Progressive lossless service aggregation and arrangement method based on active trust

Abstract

The invention discloses a progressive lossless service aggregation and arrangement method based on active trust. The method can be applied to the data-intensive Internet of Things to solve the problems of trust risk and transmission load caused by explosive data increase. Firstly, according to the method, an active trust mechanism is utilized to carry out credibility evaluation on a data reporter, unmanned aerial vehicle equipment trusted by a system is dispatched to a reference address point to obtain sample data, and the reliability of the data reporter is evaluated by comparing data consistency. Then, according to the method, trusted data are arranged as lightweight services on an edge layer, and callable services of different levels and scales are synthesized hop by hop during network routing. According to the method, a multi-stage gradual credibility and quality enhancement data service calculation and aggregation process is provided, and compared with an existing data center network, the data credibility is ensured, and meanwhile, the network transmission load and processing pressure are effectively reduced.

Description

Technical Field

[0001] The present invention belongs to the field of new generation information technology and relates to a progressive lossless service aggregation and orchestration method based on active trust. Background Art

[0002] In recent years, with the development of the Internet of Things and mobile communication technology, the scale of smart terminal devices and the amount of data they generate have shown an exponential growth trend. According to a report by Cisco, by 2030, the global mobile data growth rate will be 20,000 times that of 2010, and the number of IoT device connections will be close to 100 billion. These devices generate at least 2.5 trillion bytes of data every day. At the same time, the emergence of new business scenarios in the 5G era, such as unmanned driving, face / fingerprint / iris recognition, augmented / virtual reality, natural language processing, interactive games and other low-latency ultra-intensive computing applications, has put forward higher requirements for network transmission rate, task processing delay, energy efficiency and other indicators. Therefore, the current network faces huge challenges, which are mainly manifested in the following two aspects:

[0003] First, the emergence of the Internet of Things and 5G emerging applications has brought about an explosive growth in data. In the current data center network architecture, the raw data collected by the underlying sensing devices must be uploaded to the data center via a long network route, and returned to the user via the same long network route when requested, which brings huge transmission pressure to the network layer relay equipment. However, the current growth of equipment and transmission capacity of the backbone network in the central area of ​​the network is very limited. Most of the existing methods improve the transmission performance cost by upgrading or adding hardware, which is not only expensive, but also difficult to redeploy the network in some special environments. Therefore, the gap between the network bandwidth required by the application and the transmission rate of the existing network is getting bigger and bigger. The method of the present invention finds that the data perceived by different terminal devices at the same time or space often have a strong correlation. Therefore, by strengthening the processing of the transmitted data to reduce the network load, a data service computing mechanism is proposed. The main idea is that for the massive raw data collected by the underlying sensing device, the redundant invalid data in the data packet is first removed, and the data is arranged into services using the code program distributed by the data center at the edge layer. Experiments show that the amount of transmitted data can be reduced by more than an order of magnitude. On this basis, in the process of forwarding to the data center via the backbone routing layer, the present invention utilizes the high-trust data control principle to perform service aggregation hop by hop, thereby further improving network transmission performance.

[0004] Secondly, with the growth of connected smart terminal devices, the sources of data collection are wide and messy, and it is difficult for the system to determine the authenticity and security of the data. The processing of network events depends on reliable perception data. Unreliable data will lead to wrong decisions, resulting in huge losses of manpower, material and financial resources. In addition, the data collected by different terminals vary in quality. How to select high-quality data at the same cost is also a key issue. The trust mechanism is a simple way to ensure data security and reliability. Different trust evaluation values ​​are made according to the performance of smart terminals in the data collection process, and in the process of data calculation and service aggregation, data with higher trust is preferentially selected to ensure the quality of network data collection. In the previous trust evaluation mechanism, trust evaluation was mainly based on observing the interactive behavior and relationship of the evaluated object. However, this method is subjective, passive and not suitable for sparse interaction scenarios, and is therefore not suitable for critical, delay-sensitive, and high-reliability applications. Different from the previous trust evaluation mechanism, the present invention adopts an active trust mechanism. Its main idea is that the system sends out its own trusted drone detection equipment to perceive and collect data at specific evaluation points, and then uses this as the benchmark data to compare with the data reported by other intelligent terminal devices in the network, and adjusts the trust of the evaluation object according to the consistency of the data. This active trust identification mechanism can initiate detection of the evaluation object on demand. For example, when the network data quality is good, it can initiate no or fewer trust detections to reduce costs. When the network data quality is reduced, it can initiate trust detection in a targeted manner for a specific area. It has the application advantages of low cost, on-demand, and dynamic.

