The reason is indeed a bit exhausting in coding, so this chapter is not finished yet. I truly regret using the leave note at the beginning of the month, so the update will probably be a little late, probably around one o'clock in the morning. By then, just refresh this chapter.

Sometimes I also think, when it's actually exhausting to code like this, should I just take a leave, but really can't bring myself to do it, which leads to the current situation. Every time, I ponder life, why keep writing? Isn't it better to honestly be a Programmer?

Abstract: To reduce the delay in secure data transmission over multi-source heterogeneous networks, a machine learning-based secure data transmission technology for multi-source heterogeneous networks is designed. By selecting the importance of data sources and Attributes definitions, preprocessing multi-source heterogeneous network data, and establishing a multipath parallel transmission architecture, effective bandwidth estimation and parameter filtering processing are adopted using machine learning methods. Finally, bandwidth scheduling and channel security protocol systems are established to complete the machine learning-based secure data transmission for multi-source heterogeneous networks. The experimental results show that the machine learning-based secure data transmission effectively reduced data transmission delay, decreased data transmission interruption and packet loss rate, meeting the design requirements of data transmission technology.

Keywords: Machine Learning; Multi-source heterogeneous networks; Secure data transmission; Network data preprocessing; Parallel transmission architecture

1 Introduction

Currently, communication technology is developing rapidly, with obvious characteristics of various networks. After years of reform and innovation, the transmission rate of wireless access technology is gradually approaching the limit. In this context, to satisfy multiple business demands, multi-network writing is required. However, traditional writing mechanisms, in terms of network transmission resource usage, cannot be used simultaneously and efficiently, cannot effectively ensure high-efficiency transmission business, and will increase energy consumption issues during transmission, leading to interference problems during transmission. Therefore, many scholars have conducted research on multi-network data transmission methods. In literature [1], Shi Lingling and Li Jingzhao researched secure data transmission mechanisms in heterogeneous networks, mainly adopting an optimized AES-GCM certified Encryption algorithm combined with SHA-based Digital signature methods for data transmission; in literature [2], Zhou Jing and Chen Chen researched a data security model based on heterogeneous networks, with pre-encryption processing of data and establishing secure transmission channels for data transmission. Both methods can achieve certain effects but also have certain inadequacies. To address the above inadequacies, this paper applies Machine Learning methods to secure data transmission in multi-source heterogeneous networks to solve current issues. Experimental results indicate that the secure data transmission technology effectively solves existing problems, with practical application significance.

2 Multi-source Heterogeneous Network Data Preprocessing

In secure data transmission over multi-source heterogeneous networks, much data is useless, necessitating selecting relevant data sources for transmission to enhance accuracy and efficiency. During effective data source selection, the importance of relationships between Attributes is measured [3-4], capturing highly correlated data. The calculation expression is as follows: (1), T representing the comprehensive table number of all data sources, (i,j) representing the correlation between example source classes. Based on the importance judgment of data sources, the data table collection with the highest correlation degree can be selected, reducing unrelated tables. After completing the preferential data source selection, analyze the Attributes, as a data source is composed of a set of Attributes, whose features reflect the basic information of data to be transmitted. Mainly measured by the correlation of data tuples, analyzing the occurrence frequency of tuple data, defined by tuple data density, as shown in Figure 1. In Figure 1, ε represents the radius of the specified neighborhood. Following this idea, weights are assigned to each tuple data in the above dataset [5-7], with the expression as follows: (2), w(C) indicating Attribute weight, w(tk) indicating the number of core tuples, δ indicates outliers, w(tb) indicating the number of edge tuples.

