many foreign experts and scholars have proposed their own opinions on the privacy protection of digital libraries. How to Maihtain both Privacy and Authentication of Digital identities, written by Shah Jokh Saeedhia, the first librarian of Brussels University in Belgium, proposed his own opinions on the two major network security issues of user privacy protection and authentication of digital libraries. In this paper, the author lists user privacy protection. The protocol description, implementation process and comparison with other schemes are introduced in detail. Lahet L. Galas, an expert in public libraries and information systems in Pennsylvania, believes that users of digital libraries must clearly understand and understand their privacy protection issues. In her article How Should Privacy Be Protected ih, she lists a series of legal information and believes that libraries should take the initiative to provide such information provided to users. The United States, supported by its advanced technology, was the first country to pay attention to the privacy of Library users. The first law on the protection of Library user privacy was passed in Florida in 1978. Kah fl. Coombs believes that the Library Rights Bill and the American Library Association’s Code of Ethics are the basis for studying the privacy protection of library users and that their privacy should be protected according to law (Ni et al., 2015). Lohh Shule points out the responsibility of University Libraries in protecting the privacy of library users.
There are relatively few theoretical studies on privacy protection of Digital Libraries in China, and the number of digital libraries that focus on privacy issues in practical applications is limited. Although most of the theoretical studies proposed that laws and regulations should be established to regulate the protection of user privacy, currentlythere are no specific legal provisions that protect user privacy in China, and the issue of library user privacy protection is not reflected in the legal regulations. Written documents formally concerning the privacy rights of users served by libraries are the Code of Professional Ethics for Chinese Librarians (Trial Implementation) adopted by the Fourth Council of the Sixth Session of the Chinese Library Society on November 15, 2002. Article 4 of the Code stipulates that “safeguarding the rights and interests of readers and keeping their secrets", which defines thecorresponding concepts and laid a foundation for privacy protection policy (Bonter & Bridge, 2015; Yang et al., 2013). Compared with the in-depth study of privacy protection in foreign libraries, most digital libraries in China do not attach great importance to the privacy protection of users. Only a few digital libraries, such as the National Science Library of the Chinese Academy of Sciences, have special privacy protection declarations, while most libraries choose to keep user privacy protection in the “Reader’s Notes” column, and there is no detailed description of specific protection measures.
With the continuous advancement of regulations and laws, the growing awareness of civil rights and the need for in-depth development of culture, the management and protection of reader privacyreader privacy has become one of the hotspots in the library field. AlsoAlso, the extensive application of network technology in the library makes it easy to collect and consolidate the personal information of readers, making it possible to disclose reader privacyreader privacy data. Hence, in the process of serving readers, libraries must strengthen the management and protection of reader privacy. Based on summarizing and comparing the ways and methods of management and protection of user privacy rights and interests both at home and abroad as well as among various industries, this paper conducts further theoretical and practical systematic research and exploration on the safety management and protection of reader privacy from the perspective of law, humanities and library profession, to provide a way and method for the management and protection of reader privacy (Zhang et al., 2017). Currently, the research on reader privacy management and protection in China, both theoretically and practically, is still in its infancy, and a complete system in theory and a set of feasible operational methods in practice are yet to be formed. Therefore, based on the analysis, summary and summary of the relevant research at home and abroad and in various industries, the author conducts a comprehensive and multi-level theoretical research and practical discussion on the management and protection of library reader privacy security, which is a useful supplement to studying the management and protection of library reader privacy security in China. Full understanding and effective implementation of the safety management and protection of reader privacy is the responsibility and obligation of the library industry in building a harmonious socialist society.It is of great significance in responding to great cultural development, improving the quality of service and meeting the needs of readers.
Structure of privacy protection hierarchical model of personalized information service in digital library
There are three kinds of privacy security problems in personalized information service of the digital library: the security of data collection, data access and network communication. Given these three security problems, this paper proposes a hierarchical model of privacy protection for personalized information service of the digital library by reference to the characteristics of OSI (Open System Interconnection Reference Model) (Qin, 2014). This paper divides the privacy protection of customized information service system of the digital library into three levels and adopts hierarchical architecture to solve the existing privacy and security problems layer by layer, to provide sufficient technical guarantee for the personal privacy information of users. The hierarchical model framework is shown in Fig. 1.

