Research on upgrading the informationization of college archives management based on information technology

The design quality of the system design is directly related to the goodness of the software. The system needs to follow certain design principles in design: 1)

System design refers to the system is divided into several large modules, and each module is designed independently of each other, there is no logical relationship between each other, but through the database is connected, so as to make the system in the form of a whole to play a role in its conducive to the better realization of the system’s functions. Each module is independent of each other, so it is more secure and reliable, and the structure of the system is more perfect.

When designing the system, in addition to following certain design principles, it is also necessary to take into account the reality of the actual situation, mainly in the following aspects: 1)

Practicality

Practicality refers to the design of the system should be based on the reality of the school’s needs, so as to facilitate the school’s management of teacher and student records, to achieve a high degree of data informatization, and to enhance the practicality of the system.

The digital certificate of the management system can be used to securely authenticate the user’s identity, and its main function is to safeguard the security of the file in the transmission process and the object of transmission. This system can guarantee system security by using digital certificates that uniquely identify the user’s identity.

Authentication is required for various operations of staff within the university in the records management system. So a new CA authentication support is needed. Through the use of middleware technology, ASP, XML, WebService, CORBA, COM components, and so on, so as to realize the underlying security device hardware differences and cryptographic logic for shielding, the user only needs to add its needs in the business of the security function, and then the operation of a simple configuration, you can realize the security applications based on PKI, and, therefore, will be able to greatly make the development cost can be reduced, thus maximizing the efficiency of development [17]. The current extranet PKI security services middleware platform has been used by a large number of universities and colleges unified construction, the project construction of this system can be realized by using the platform that has already been constructed.

In order to ensure database security and information leakage, this system adopts MD5 encryption to encrypt the key information of the system, such as ID card number, home address and other sensitive information. And in order to distinguish between different account roles, the system introduces a privilege control function, which strictly distinguishes the functional privileges of users, and strictly restricts the functional privileges of users with different roles to ensure the framework security and stability of the system.

In this section, we present the architecture of the blockchain-based traceable digital archive sharing scheme. The system framework is shown in Fig. 3, and the architecture consists of a user interface layer, a smart contract layer, and a storage layer.

Figure 3.

System framework

1)

Extraction and protection of archive sharing metadata (1)

Extraction of archive sharing metadata

Archive sharing metadata is a data structure extracted from the archive data sharing process, containing data description information, data operation records and security control information, which is stored in the form of transactions in the distributed ledger of the blockchain.

Keyld is the unique identifier of the archive sharing metadata, which is obtained by calculating the hash value of the privacy preserving metadata topic. The calculation formula is shown in (1): (1)KeyId=HASH(ParentKeyId|SenderId|Re cipientId\Archi veAttribution)

Digitalsignature is used to ensure the integrity and non-repudiation of archive sharing metadata by generating a signature on the archive sharing metadata using the sender’s private key. The calculation formula is shown in (2): (2)Digital signature=SIGN(KeyId, Sender PrivateKey)

(2)

File sharing metadata storage

The storage of archive shared metadata is encrypted and packaged to be stored on the alliance chain through smart contract to ensure its tampering and traceability. Through the hash function of the file sharing metadata, the file sender Id, file receiver Id, file attribute data and other hash operations to get the unique identification keyId of the current data block, and then encrypt the keyId through the file sender’s private key sender Private Key to generate a digital signature, and this keyId will be sent to the file receiver, file The receiver of the archive will use this keyId to query the shared metadata of the archive on the federation chain. Finally, call the uplink method of the federation chain to uplink the generated file sharing metadata.

Figure 4.

File sharing process

3)

Archive leakage tracking

By storing the archive sharing metadata on the blockchain, this ensures transparency and tamper-proof data sharing, and thus this paper solves the problem of traceability of the sharing process in the archive sharing process. The relationship between the archive sharing metadata and the blockchain ledger transaction record structure is shown in Figure 5.

Figure 5.

The relationship between the metadata and the record structure of the account

1)

Practical Byzantine Fault Tolerant Consensus Algorithm

The PBFT consensus algorithm provides a high level of security to the system. The PBFT consensus algorithm proves that the ultimate consistency of the system can be guaranteed when there are f malicious nodes in the system as long as the total number of nodes N in the network satisfies the number N ≥ 3f + 1. To ensure that the nodes of the blockchain network are in agreement, the PBFT consensus algorithm implements a consistency protocol, a view switching protocol and a checkpointing protocol. The consistency protocol ensures that all nodes are able to save correct and consistent messages through a system node peer-to-peer communication and reply confirmation mechanism. Consistency protocol is the key process to realize consensus in PBFT.

Where d_max denotes the maximum delay allowed by the system, communication delay exceeding this value will cause the task node to go down.