[0005] In summary, it can be seen that the present invention is particularly suitable for large-scale data-intensive Internet of Things. On the one hand, by performing trust evaluation on the underlying sensing devices, the authenticity and credibility of the collected data is guaranteed from the source; on the other hand, through data service computing and multi-hop lossless service aggregation, the amount of transmitted data is reduced. Especially in long-term data-intensive Internet of Things, the method of the present invention can greatly save network transmission costs while improving service quality. Summary of the invention

[0006] The present invention provides a progressive lossless service aggregation and orchestration method based on active trust. First, an active trust mechanism is used to implement credibility assessment on data reporters. Then, the trusted data is orchestrated into lightweight services at the edge layer. Multi-level progressively trusted and quality-enhanced data service calculation and aggregation are performed during network routing. The purpose is to solve the trust risk and transmission load problems caused by the explosive data growth in the current data-intensive Internet of Things.

[0007] The technical solution adopted by the present invention is as follows: Step 1, various types of information data such as urban weather, sound, and images are collected through various types of sensing devices in the data collection layer, and trust evaluation is performed on the massive data reporter according to the active trust evaluation mechanism. The trust benchmark evaluation point is first determined based on the data point perception density coefficient, and then the system-trusted drone equipment is dispatched to the benchmark address point to actively collect sample data, and this is used as the benchmark data to compare the consistency and difference value between the benchmark data and the submitted data, and the credibility of the data reporter is evaluated based on the system's trust threshold δ1; Step 2, retain the true and credible The original collected data is arranged into lightweight services at the edge layer. First, the original data is cleaned and filtered to remove redundant, invalid, and false information fields; then, the retained valid data is fused and calculated, and multiple discrete data are arranged into a lightweight service package through association, reconstruction, and integration, and the service package is formatted to reduce the amount of transmitted data; Step 3, multi-hop lossless progressive service aggregation is performed during network routing. For each hop router passed through, a multi-hop lossless aggregation model is used to aggregate each newly arrived service data packet with the existing message. Assuming that the existing message is S old , the new arrival service is S new , the service aggregation operation can be expressed as:

[0008]

[0009] in, is the data correlation coefficient, d is the physical distance, θ is the correlation coefficient constant, and when θ→0, it means S new and S old Unrelated, when θ→1, the higher the correlation between the two; service aggregation is completed by the built-in software code of the router, and the principle of high-trust data dominance is adopted to build multi-level gradually trusted and quality-enhanced callable services. Finally, when the user initiates a service request through the application terminal, the request will first be locally parsed and service queried at the edge layer. If it can be satisfied, the data will be returned directly, otherwise it will be forwarded to the backbone network layer and the data center layer, and finally the result will be returned to the requesting terminal.

[0010] The method of the present invention has the following beneficial effects: first, the method of the present invention can solve the problems of previous trust evaluations that rely on interactive relationships, are inaccurate, and are not applicable to sparse interaction scenarios through an active trust evaluation mechanism; second, the method of the present invention introduces indirect trust evaluation on the basis of active trust evaluation, which can speed up the trust evaluation speed and expand the scope of trust evaluation; finally, the method of the present invention can reduce the amount of transmitted data by more than one order of magnitude through a data service computing mechanism, and in service aggregation, high-trust data dominates low-trust data, thereby improving service credibility and quality. In general, compared with the existing data center network, the present invention effectively reduces the network transmission load and processing pressure while ensuring data credibility. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 A schematic diagram of a network architecture for application of the present invention;

[0012] Figure 2 A schematic diagram of implementing active trust assessment in the present invention;

[0013] Figure 3 A schematic diagram of trust reasoning and evolution for the present invention;

[0014] Figure 4 A schematic diagram of service aggregation for implementing the present invention;

[0015] Figure 5 A schematic diagram showing the comparison of the data credibility between the method of the present invention and the comparative method;

[0016] Figure 6 It is a weighted cost comparison diagram of the method of the present invention and the comparative method;

[0017] Figure 7 A schematic diagram showing a comparison of energy consumption between the method of the present invention and a comparative method;

[0018] Figure 8 It is a schematic diagram of delay comparison between the method of the present invention and the comparative method. DETAILED DESCRIPTION

[0019] The present invention is further described below in conjunction with the accompanying drawings and examples.