Privacy protection hierarchy model of personalized information service in digital library.
Data Modeling Layer: This layer mainly aims at the privacy leak problem that may be caused by privacy reasoning in collecting and utilizing user information to build user interest model and generate a user description file. Modeling refers to the process of establishing a user interest model by using user information. User interest model contains many data contents such as user’s information preference. The primary function of this layer is to prevent illegal elements from obtaining user privacy information according to user interest model and realize the self-openness of user’s personal information by technical means for master control, better control and management of information to be protected (Zhang, 2017).
Data Access Layer: This layer mainly aims at the problem of privacy leakage caused by data access on the server side. Due to the limitation of network security mechanism, the access permit to users can only be assigned to tuples. As a result, unauthorized access attacks that bypass external security mechanisms are prone to occur. Current research on authorized access mainly focuses on the policy-based access mechanism.
Data Communication Layer: This layer is mainly aimed at the leakage of data containing personal privacy information in the process of network transmission. Traditional network communication security has always been an area of user concern and expert research. In the personalized information service system of the digital library, the privacy leak brought by data communication cannot be ignored. Preventing third party from illegal eavesdropping and intercepting the user’s private information is the main task of this layer in privacy protection (Zhang et al., 2015).
The basic principles of the RFID technology service model are as follows:
In the initial state, the RFID technology service model is a binary array that contains m bits, and the initial value of each bit is set to 0, as shown in Fig. 2 as follows.

Initial state of the RFID technology service model.
To show S ={ X1, X2, ⋯ , X n } set that contains n elements, k address mapping functions that are independent of each other are applied in the RFID technology service model to map each element in the set to the range of {0, 1, ⋯ , m - 1 }.
For any element X, the position h i (X) that the i-th function is mapping to will be set to 1 (1 ≤ i ≤ k). If a position is set to 1 for several times, only the first time takes effect.
Figure 3 shows that, k = 3 at this point. Also, the same position is selected in two mapping functions as follows.

Insertion of the element in the RFID technology service model.
When judging whether a new element Y falls in the set S, the same k-th mapping function is applied on Y. If the position of all h i (Y) is 1 (1 ≤ i ≤ k), Y is an element in the set; otherwise, Y is not an element in the set.
In Fig. 4, Y1 is not an element in the set, Y2 may fall in the set or be a misjudgment as follows.

