A verifiable trust evaluation mechanism for ultra-reliable applications in 5G and beyond networks

Highlights

• A novel verifiable trust evaluation joint UAV (VTE-UAV) mechanism is proposed.

• An aggregation-based trust evaluation and an active trust acquisition joint UAV are proposed for cross trust evaluation and verification of MEUs.

• A trust-aware greedy selection algorithm is designed to recruit trustworthy MEUs to conduct the task with low cost.

• Simulations demonstrate the better performances of VTE-UAV scheme.

Abstract

With the development of Internet of Thing (IoT) joint 5G and Beyond Networks, Mobile Edge Users (MEUs) can act as mobile data collectors to collect data for various applications. However, some malicious MEUs reporting false or malicious data can cause serious harm to applications, especially for ultra-reliable applications. A novel Verifiable Trust Evaluation joint UAV (VTE-UAV) mechanism is proposed to select trustworthy MEUs to conduct the task for ultra-reliable applications. The VTE-UAV strategy adopts two novel trust evaluation methods, one is the aggregation-based MEU trust evaluation mechanism, when malicious evaluation objects are in the minority, the mechanism takes most evaluation results as baseline data. The other is an active trust acquisition mechanism, it takes the data obtained by Unmanned Aerial Vehicles (UAVs) as baseline data to actively validate the authenticity of the data. Through these two cross evaluation strategies, we obtain more accurate trust evaluation results. Finally, this paper transforms the trust evaluation optimization problem into the optimization of the accuracy of trust evaluation with reducing the cloud payment and the dispatch cost of UAVs. Extensive experiments have verified the validity of the VTE-UAV strategy. Compared with the previous strategies, the VTE-UAV improves the cloud recruitment performance by 7.74%-25.91%, increases the accuracy of trust evaluation of IoT devices by 2.24%-11.72%, and reduces the cloud payment and the cost of UAVs by 3.11%-10.20% and 58.23%, respectively.

Introduction

The overwhelming growth of the IoT [1], [2], [3] along with the rapid advances in 5G and beyond networks [4], [5], [6] have led to the emergence of various data-based applications [7], [8], [9]. The year 2020 is expected to witness approximately 26 billion IoT devices that will be connected to the Internet [10], [11], [12]. These IoT devices generate a massive amount of data at 2.5 quintillion bytes of data every single day [10,12,13]. In the face of the rapidly increasing volume and complexity of data, the emergence of 5G and beyond networks has dramatically improved the ability of network data transmission, thereby ensuring that the data transmission capacity required by the application meets the physical requirements [2], [3], [4], [5]. According to references, 5G and 6G technologies are expected to provide 20Gbps peak data rate and 1ms round-trip latency for critical applications [2], [3], [4], [5]. On the one hand, a huge number of IoT devices monitor the environment and obtain sensing data [14]. On the other hand, some mobile IoT devices can act as Mobile Data Collectors (MDCs) because of their mobility and large computing and storage capacity [15,16]. MDCs can collect data generated by some static and communication-limited IoT devices to make applications obtain data more economically and expediently [17], [18], [19]. There has been a lot of research about some common MDCs such as mobile vehicles and UAV which act as data mules to collect data [1,11,19,20]. In a crowdsource network, the phone can also work as the MDC to implement large-scale data collection. Due to the MDC plays a crucial role in data collection [21,22], and the collected data is used to construct various applications, so that the reliability of MDC is essential for data collection [1,3,11,13,16,21,23,24].