[0020] Figure 1 The figure shows a schematic diagram of the network architecture of the present invention. The entire system is divided into a data collection layer, an edge layer, a backbone network layer, a data center layer, and an application layer. The data collection layer is composed of various sensing devices. By deploying a large number of devices with sensing capabilities in a specific area, data sensing and acquisition of the surrounding environment can be achieved. Figure 1 Taking the urban data collection scenario shown in the figure as an example, a large number of sensing devices in the city, such as mobile phones, smart computers, smart cars, monitoring probes, etc., can be used to collect various types of information such as urban weather, sound, and images. These devices deployed in different scenarios for massive data collection are called data reporters. In order to ensure the authenticity and credibility of the perceived data, the method of the present invention introduces an active trust evaluation mechanism in the data collection layer. The mechanism first selects the benchmark evaluation point and performs a global analysis of the current data credibility of the system according to the application requirements. Assume that the credibility required by the system for the data is C req , the actual data credibility of the current system is C cur , calculate the credibility difference between the two ΔC=C req -C cur. Next, the area and big data reporter of the data that need to be trusted are determined based on ΔC. When selecting the benchmark evaluation point, the perception density coefficient of the evaluation point is used as a reference. Suppose the number of data reporters perceived by a certain evaluation point in an average time unit is n, and the area covered by the evaluation point is S, then the perception density coefficient is The larger the perception density coefficient, the fewer benchmark data points can be selected to evaluate the trust of more data reporters. Then comes active trust acquisition and evaluation. After determining the benchmark evaluation point, the system sends a trusted detection device (such as a drone) to obtain data from the evaluation point. Suppose the data value perceived by the device sent by the system is Data Reporter BDC i The value of the perception data is For the same observation object at the same time, if (where δ1 is a pre-set direct trust threshold), then the BDC is increased i trust; if Then reduce BDC i The trust level of each increase or decrease is Where ɑ is a coefficient adjusted according to the data difference. The edge layer is composed of various heterogeneous distributed computing devices with different computing and storage capabilities, such as gateways, routers, and wireless access points. These devices can not only realize data routing, but also realize simple calculation and processing of the original data uploaded by the data acquisition layer through the software code distributed by the data center. These devices are located at the edge of the network and are the key link between the data acquisition layer and the backbone network layer. After the original data is uploaded to the edge network layer, the servers and distributed computing devices of the edge layer use the locally resident software code and hardware programming to process the data. First, the data is cleaned to remove redundant, invalid, and false information fields. Then, the built-in hardware and software code of the edge device is used to organize multiple discrete data into a lightweight service package through association, reconstruction, and integration, and the service package format is encapsulated to make it more suitable for transmission and subsequent processing. The encapsulated service package has a unified format and specification, which is convenient for transmission in the network. It is also conducive to the backbone network layer and the data center layer to identify and process it. The processed and organized service package will be uploaded to the backbone network layer for further aggregation and transmission to the data center layer. The backbone routing layer is composed of core routers with stronger computing and storage capabilities. These core routers are key nodes of the network, equipped with high-performance processing chips and large-capacity storage devices. They can quickly process large amounts of data traffic and ensure stable data transmission. They are also responsible for forwarding services uploaded by the edge layer to the data center. The core routers are connected by high-speed links to form an efficient transmission network. When the service packet uploaded by the edge layer reaches a core router of the backbone routing layer, the router will accurately forward the service packet to the next core router according to the pre-set routing strategy until the service packet finally reaches the data center. For each hop router passed through, a multi-hop lossless aggregation model is used to aggregate each newly arrived service data packet with the existing message. Assuming that the existing message is S old , the new arrival service is S new , the service aggregation operation can be expressed as:

[0021]

[0022] in, is the data correlation coefficient, d is the physical distance, θ is the correlation coefficient constant, and when θ-0, it means S new and S old No correlation, when θ→1, the higher the correlation between the two. The specific process is as follows: First, the data correlation is determined: the router first determines the data correlation coefficient. Determine the newly arrived service S new and the existing message S oldIf θ is close to 0, it means that the two are less correlated, and they may come from different data sources or data collection areas; if θ is close to 1, it means that the two are highly correlated, and they may be related data of the same object or event; then the aggregation operation is performed: according to the judgment result, if S new and S old If the correlation is high, the aggregation will tend to retain the high-quality data in both and calculate according to the aggregation formula. new A data field in S old If the corresponding field in is more accurate or more valuable, S will be preferred in the aggregation process. new If S new and S old The correlation is low and the calculation is still performed according to the aggregation formula, but Min(S new , S old) is relatively small, and its impact on the aggregation result is relatively small. Each time a hop router is passed, an aggregation operation will be performed on the service. As the service data packet is transmitted in the backbone routing layer, its data volume will gradually decrease, while the quality and relevance of the data will gradually increase. Finally, data quality and credibility are enhanced: in the service aggregation process, the principle of high-credibility data dominance is adopted to build a multi-level gradually credible and quality-enhanced callable service. That is, when multiple service data packets pass through a router, the aggregation operation is performed according to the above model and principle, and the related service data packets can be effectively integrated, further reducing the amount of transmitted data while ensuring the credibility and quality of the service. For example, if the trust of one service data packet is high and the trust of another service data packet is low, the data in the data packet with high trust will be preferred during aggregation, thereby improving the overall credibility of the aggregated service. This multi-level gradually credible and quality-enhanced mechanism enables the service to be continuously optimized during the transmission process of the backbone routing layer, and the service that finally reaches the data center has high quality and credibility, which can better meet the needs of users and application requirements. Through the above steps, even service data packets with low relevance can be integrated to a certain extent to reduce the amount of transmitted data. The data center layer is composed of large-scale equipment with huge storage capacity and computing power, such as large-scale server clusters, high-performance storage arrays, and processors with powerful computing power. These devices work together to form a core hub that can process massive amounts of data and is responsible for receiving services from the backbone routing layer. After multi-hop lossless service aggregation, the backbone routing layer transmits high-quality and reliable services to the data center layer. The data center layer has a strong storage capacity and can store the received services for a long time for subsequent analysis and processing, and can also perform in-depth analysis and processing on the received services. When a user at the application layer initiates a service request through an application terminal, the request will be processed by the edge layer and the backbone network layer. If the request cannot be met at these two layers, it will eventually reach the data center layer. The data center layer extracts relevant information from the stored services based on the user's request content and returns the processed results to the application layer to meet the user's needs. And manage and optimize the entire service life cycle. In the method of the present invention, since the operation of converting data into services is realized at the edge layer, the data center mainly further processes the received services and returns them to the user when the user requests. The application layer is composed of various application terminals, which are used to request service data from the system. For example, for applications targeting urban security, the application terminals include city buses, private cars and other transportation vehicles that request real-time traffic information; user mobile phones, computers and other communication devices that request weather information and public data. The main function of the application layer is to request service data from the system.Different application terminals initiate corresponding service requests based on the user's specific needs and provide users with interactive interfaces so that users can easily operate and obtain the required services. They can also process and present service data according to the specific logic of the application, and feed back the user's operations to the system, and update according to the service results returned by the system.

[0023] The system under the method of the present invention includes three transmission streams, namely data stream, service stream and service request stream. First, various data reporters located in the data collection layer perceive and collect data, and transmit the data to the edge layer in the form of data stream. In this process, the system performs trust evaluation on the data reporter and the data collected based on the active trust evaluation mechanism. Then, the edge layer arranges the received raw data into services and transmits them to the backbone routing layer in the form of service stream. The backbone routing layer aggregates the services hop by hop, and forwards the aggregated new services to the data center, which stores the services after in-depth processing. When a user requests service data, a service request is initiated to the network system through the application terminal, a service request stream is generated and forwarded to the edge layer, and the edge layer device locally parses and performs service query. If the requested service can be satisfied at the edge layer, it is directly returned from the edge layer, otherwise it is forwarded to the backbone routing layer. If the request cannot be satisfied at the network layer, it is finally forwarded to the data center, and the data center returns the service to the application layer.