Completion index, which refers to the success rate of nodes in completing message consensus during the consensus period. Completion index τ is denoted as: (4)τi=ni−viE∈(0,1]

Where E represents the total number of times the node participates in the consensus, the more times the node participates in the consensus proves that the sub-node is more credible, and the degree of completion will be relatively higher, but the more times it submits incorrect data, the lower the degree of completion.

The node activity index refers to the ratio of nodes’ online participation in the consensus process. The activity index μ is expressed as (5)μi=e−tiwi∈(0,1)

From the index, it can be seen that the longer the node is online the greater the active index, on the contrary the longer the node is offline the lower the active index, and in the selection of the consensus node is more inclined to choose the more active node.

The method of calculating credit value according to the node credit evaluation index is shown in equation (6). (6)Ri=ω1×σi+ω2×τi+ω3×μ1

(2)

Node Partitioning Mechanism

In CP-PBFT improved consensus algorithm, let the total number of nodes be N, the maximum number of Byzantine nodes that the system can withstand is f, and N ≥ 3f + 1 is required for the system to reach consensus correctly. This paper introduces the Sturges empirical formula as a network grouping method. The Sturges empirical formula is widely used in statistical research to control the number of data groupings, ensuring that the number of data groupings should be large enough to reflect the distribution of the data but not too large to make the distribution of the data unclear. Using its grouping by credit value adequately reflects the overall credit value situation. In partitioning, let the number of partitions be k, then k=Floor(1+3.3×log(N)) , pi∈{p1,p2,p3,⋯,pk} denote the node partitions in the blockchain network. By evenly distributing nodes according to their credit values, the credit value status of nodes in each partition group remains consistent, preventing a certain partition from failing to reach consensus due to the presence of too many Byzantine nodes, setting the sequence of nodes’ credit values as ri∈{r1,r2,r3,⋯,rN} , with credit values increasing from r_i to r_N, and ensuring that the credit value status of each partition is similar by means of the S-shaped credit value filling method.

After the network node partition, the average number of nodes owned by each node numpi is rounded to NFloor(1+3.3×log(N)) . In this paper, the partition is carried out in accordance with the credit value, a credit value corresponds to a node, and the credit value belonging to a partition corresponds to the node belonging to a partition.

C4.5 algorithm is the more mature decision tree algorithm and the mainstream decision tree algorithm, which is characterized by fast classification speed and high classification accuracy. C4.5 algorithm is an improved algorithm based on ID3 algorithm, and ID3 algorithm, compared to the increase of continuous attributes, attribute value of the vacant case of processing, in the efficiency of a great improvement in the C4.5 algorithm of the idea is: assuming that the S is a training sample set, the training sample set S to build a decision tree for the training sample set, the largest value of Gain−Ratio(x) is selected to be an attribute for splitting the node, in accordance with this criterion can be S divided into n subsets. If the tuples contained within the ith subset S_i are of the same category, then this node acts as a leaf node of the decision tree and stops splitting. For those S_i that do not satisfy the above conditions, the tree is generated sequentially and recursively using the above method until all the tuples contained in all the subsets belong to the same category [18]. It is based on the following principle:

Define a category information entropy: let the training set S has s samples, the training set is divided into m classes, the number of instances in the ith class is s_i, s_i/s that is the probability p_i, Info(S) is the category information entropy, based on the information entropy calculation of the formula is: (7)Info(S)=−∑i=1mpilog2(pi)

Definition 2 Conditional information entropy: If attribute A is chosen to divide the training set S, and the training sample set S is divided into k subsets {S1,S2,…,SK} , and let attribute A have k different values {a1,a2,…,ak} , then the number of training instances belonging to the ith class in Definition S_j is s_ij, and Info_A(S) is the conditional information entropy of attribute A, and the information entropy of the subset divided into subsets by A is given by Eq. (8): (8)InfoA(S)=∑j=1ks1j+s2j+…+smjs×Info(Sj)

where Info(Sj)=−∑i=1mpijlog2(pij) ; pij=sijsj is the sample probability of class i in S_j.

Definition 3: The information gain for classifying attribute A is calculated as: (9)Gain(A,S)=Info(S)−InfoA(S)

Definition 4 Split information entropy: Let attribute A have k different values, and the sample set S can be divided into k subsets using attribute A. One of them S_j contains some such samples in S: they have value a_j on attribute A. If the samples are partitioned in terms of the value of attribute A, Info(A) is the split information entropy of attribute A, as in equation (10): (10)Info(A)=−∑j=1kpjlog2(pj)

Definition 5: The formula for dividing the information gain rate for attribute A is: (11)Gain−Ratio(A)=Gain(A,S)Info(A)

1)

Theoretical foundation

The Taylor series is a representation of a function in terms of series, that is, a concatenation of infinite terms, which are computed from the derivatives of the function at a given point, and Taylor’s Median Theorem is defined as follows:

If for any function f(x), it contains derivatives up to order (n + 1) in some open interval (a, b) containing x₀, then for any x ∈ (a, b), there are: (12)f(x)=f(x0)+f′(x0)(x−x0)+f−(x0)2!(x−x0)2+…+fn(x0)n!(x−x0)n+f(n+1)(ξ)(n+1)!(x−x0)n+1

The ξ in the above equation refers to a value between x₀ and x.