3 Multipath Parallel Transmission Architecture

Upon completing the above preprocessing, establish a multipath parallel transmission architecture, mainly as follows: pre-divide traffic, traffic division is used at the sending end to split large data blocks into data units of different or same sizes [8], with size determined by the grain size of communication flow division. It mainly falls into the following categories: First, in packet-level business segmentation, packets are the smallest constituent units of data flow, so the granularity of segmentation methods is minimal, and packet probabilities are mutually independent and can be sent to the sending end; Second, flow-level traffic division [9], encapsulating specific destination addresses in header, then aggregating packets sharing the same destination address into data flows, these distinct data flows are independent and distinguished by unique flow identifiers. Flow-level division techniques effectively resolve data distortion impacts on multipath transmission [10]. Third, sub-flow level traffic division, the same destination header data flow is split into multiple sub-flows, all sub-flows' packets have the same destination address, somewhat resolving load imbalance in flow division algorithms. Multipath parallel transmission architecture is as shown in Figure 2. Besides, in bandwidth aggregation architecture, the scheduling algorithm determines business transmission patterns and business sub-flow scheduling order [11], ensuring orderly arrival of business sub-flows at the receiving end, next discussing data scheduling.

4 Bandwidth Scheduling Formulation

For data transmission over multi-source heterogeneous networks, when the bandwidth of a path reaches a certain value, the network bandwidth will continuously increase, and transmission performance becomes relatively stable. Excessive bandwidth allocation will reduce spectral utilization rate, leading to a waste of spectral resources. Under the increasingly tight spectral resources, scheduling and managing bandwidth across various paths in multipath parallel transmission not only ensures transmission performance but also efficiently utilizes resources. Processing mainly is as follows: First, adopting Machine Learning methods for effective bandwidth estimation, reasonably estimating wireless bandwidth resources each sub-flow can fully utilize, meeting high throughput requirements with less bandwidth resources, which is key to bandwidth scheduling algorithms. Therefore, a coupling congestion control algorithm is employed for combined control of sub-flows, with its expression as follows: (3), MSS represents the maximum length of messages constant, set by protocol, RTTi, PLRi respectively represent the return delay and packet loss rate of the sub-flow's path. Second, parameter filtering processing, due to wireless channel diversity and time-varying characteristics, link parameters and path effective bandwidth undergo dynamic changes and have errors. To remove errors, Kalman filter is applied to network parameters for precise estimates. The Kalman filter is a discrete time recursive estimation algorithm, computing more accurate current moment Status using differential recursion, current Status measurement, last moment Status, and forecast error. In discrete control systems research, linear stochastic differential equations are adopted as follows: (4), xk, xk-1 respectively represent Status parameters at moment k and k-1, Ak, Bk respectively represent system parameters, matrix form in multi-model systems, representing Status transition matrix and input matrix, uk represents control input parameters, wk represents noise during computation. Third, bandwidth scheduling, assuming multipath connection C contains n sub-flows, each sub-flow is independent, each occupying a single path for data transmission, the scheduling process is as shown in Figure 3. Based on the above bandwidth scheduling process, finally establish channel security protocol, ensuring secure data transmission over multi-source heterogeneous networks. Security protocol is composed of SSL protocol, rule establishment protocol, tunnel Information protocol, etc. Among them, SSL protocol mainly includes certification algorithm and Encryption algorithm, with all server-side messages encrypted through SSL protocol to ensure security of message communication, rule establishment protocol includes connection information and message identification, record table successfully matches, generating socket, relaying ensures information relay on VPN technology channel application. OpenVPN programming is the main method for realizing tunnel Information protocol. The client sends request command messages to establish connection with the server. After connection, the server writes encrypted and authenticated Information into tunnel Information area according to SSL protocol, achieving data exchange and transmission with the client. Channel security protocol structure is as shown in Figure 4. During data transmission, transmit according to the above channel security protocol to complete machine learning-based secure data transmission over multi-source heterogeneous networks.

5 Experimental Comparison

To verify the effectiveness of the designed machine learning-based secure data transmission technology for multi-source heterogeneous networks, experimental analysis was conducted, comparing the heterogeneous network secure data transmission mechanism in literature [1], heterogeneous network-based data security model in literature [2], and effectiveness of three systems. Experimental datasets as shown in Table 1. Based on collected experimental data, it's evident that selected experimental data is increasing, better verifying the effectiveness of three methods, mainly comparing transmission delay, data transmission interruption situations, and link packet loss rate, as detailed below.