[0024] Figure 2 The figure shows a schematic diagram of the active trust evaluation implemented in the present invention. The active trust evaluation includes three parts, namely, the selection of benchmark evaluation points, active trust acquisition and evaluation, and trust reasoning and evolution.

[0025] First, the process of selecting benchmark evaluation points is as follows: The method of the present invention first comprehensively analyzes the specific requirements of various applications for data credibility in the current operating state of the system. Different application scenarios may have different emphases and standards for data accuracy, completeness, consistency, etc. For example, in intelligent transportation systems, the accuracy of vehicle position and speed data is extremely high; while in environmental monitoring applications, there are high requirements for the stability and reliability of environmental data collected by sensors. Suppose the system's quantified value for the credibility of data for a specific application is C req By comprehensively evaluating the data currently collected and processed by the system, the actual quantitative value of the data credibility of the current system is obtained as C cur . Calculate the credibility difference between the two ΔC=C reg -C cur, this difference value reflects the degree of gap between the current data credibility of the system and the application requirements. Next, determine the regional scope of the data to which the trust enhancement needs to be focused and the related big data reporter group based on ΔC. Specifically, if ΔC is large, it means that the current data credibility is far from the application requirements, and it is necessary to improve the data trust in a wider area; conversely, if ΔC is small, the focus can be placed on a relatively small specific area. When actually selecting the benchmark evaluation points, the perceptual density coefficient of the evaluation points is introduced as a key reference indicator. Suppose the number of data reporters perceived by a certain evaluation point within the average time unit is n, and the area covered by the evaluation point is S, then the perceptual density coefficient The larger the perception density coefficient, the denser the distribution of data reporters in the area, and the more frequent and large the data collection frequency. In this case, the data provided by each data reporter can be more mutually verified and compared, so that fewer benchmark data points can be selected to more accurately evaluate the trust of more data reporters. For example, in a central area of ​​a city, where smart devices are densely distributed and the perception density coefficient is high, it may only take a few key locations as benchmark evaluation points to effectively evaluate the trust of a large number of data reporters in the surrounding area; while in some remote areas, where devices are sparsely distributed and the perception density coefficient is low, more benchmark evaluation points need to be selected to ensure the accuracy and reliability of the trust evaluation.

[0026] Then, the active trust acquisition and evaluation process is as follows: After the benchmark evaluation point is determined, the system dispatches its own trusted detection equipment to actively obtain data from the evaluation point. The data perceived by these devices dispatched by the system itself is used as the benchmark data of the evaluation system, and the data submitted by the data reporter is called the report data. For the same observation object at the same time, if the report data is consistent with the benchmark data, it means that the report data is true. At this time, the data reporter is considered to be trustworthy, and his trustworthiness is improved to facilitate the system to ensure data quality when selecting data later. If the report data is inconsistent with the benchmark data, it is considered that the data reporter has distorted the real data, and the credibility of the data reporter is reduced. Assume that for a certain observation object, the data value perceived by the device dispatched by the system is φ B , this data value will serve as one of the important reference standards for subsequent trust evaluation. It represents the most reliable data sample obtained by the system at a specific time and location. i The value of the perception data is Where i represents the number of different data reporters. Each data reporter has its own unique location and perception characteristics, so the data they perceive may be different. The system pre-sets a direct trust threshold δ1, which is one of the key criteria for measuring the trust of data reporters. It is based on the system's requirements for data accuracy and reliability, as well as the analysis and experience of historical data. If Then increase BDC i If Then reduce BDC i The degree of trust each time is increased or decreased Adjust according to the data difference, where α is the adjustment coefficient. The value of ɑ is not fixed, it will be adjusted appropriately according to the size of the data difference and the specific application scenario. When the data difference is small, that is, Close to δ1, ɑ can be relatively small to avoid over-adjustment of trust. This is because in this case, the data reporter may only be slightly affected by interference or error, and his overall credibility is still high. When the data difference is large, α will increase accordingly to more significantly reflect the change in the credibility of the data reporter. For example, if If it is far more than δ1, it means that there may be serious problems with the data reporter, and the trust level needs to be reduced more significantly to ensure that the system's trust evaluation of the data is accurate and reliable. Through the above active trust acquisition and evaluation process, the system can monitor and evaluate the credibility of the data reporter in real time, providing an important basis for subsequent data processing and decision-making.