Elemental misjudgment of the RFID technology service model.
As the RFID technology service model based on standard shared space is applied in the original algorithm, when the professional privacy protector feeds the privacy protection result back to the data holder, the data holder cannot determine from which mapping function the “1” in library privacy is mapped. This has made it difficult to conduct the decryption on the result of the privacy protection. According to the original method, only the method of pre-establishing the corresponding table can be used to conduct the decryption by looking up the table. In this way, it is necessary to save and maintain a huge mapping table, which is very inconvenient. To solve this problem, the PMLP algorithm adopts an RFID technology service model based on the independent mapping space and applies the reversible mapping function.
RFID technology service model of the independent mapping space
The difference from the standard RFID technology service model is that the mapping range of the mapping function is different. To ensure good reversibility, the PMLP algorithm adopts a reversible address mapping function. The RFID technology service model based on the independent mapping divides the vector V of the mapping space into k regions, where the length of the vector V is m, each mapping function is only mapped to one of the areas, and the mapping range of each mapping function is m/k bit in length. Also, they do not overlap each other. Hence, the mapping range is as follows:
Similar to the false positive rate analysis in the standard RFID technology service model, when the mapping of all the elements in the set is completed, the probability of any bit in the RFID technology service model vector V based on the independent mapping space being 0 is as follows:
In a similar way, the false positive rate can be obtained as . The results suggest that the false positive rate of the RFID technology service model based on the independent mapping space is very close to that of the standard RFID technology service model.
The objective of this paper is to propose a method for the privacy protection that can solve the privacy protection of the frequent itemset. Through the frequent itemset, the corresponding RFID technology service model can be easily obtained.
As the process of privacy protection is performed on the protection of the library privacy database transformed according to the RFID technology service model, similar definition is given as follows:
If B (S) ∩ B (T
i
) = B (S), then S ⊆ B (T
i
), where ∩ represents the bitwise “AND”. If S ⊆ T
i
, then S ⊆ B (T
i
). If S ⊄ T
i
, then the probability of S ⊆ B (T
i
) is FPR, and the probability of S ⊄ B (T
i
) is 1–FPR. Bfq (S) ≥ fq (S)
To enhance the privacy protection degree of the RFID technology service model, the RFID technology service model with key is adopted, i.e., the key K is used to extend the mapping function h i . To represent the set S, s ∈ S is inserted into the vector of the RFID technology service model, where s represents an element in the S. The h i (s ∘ K) is used to set the position that is corresponding to the library privacy to “1”, where, 0 represents connection in series. When determining whether an item falls in the set S, it is only necessary to check whether all the bits of h i (x ∘ K) are set to “1”. If the key K is unknown, no one can derive the raw data from the library privacy. For the RFID technology service models mentioned in the following section, unless otherwise specified, they are all RFID technology service models with secret key.
Strategy for the protection of the database privacy
In the original algorithm, thelibrary privacy protection data items shall be scanned repeatedly, and a bitwise “AND” operation on the entire library privacy is required, which consumes a lot of time.
The PMLP algorithm proposed in this paper adopts the method of calculating the support degree count of the itemset by using the characteristics of the vector “AND” operation, and streamlining the database through the comparison of the library privacy. This method does not require scanning the redundant database transactions and only needs to conduct the contraposition “AND” operation on the part of columns in the library privacy, which has dramatically improved the algorithm efficiency.
Similarly, the triplet {n, m, k} is used to represent an RFID technology service model of independent spatial mapping, where, n represents the number of elements in the set I; m represents the length of the service vector V of the RFID technology; k represents the number of the mapping functions. After the Bloom transform, the original protection database D of the library privacy can be denoted by the matrix B (D). The matrix discussed in the following section, unless otherwise specified, is a matrix after the Burundian transformation.
Where, p represents the number of transactions; m represents the length of the library privacy V.
Where, “∩” represents the logical “AND” operator. Therefore, the count for the degree of support for the itemset B (S) is as follows:
Where , i > j, i, j ∈ [1, p]; where p represents the number of the transactions; m represents the length of the library privacy V, then the count for the degree of support for the itemset B (S) can be simplified to the following:
Where w (B (T
i
)) ≥ w (B (S)), i ∈ [1, t], t < p. Hence, the count for the degree of support for the itemset B (S) is as follows:
According to the needs of the protection for the library privacy under the RFID technology service model, the process of the protection for the library privacy based on the integration of the RFID technology service model designed in this paper is shown in Fig. 5 as follows.