IoT combined with the current 5G and beyond networks has caused major changes in the current network structure. Lots of computing resources are deployed at the edge of the network [24,[26], [27], [28]]. The processing and analysis of data can be performed at the edge of the network, reducing the energy consumption, bandwidth occupation, delay of uploading data to the cloud, and achieving better QoE for users [[29], [30], [31], [33]]. A large amount of data obtained enables the types of structured applications to increase rapidly. At the same time, the real-time, frequency, breadth and granularity of the data obtained by the application have been more abundant, which has greatly expanded the function and scope provided by the application. The above changes have promoted the development of current data-based applications rapidly. Among these applications, one of them is ultra-reliable applications which are deployed for critical infrastructure such as smart industry, agriculture, smart grid, autonomous vehicles [29,31], e-healthcare, smart education, etc. The aim is to provide QoS and QoE to end users with high reliability [34]. False, low-quality and even malicious data has a tremendous impact on this type of ultra-reliable applications since the application is based on collected data. Therefore, the key to building ultra-reliable applications is to ensure the reliability of the acquired data [1,3,11,[32], [35]], only trustworthy data can construct ultra-reliable applications. The untrustworthy data will lead to incorrect decisions of the application.

For the data acquisition mode of applications [21,22], the application publishes data collection tasks through agents or the cloud. In order to obtain trustworthy data, we should select IoT devices with high trustfulness to collect data. Based on the existing research [52], the network can use the crowdsourcing mechanism to perform evaluation tasks for IoT devices by recruiting the trustworthy MEUs as the third-party evaluators, and give MEUs certain rewards to incentive them to carry out tasks. MEUs act as mobile data collectors, during the execution of tasks, there may be some untrustworthy MEUs to provide some false even malicious data to get rewards, resulting in the system cannot choose trustworthy IoT devices to collect data, which damages the application reliability substantially. Therefore, an effective trust evaluation mechanism is needed to evaluate MEU trustfulness for supporting ultra-reliable applications. Trustworthy MEUs are selected to perform evaluation tasks, untrustworthy IoT devices will be excluded, thus ensuring that the data collected by selected devices are reliable [21,23,24]. The main challenge issues are as follows:

(1) Compared with new open and distributed IoT networks, traditional trust evaluation mechanisms have difficulties in obtaining trust relationships and evaluating accurately. The traditional trust acquisition method is passive [13,16], by observing the interaction behavior between the evaluated objects and evaluating whether the interaction behavior meets expectations. However, in the current IoT, the number of IoT devices is enormous [25]. Mobile IoT devices can become MDCs [21], each MDC whether submits data to the system or not is determined by itself. On the one hand, the trust evaluation system is restricted by observation methods, cost, and privacy, so it cannot obtain the interactive behavior of MDCs [37,38]. On the other hand, even if the interactive behaviors of MDCs are known, the trust evaluation system cannot give an effective trust evaluation result because it cannot evaluate whether the data submitted by MDCs is authentic and trustworthy [1,2,16].

(2) There are problems in traditional trust evaluation that the system is easy to be deceived and the data is difficult to be verified, which makes it challenging to support ultra-reliable applications. Although researchers have proposed some trust reasoning mechanisms, such as direct trust evaluation, indirect trust evaluation. The direct and indirect trust evaluation fusion is performed to obtain a comprehensive trust evaluation result. Essentially, in the traditional trust reasoning and evolution mechanism, whether it is direct or indirect trust evaluation, the evaluation information is derived from the evaluation results submitted by the evaluation object itself. Obviously, it is difficult to guarantee the accuracy of malicious evaluated objects [39]. Even if some evaluation systems adopt a method that comprehensively considers the evaluation results submitted by the interacting parties, it is easy to be deceived by the joint collusion of the interacting parties. Besides, for traditional evaluation, since the evaluation data comes from the evaluated object itself, it is impossible to test whether the evaluation results accord with the actual situation. The fundamental reason for these deficiencies is that the source information for trust evaluation is only from the evaluation object's feedback without the third party's reliable evaluation information source, which makes it difficult to verify the evaluation results and effects. The effectiveness cannot be guaranteed, and it is difficult to support ultra-reliable applications [39]. At the same time, some smart malicious objects can actively evade or reduce malicious behavior at the appropriate time after being aware of the trust evaluation mechanism and method, and disguise trustworthy behavior to maintain their trust value at a high level. Although the existing security algorithms [40] can suppress their attacks to a certain extent, it protects malicious attackers from being detected, which will continue to cause damage to the system.