[0027] Figure 3 The figure shows a schematic diagram of trust reasoning and evolution implemented in the present invention. The method of the present invention uses the trusted data reporter obtained in the direct trust evaluation to perform trust reasoning and evolution. The main idea is: the present invention uses the trusted data reporter obtained in the direct trust evaluation to perform trust reasoning and evolution. In the direct trust evaluation process, the data value (set as φ) perceived by the device (such as a drone) dispatched by the system B ) and data reporter BDC i The value of the perception data (set to ) for comparison, based on The relationship between the direct trust threshold δ1 determines the direct trust of the data reporter. Then increase BDC i trust; if Then reduce BDC i The trust level of each increase or decrease is Where α is a coefficient adjusted according to the data difference. If the system evaluates that the data reporter is trustworthy, then the data of other address points reported by him are also trustworthy. At this time, the data reported by those trustworthy data reporters are used as sub-benchmark numbers, and the sub-benchmark data is used as the evaluation standard to conduct indirect trust evaluation on other uncertain data reporters who interact with the trustworthy data reporter. Similar to direct trust evaluation, it is assumed that for a certain benchmark evaluation point, the trustworthy data reporter BDC k The perceived data value is φ T_k , Uncertain Data Reporter BDC m The value of the perception data is if Then increase BDC m If Then reduce BDC m The trust level of Where γ is the indirect trust adjustment coefficient. k The trustworthiness of the system equipment is not as high as that of the direct trust. The method of the present invention allows the data difference δ2 of indirect trust to be slightly larger than the data difference δ1 of direct trust, δ2>δ1. At the same time, the adjustment range of each trust degree of indirect trust is lower than the adjustment range of direct trust degree, γ<α. Through the trust reasoning and evolution process, the system can further expand the scope of trust evaluation on the basis of direct trust evaluation.

[0028] Figure 4 The figure shows a schematic diagram of service aggregation for the implementation of the present invention. Assume that three data packets D1, D2, and D3 are transmitted via an edge device. These data packets are raw data collected by data reporters, so there may be invalid or redundant data. Therefore, the first step in converting data into services is data processing. According to the trust evaluation of the data reporter and the data by the data acquisition layer, it can be determined whether each data is authentic and reliable. Invalid or false data is directly eliminated. For redundant data between multiple data packets, high-quality data is preferentially selected based on their trustworthiness. Then, the valid data retained after data processing is subjected to data fusion, and multiple discrete data are placed in a service package through association and integration. Finally, the service package is encapsulated. When the data packet is converted into a service package, its package format changes.

[0029] Under the method of the present invention, the service package will also be gradually trusted and quality-enhanced service aggregation at the backbone network layer. For any router, service aggregation refers to the process of processing all the service package data received according to certain aggregation rules every unit time, and encapsulating the data retained after processing to form a new service package. When the service is aggregated, the stronger the correlation between the service packages, the higher the service aggregation rate, and the smaller the amount of data generated after aggregation. The input of service aggregation is the service package from the edge layer. The edge layer preprocesses the original data, including removing redundant, invalid, and false information fields, and arranging multiple discrete data into a lightweight service package. For example, from a large amount of sensor collection data, after processing by the edge layer, a service package for a specific application scenario (such as a meteorological data service package for a certain area in environmental monitoring) is formed. These service packages have a certain format and structure to facilitate transmission and processing at the backbone network layer. Each service package contains filtered and sorted data, and its data volume has been greatly reduced relative to the original data, but the key information is retained. As the executor of service aggregation, the core router has powerful computing and storage capabilities. They can receive, process and forward service packets, and perform aggregation operations on services according to the multi-hop lossless aggregation model. The present invention adopts the multi-hop lossless aggregation model to aggregate each newly arrived service data packet with the existing message. Assuming that the existing message is S old , the new arrival service is S new , the service aggregation operation can be expressed as:

[0030]

[0031] in, is the data correlation coefficient, d is the physical distance, θ is the correlation coefficient constant, and when θ→0, it means S new and S old Unrelated, when θ→1, the higher the correlation between the two. new A data field in S old If the corresponding field in is more accurate or more valuable, S will be preferred in the aggregation process. new This is because high-quality data is more important for the final service quality and accuracy. new and S old The correlation is low and the calculation is still performed according to the aggregation formula, but Min(S new , S old) is relatively small, and its impact on the aggregation result is relatively small. In this way, even service data packets with low relevance can be integrated to a certain extent, reducing the amount of transmitted data. This reflects the flexibility of the multi-hop lossless aggregation model, which can adapt to service packets with different relevance for effective aggregation operations. Every time a hop router is passed, the service is aggregated once. As the service data packet is transmitted in the backbone routing layer, its data volume will gradually decrease, while the quality and relevance of the data will gradually increase.

[0032] Service aggregation includes three factors, input, aggregation rules and output. The input is the service package that needs to be aggregated, and the output is the new service package obtained after service aggregation. Aggregation rules are specific operations to be performed for service aggregation. Due to different data requirements, different applications often have different aggregation requirements. The present invention lists four different aggregation rules. The first is redundant aggregation, which means retaining the full set of valid data of input data in service aggregation. The second is collaborative aggregation, which means eliminating heterogeneous attribute data during aggregation. The third is service aggregation similar to data extraction, which only retains the maximum or minimum value in the data, and sometimes it can also be the average or mode. The fourth is to screen data based on set thresholds. For example, in a forest fire prevention system, only when the temperature is higher than the threshold, statistical information is needed and sent to the data center. Therefore, in this case, only valid data that meets the conditions is retained. In addition to the above, there may be other aggregation rules, which are implanted into the processing unit of the network device through software programming, and then implemented by calling hardware technology.

[0033] The realization of service aggregation is completed through the built-in aggregation unit of the router, and the device automatically associates, relates, estimates and combines the data. In the past, service aggregation was generally divided into three aggregation areas and one management unit. The first is the data preparation area, which mainly performs data preprocessing and zero-level joint detection; the second is the target aggregation area, which performs aggregation processing of the detection target, including first-level target state estimation and second-level target identity judgment; the third area is the situation assessment area, which includes third-level situation assessment and fourth-level security estimation; the management unit refers to the fifth-level aggregation management, which mainly estimates and optimizes the efficiency of the service aggregation system. In the method of the present invention, we introduce trust-based data selection between the first area and the second area. After preprocessing the data in the first area, the trust of each data packet is evaluated using the data dominance principle, and data with high credibility is preferentially selected as the aggregation target. When a high-credibility data packet is selected, a low-credibility data packet covering similar data will be discarded, thereby ensuring that the data of service fusion is credible and high-quality.

[0034] Figure 5Shown is a schematic diagram comparing the data credibility of the method of the present invention and the comparative method. Shown is the data credibility results under different network operation rounds obtained by comparing the method of the present invention with three other communication schemes that use the same routing algorithm and data processing technology as the present invention. Among them, one round of network operation refers to the time when all data packets generated by the data acquisition device within one cycle are received by the data center. Since the present invention and comparative method 3 introduce a trust evaluation mechanism, high-quality data is preferentially selected for service fusion during the data transmission process, while low-quality malicious data is discarded, so their data reliability is higher. At the same time, compared with comparative method 3 of passive trust evaluation, the data credibility of the active trust evaluation proposed in this article is higher, indicating that the ability of the present invention to identify malicious data is stronger than the previous passive trust evaluation scheme.

[0035] Figure 6 The figure shows a weighted cost comparison diagram of the method of the present invention and the comparative method. Energy consumption and delay are two important evaluation indicators. They are weighted and recorded as network transmission costs. The cost of the four solutions can be shown as follows: Figure 6 Among them, the cost of the method of the present invention is the lowest, followed by comparison method 3, while the transmission cost of the previous data center communication solution is relatively high.