In Taylor’s formula, let x₀ = 0, then ξ is between 0 and x. Let ξ=θx(0<θ<1) , and so Taylor’s formula becomes the simple form, which is often referred to as McLaughlin’s formula with a Lagrange cosine term: (13)f(x)=f(0)+f′(0)x+f′(0)2!x2+…+fn(0)n!xn+f(n+1)(θx)(n+1)!xn+1(0<θ<1)

Removing the Lagrange residual term gives the approximate formula: (14)f(x)≈f(0)+f′(0)x+f*(0)2!x2+…+fn(0)n!xn

2)

Improvement Ideas

From the basic principle of C4.5 algorithm, it can be known that the selection of attributes when generating the decision tree is based on the principle of information theory, due to the fact that in the process of calculating the formula for the information gain rate involves multiple logarithmic function operations, which requires calling the library function several times during the calculation, which greatly increases the calculation time. To address this problem, an improvement method for the information gain rate formula is proposed, i.e., to simplify the calculation of the information gain rate of the C4.5 algorithm by using the mathematical Taylor’s formula and McLaughlin’s formula, which reduces the complexity of the algorithm’s computation to a large extent, and the improved C4.5 algorithm is named as the TAM-C4.5 algorithm.

Since the derivative of ln(x) at x₀ = 0 is meaningless and the range of values of the probability commonly used under the information gain rate formula is between [0,1] , the McLaughlin’s formula of ln(x + 1) is chosen in this paper to improve the formula of the information gain rate in the traditional C4.5 as in Eq. (15): (15)ln(x+1)≈x−12x2+13x3…+(−1)n−11nxn

So there: (16)ln(x)≈(x−1)−12(x−1)2+13(x−1)3…+(−1)n−11n(x−1)n

When x ∈ (0, 1). (17)ln(x)≈(x−1)−12(x−1)2+13(x−1)3

Through the above approximate simplification, it is possible to realize the conversion of logarithmic operations into non-logarithmic operations, which can eliminate the complex logarithmic operations in the formula of the rate of information gain by using the above transformation characteristics, so as to achieve the purpose of simplifying the calculation formula and improving the efficiency of tree building. And the formula is more accurate than the simplification of the equivalent infinitesimal ln(1 + x) ≈ x.

The transformation formula of category information statement is shown in (18): (18)Info′(S)=−∑i=1mpilog2(pi)=−∑i=1msislog2(sis)=−∑i=1msisln(sis)ln2=−1ln2×s∑i=1msi×ln(sis)=−1ln2×s∑i=1msi×[(sis−1)−12(sis−1)2+13(sis−1)3]=−1ln2×s∑i=1m[si(si−s)(11s2+2si2−7sis)6s2]

Similarly, the transformation formula for conditional information stated and split information entropy is shown in (19): (19)InfoA′(S)=−1ln2×s∑j=1k∑i=1m[sij(sij−sj)(11sj2+2sij2−7sijsj)6sj2] (20)Info(A)=−1ln2×s∑j=1k[sj(sj−s)(11s2+2sj2−7sjs)6s2]

Therefore, the formula for the transformed information gain rate is shown in (21): (21)Gain−Ratio′(A)=Info(S)−InfoA′(S)Info(A)

In this paper, the C4.5 algorithm is used to accurately classify the access customers of college archives according to whether their access frequency is frequent or not, and the assessment of whether the access frequency of college archives is frequent or not should be combined with the specific behaviors and actual situation of each customer, including the specialty, age, etc. of each customer. According to the identity (academic background), gender, age and specialty of each customer, the classification model of C4.5 decision tree is established to evaluate the frequency of customers’ visits to college archives, the essence of which is to use the C4.5 algorithm to mine the data to obtain the classification law, that is, to obtain the relationship between the personal information of the visiting customers and their visit frequency, and to derive the classification rules, i.e., the customers’ evaluation of the frequency of visits to college archives. Intelligent model.