(3) The traditional trust evaluation mechanism has the problem of long timeliness required for evaluation, which is difficult to meet the requirement of ultra-reliable applications. The traditional trust evaluation mechanism needs to observe the interactive behavior of the evaluated object to obtain the evaluation result, and the system cannot determine the time required for evaluation. For example, if the evaluated object has no interactive behaviors for some time, the system cannot evaluate its trust level for a long time. Therefore, it is necessary to find a new method to actively conduct trust evaluation, rather than passively relying on the interactive behavior of the evaluated object.

To tack such challenge issues, a novel verifiable trust evaluation joint UAV (VTE-UAV) mechanism is proposed to choose reliable MEUs to conduct the task for ultra-reliable applications in 5G and beyond networks. The VTE-UAV mechanism mainly adopts two types of information of non-evaluated objects for trust evaluation to avoid the deficiency of passive trust evaluation in the past. Such as the inaccuracy of trust evaluation, easy to be deceived and the inability to perform trust verification. The main innovations of this paper are as follows:

• A trust evaluation and reasoning mechanism based on aggregation is proposed. The mechanism no longer uses the information fed back by the evaluated object for trust evaluation, but adopts the information reported by the third party, thereby improving the objectivity and accuracy of trust evaluation. In the IoT, most MEUs are reliable (this is in line with the conclusions of most studies [39], if the malicious object accounts for the majority, the normal operation of the network cannot be guaranteed [39]). Therefore, by aggregating the data submitted by the MEUs, those data located in the aggregation center can be considered to represent the real value (that is, representing the value submitted by most MEUs, because most MEUs are reliable). Therefore, these data are used as baseline data. If the results reported by a certain MEU are close to the baseline data, it means that the data submitted by this MEU is authentic, and the trust of this MEU is improved. The MEU is likely to be malicious if the difference between the data submitted by this MEU and the baseline data exceeds a certain threshold. And the trust of this type of MEU will be reduced because of reporting false data, which can effectively identify trustworthy MEUs.

• An active trust evaluation mechanism is used to verify the reliability of MEUs by using the data obtained by UAVs as baseline data. To prevent MEUs from joint deception, we actively dispatch UAVs to collect some key data. Because the UAV is dispatched by the system, it is considered reliable. Correspondingly, the data collected by UAV is trustworthy, so it can be treated as trustworthy baseline data. Then the data collected by MEUs can be compared with the baseline data, and the reliability of MEUs can be evaluated, which makes the trust evaluation mechanism for MEUs become more robust. On this basis, a set of trust reasoning methods are established. After the verification of baseline data, the data collected by those MEUs with high trustfulness can be added to the baseline, so that more MEUs can be tested for reliability. In this way, the scope of trustworthy evaluation can be expanded through the trust evaluation method.

• A trust-aware greedy selection algorithm is proposed, which selects MEUs with high trustfulness to perform evaluation tasks in a low-cost manner. It can improve the accuracy of trust evaluation and the reliability of data collection for applications while reducing the cost. MEUs with high trustfulness are preferred as candidates, then the MEU with the maximum performance-price ratio will be selected from the candidate set by the cloud. Experimental results show that the trust-aware greedy selection algorithm can reduce the cost of cloud recruitment while maximizing cloud recruitment performance. Finally, through the systematic experiment of the VTE-UAV mechanism proposed in this paper, the results show that VTE-UAV can improve the cloud recruitment performance and increase the accuracy of trust evaluation of IoT devices.

The rest of the paper is organized as follows: Section 2 reviews related works. Section 3 presents the network model and problem statements of this paper. Then, the VTE-UAV mechanism is introduced in Section 4. Section 5 proposes experimental results for the VTE-UAV mechanism. Finally, we conclude this paper in Section 6.