[0036] Figure 7 The energy consumption comparison diagram of the method of the present invention and the comparative method is shown in FIG. By converting data into services and performing hop-by-hop service aggregation, the amount of data transmitted is greatly reduced, so the energy consumption of the method of the present invention is significantly minimized.

[0037] Figure 8 The figure is a schematic diagram of the delay comparison between the method of the present invention and the comparative method. Since the amount of transmitted data is reduced, the transmission delay of the method of the present invention is reduced, and at the same time, the edge server is introduced to provide some nearby service storage and computing services, and the request response speed is improved.

Claims

1. A progressive lossless service aggregation and orchestration method based on active trust, characterized in that: The active trust mechanism is used to conduct credibility assessment on the reporters of massive data, and the data is arranged into lightweight services during transmission to build a multi-level integrated, gradually trusted, quality-enhanced service aggregation and transmission architecture, which mainly includes the following steps: Step 1: Active trust evaluation is performed on the reporters of massive data. First, the trust benchmark evaluation points are selected based on the data point perception density coefficient. Then, the drone equipment trusted by the system is dispatched to the benchmark address point to actively collect sample data. Finally, the consistency and difference values ​​of the benchmark data and the submitted data are compared, and the credibility of the data reporter is evaluated based on the system's trust threshold δ1. Step 2: Arrange the original data into lightweight services at the edge layer. First, clean and filter the original data to remove redundant, invalid, and false information fields. Then, perform fusion calculations on the retained valid data, and arrange multiple discrete data into a lightweight service package through association, reconstruction, and integration, and encapsulate the service package in a format. Step 3: Perform multi-hop lossless progressive service aggregation during network routing. For each router passed through, a multi-hop lossless aggregation model is used to aggregate each newly arrived service data packet with the existing message. Assume that the existing message is S old , the new arrival service is S new , the service aggregation operation can be expressed as: in, is the data correlation coefficient, d is the physical distance, θ is the correlation coefficient constant, and when θ→0, it means S new and S old Unrelated, when θ→1, the higher the correlation between the two, the higher the degree of correlation; service aggregation is completed by the built-in software code of the router, and the principle of high-trust data dominance is adopted to build multi-level gradually trusted and quality-enhanced callable services.

2. The method for progressive lossless service aggregation and orchestration based on active trust according to claim 1, characterized in that: The specific implementation of step one is as follows: ① Determine the trust benchmark evaluation point selection strategy based on the data point perception density coefficient. The data point perception density coefficient refers to the number of data reporters whose data is collected per unit time; ② Send a drone device with high-precision positioning and data collection functions that the system trusts to the benchmark address point. The drone device follows the preset safe route and energy-saving mode during the flight. After arriving at the benchmark address point, it actively collects sample data according to the predetermined collection specifications and frequency; ③ Use an accurate data analysis algorithm to compare the consistency and difference value between the benchmark data and the submitted data. The algorithm takes into account the numerical difference, data structure difference and time series characteristics of the data, and evaluates the credibility of the data reporter based on the system preset trust threshold. Different weight coefficients are set for different types of data differences during the evaluation process.

3. The method for progressive lossless service aggregation and orchestration based on active trust according to claim 1, characterized in that: The specific implementation of step two is as follows: ① Use a combination of rule-based and machine learning methods to clean and filter the original data. The rules include preset invalid data patterns and redundant data judgment criteria. The machine learning model identifies false information fields by learning from a large amount of historical data; ② Use an optimized fusion calculation algorithm to perform fusion calculations on the retained valid data. When associating, reconstructing, and integrating multiple discrete data, the algorithm fully considers the semantic associations, logical relationships, and business needs between the data, and arranges them into a lightweight service package; ③ The service package is formatted and encapsulated according to specific communication protocol standards. During the encapsulation process, the header information, data segment format, and check bits of the service package are strictly defined.

4. The method for progressive lossless service aggregation and orchestration based on active trust according to claim 1, characterized in that: In step three, the calculation of the data correlation coefficient comprehensively considers factors such as network topology, data transmission path stability, and data content correlation; service aggregation is completed by the router's built-in efficient software code, which is designed based on the principle of high-trust data dominance and has data quality detection, error recovery, and service optimization functions, building a multi-level, gradually trusted and quality-enhanced callable service.

Images