5.1 Comparison of Transmission Delay

We compared the transmission delay of three methods, and the comparison results are shown in Figure 5. By analyzing Figure 5, it is found that for transmission on Google's public dataset, the transmission delays of all three methods are relatively small. However, as the amount of transmitted data increases, the transmission delays in all three methods increase. Nonetheless, by comparison, it is evident that the transmission delay of the machine learning-based multivariate heterogeneous network data security transmission technology in this study is the smallest, less than the other two traditional methods.

5.2 Comparison of Data Transfer Interruption

We compared the data transfer interruption situations after applying three transmission technologies, and the comparison results are shown in Figure 6. Figure 6 reveals that the transmission technology in this study experiences the least data transfer interruptions, with fewer occurrences than the two traditional transmission technologies in several experiments.

5.3 Comparison of Packet Loss Rate

The study's machine learning-based multivariate heterogeneous network data security transmission technology and two traditional transmission technologies were used for data transmission, and the packet loss rate comparison results of the three methods are shown in Figure 7. By analyzing Figure 7, it is found that the link packet loss rate in traditional heterogeneous network secure data transmission mechanisms is the highest, greater than the data security model based on heterogeneous networks and the transmission technology in this study. To sum up, the machine learning-based multivariate heterogeneous network data security transmission technology in this study has lower transmission delay and packet loss rate compared to the two traditional transmission technologies. This is because the study's transmission technology preprocessed multivariate heterogeneous network data in advance, formulated bandwidth scheduling schemes, and established secure transmission protocols, thus improving the effect of secure transmission in multivariate heterogeneous networks.

6 Conclusion

This paper designed a machine learning-based multivariate heterogeneous network data security transmission technology, and experiments validated the effectiveness of this research technology. This technology can improve the efficiency of data transmission and reduce the packet loss rate of data transmission, and has significant practical application value. However, due to research time limitations, the multivariate heterogeneous network data security transmission technology in this study still has certain shortcomings. Thus, further optimization is required in subsequent research.

Abstract: It explains that Virtualization Technology ensures the stability and smoothness of information usage, Cloud Storage Technology ensures the rational allocation of data volumes, and Information Security Technology ensures the security of big data usage and browsing.

Keywords: Computer System, Big Data, Cloud Storage, Virtualization.

0 Introduction

Computer software technology can process large amounts of data in a short period, conduct analysis by employing certain logic, extract relevant data information needed by users, perform reprocessing, and determine the data content associated with user-needed data analysis.

1 Virtualization Technology

Virtualization Technology is an innovative technology of computer software technology, capable of creating a new virtual machine for users within a short period. Virtualization Technology has effectively achieved rational utilization of information resources, efficient configuration of software resources and mobilization, rational allocation and usage of computer software resources, and ensures that computers will not become stuck or slow due to uneven allocation of software resources. Flexibility is a prominent feature of Virtualization Technology; it can perform operations and calculations on the virtualized computing components, realizing cross-domain sharing and cooperation of the computer, processing, and switching resources required by users to form a new resource chain. Virtualization Technology mainly includes the following categories: Server Virtualization and Docker Container Technology, with particular emphasis on Docker Container Technology. Server Virtualization is built on the basis of computer multi-dimensional virtualization, transforming one mainframe computer into multiple virtual logic-related computers, establishing a virtualization hierarchy to interlink the computer's hardware and logical associative systems, and achieving specific functions through decoupling connections. The so-called virtualization hierarchy refers to the ability to run multiple virtualized operating systems on one physical computer, which can be switched between each other, and these virtual computers at these levels can share one or several exclusive software and hardware resources.

If you find any errors ( broken links, non-standard content, etc.. ), Please let us know < report chapter > so we can fix it as soon as possible.