Take the information of 2023 customers visiting college archives in a college as an example to establish a data table, including fields customer number, identity, gender, age, major and access frequency, use data conversion and integration and other related technologies to remove noise and other irrelevant information in the data, and transform data types and values of data sources into a unified standard format. For clients who are both teachers and postgraduate students, the status is defined as teachers. The attributes of the clients’ visit frequency are treated as follows: ① Clients who visit more than 3 times a week are designated as frequent visitors. ② The remaining customers are designated as infrequent visitors. The information table of converted customers is obtained after data preprocessing, and the personal information feature set of customers after data conversion is shown in Table 4.

After pre-processing the data, according to the C4.5 algorithm, the attributes identity, gender, age and specialty are taken as the object attributes of the algorithm, and the attribute “whether to visit frequently” is taken as the target attribute, and the attributes are sorted one by one by applying the meaning of the rate of information gain, and the attribute with a high rate of information gain is selected as the test attribute, and a root node is constructed, and this attribute is used as a marker to construct a branch for each value of the decision tree recursively. Using this attribute as a marker, we build a branch for each attribute value, and recursively build a tree to construct a corresponding decision tree, the specific processing of the algorithm is as follows. The attribute “identity” is chosen as the root node of the decision tree because it has the largest information gain rate value. For each branch, the above steps are repeated to generate the decision tree, and the final decision tree is shown in Figure 6.

Figure 6.

The resulting decision tree

1)

If status = “undergraduate” AND age <20 AND major = “chemistry” THEN visits were more frequent.

From the above rules, it can be seen that the degree of influence of each information of customers on the frequency of visit is “identity”, “age”, “specialty”, “gender”. According to this rule, university archives can classify their clients into frequent and infrequent visitors. In addition, according to the customer’s personal information can be provided timely information content, not only to maintain the access of frequent customers to the college archives, but also to attract infrequent customers to visit the college archives, and thus improve the management of college archives, so as to provide customers with better personalized service.

Taking the number of nodes n as 100, 200, 400, 500, 800, 1000, the number of PBFT consensus algorithm communication is shown in Table 5. Taking the number of groups of PAPBFT consensus algorithm as 5, 10, 15, 20, 25, the number of PAPBFT consensus communication is shown in Table 6. As shown in Table 5 and Table 6, when the number of nodes n is 100, 200, 400, 500, 800, 1000 respectively, the number of PAPBFT consensus algorithm communication is reduced by many orders of magnitude than the traditional PBFT consensus algorithm.

The communication complexity of PBFT is known to be O(n2) and the communication complexity of PAPBFT is calculated to get O(nk*log kn) where n is the total number of nodes in the network and k is the number of groups. The communication complexity comparison is shown in Fig. 7. After the above comparative analysis, when the number of nodes is less, the PAPBFT communication complexity is closer to PBFT, and as the number of nodes increases, the PAPBFT communication complexity is much smaller than the traditional PBFT, and the increase is smaller and the communication is stable. It can be seen that when the number of nodes is large, PAPBFT can significantly reduce the system communication complexity and improve the efficiency of the system. In the university file management framework, a large number of school nodes no longer need to maintain a P2P link state with all other school nodes, and only need to maintain a network connection with the same node from the consensus cluster, which reduces the resource consumption of the school nodes in hardware and network.

Figure 7.

Comparison chart of communication complexity

The experiment uses docker containers as the isolation technology. Due to the lightweight nature of Docker containers, thousands of containers can be run in a single machine environment. Multiple node containers are started based on the customized Docker image using Docker commands. Each container represents a consensus node, which collectively constitutes the consensus domain. The experiment is conducted using the Golang programming language and supports multiple concurrent program design schemes. The Golang programming language is used and simulation tests are conducted using JetBrainsGoLand version 2023.1.2. Golang provides efficient concurrent program design solutions and supports multi-threaded work modes, which facilitates distributed consensus algorithm implementation. Multi-computer network communication is implemented in the same LAN for consensus process simulation. This experiment compares the throughput of traditional PBFT algorithm as well as PAPBFT under different number of groupings based on the number of nodes, the number of nodes are taken as 25, 50, 125 and 200 respectively and the PAPBFT groupings are taken as 5, 10, 15 and 20 to draw the histograms and analyze the data comparatively and the throughput when N and K are taken at different values is shown in Fig. 8. According to the above shown, the throughput of PAPBFT consensus algorithm increases as the number of nodes N increases. From this point of view, the larger the number of consensus nodes, the larger the throughput of the system, which greatly increases the scalability of the university file management framework. As can be seen from the figure, the throughput of the traditional PBFT consensus algorithm gradually decreases with the increase of the number of nodes, when N=25, the PBFT throughput is 1250, when N=200, the PBFT throughput is 215. The improved consensus algorithm effectively alleviates the problem of the obvious decrease in throughput due to the continuous increase of the number of nodes, and is suitable for application in the university archive management framework.

Figure 8.

The throughput of N and K at different values