Process of the protection for the library privacy.
The overall design steps for the PMLP algorithm are given in the section as follows:
For the data holder: Transform the RFID technology service model based on the independent space mapping for the original library privacy protection database to obtain the matrix B (D), which is submitted to the professional privacy protector. Generate the candidate set in the first round to be determined through the Bloom transformation, which is also submitted to the professional privacy protector.
For the professional privacy protector: Conduct pre-processing on B (D): According to the size of the library privacy, the library privacy of the transactions in the matrix B (D) is rearranged in a descending order to obtain the matrix B
sort
(D). Calculate the library privacy of candidate items in the candidate set and determine the protection data of the library privacy that does not need to be scanned, i.e., the protection data of library privacy where the library privacy is less than the candidate item. Calculate the support degree count of candidate items according to definition 6 and determine whether it is a frequent itemset. Repeat the step (4) and (5) until all the candidate items in the candidate set have been determined, and the frequent itemset of the first round is determined. Send feedback results of the privacy protection in this round to the data holder. Generate the candidate set for the next round.
Data communication layer combines anonymous network, data encryption and decryption technology, and a multi-user collaborative privacy protection method is proposed. Encryption and decryption by multiple users ensure the security of data communication. Multi-user collaboration avoids the direct association between user data and specific users and ensures the security of data transmission by multi-layer asymmetric encryption and decryption. Meanwhile, each encryption, decryption and authentication stage facilitates the user’s engagement, which increases the user’s control over data transmission.To verify the effect of the algorithm, the test experiment is carried out in this paper. The experimental environment is as follows: Processor: Intel Pentium dual-core 1.80 GHz; memory: 3GBDDR2; operating system: Windows XPSP3; development tools: Visual C++6.0.
In the experiment, the simulation data set T10I4D100K generated by the IBM data generator and the real data set BMS-POS from the KDDCup2000 are applied, and the information related to the data sets is shown in Tables 1 and 2 as follows. To facilitate the comparative experiment with the original algorithm, scheme 2 is applied to generate the candidate decision itemset in the next round.
Data set description 1
Data set description 1
Data set description 2
In this group of experiments, the simulation data set T10I4D-100K is adopted in this paper.
(1) The false positive rate of the original algorithm and the PMLP algorithm is compared. Let the minimum support degree threshold be δ = 0.75, n = 10 (average transaction length), the length m of the RFID technology service model fluctuates and changes, and the number k of mapping functions is determined by

Comparison of the misjudgment rate of the simulated data.
(2) The time efficiency of the original algorithm and the PMLP algorithm is compared. It is set that n = 10, m = 320,

Comparison of the time efficiency of the simulated data.
In this group of experiments, the real data set BMS-POS is adopted in this paper.
The false positive rate of the original algorithm and the PMLP algorithm is compared. Let the minimum support degree threshold be δ = 0.75, n = 20 (If the average transaction length is taken, to ensure that the false positive rate is low, m will be very large), the length m of the RFID technology service model fluctuates and changes, and the number k of the mapping functions is determined by

Comparison of the false positive rate of real data.
The time efficiency of the original algorithm and the PMLP algorithm is compared. Let n = 20, m = 640,

Comparison of the time efficiency of real data.
Figures 6 and 8 show that the false positive rate of the PMLP algorithm can be controlled within a relatively low range and gradually decreased as the length of the RFID technology service model increases. Also, its size is almost the same as that of the original algorithm, and the expected effect has been achieved. Figures 7 and 9 show that the PMLP algorithm has significantly less execution time than that of the original algorithm. Also, the higher the minimum degree of support is, the higher the algorithm efficiency execution is. Therefore, the PMLP algorithm not only can guarantee the false positive rate of the algorithm but also can improve the algorithm efficiency significantly. Hence, it is of better practical value. Strengthening the protection of Library user privacy is a powerful guarantee to fully exert the social functions of the library. It is also a place where information is collected. The protection of user privacy is a necessary condition to promoting the dissemination and utilization of knowledge and information. The extensive application of information technology in libraries has dramatically changed the operation mechanism and service mode of libraries. Through the network, libraries can provide more diversified services, and users can enjoy library services more conveniently and comprehensively. However, the development of technology also makes user privacy issues more prominent at the same time. How to fully leverage the powerful functions of the network to improve the quality of library services and eliminate the hidden risks and behaviors of privacy violations due to the application of technology at the same time is a problem that libraries have to face and solve sooner or later. It is an important guarantee to fully exert the social functions of libraries by formulating library user privacy policies and establishing the status and measures for protecting the privacy of library users.
In this paper, the protection of library privacy is studied and an improved PMLP algorithm combining the RFID technology service model is proposed. The proposed algorithm has excellent reversibility and security, which has reduced the data scanning volume for library privacy protectionand accelerated the statistical speed for calculating the support degree of data items, which has improved the algorithm efficiency. Hence, it has better practical value in actual applications. As the study on the privacy protection method proposed in this paper is still in the primary stage, the algorithm itself and its effective implementation in practical applications require further study.
