Trust Management System in Internet of Things: A Survey

Secured communication channel establishment based trust score of sensor nodes is considered as important parameter while designing secured routing solution. Several cryptographic based approaches proposed to ensure security and privacy during message transmission in different environments discussed in Rathee et al. [3]. These approaches cannot be applicable to resource constrained environment. The cryptographic solution increases storage area, communication and computation overhead. However, due to unbound connection and lack of specific method for network monitoring, attacks impose severe consequences in Internet. Further delays in transmission result from these challenges.

Trust management system computes trust value of particular node based on current or previous interactions. Trust computation techniques improve security without degrading network performance. Trust based approach for IoT system must also deals with scalability and heterogeneity. To date, limited work on trust management system for security improvement is proposed. It is particularly important to identify misbehaving nodes that deliver services to other devices. Using IoT systems, the paper discusses trust management schemes that are at the forefront of innovation.

Rathee et al. [4] proposed trust computation framework for fog environment. It calculates trust score for each device and identifies malicious devices in network. The framework considered the fog devices and IoT devices which will be converted into malicious devices during handoff. A secured routed mechanism is proposed to make system attack resistance. Trust system created between fog and IoT layers. All fog nodes within the table are recorded and malicious nodes are detected. Trust calculation model for IoT system is presented in the study [5]. The author provided survey of existing trust computation techniques and further trust calculation method classification is shown. The article also highlighted with challenges and research ideas in Trust management systems.

Guo and Chen [6] present a trust based algorithm for model driven Internet of Things systems. This approach classifies trust model based on trust dimensions. Author summarizes pros and cons of each trust dimensions and highlights defence scheme against malicious nodes. Bahutair et al. [7] proposed trust management framework for potential producer and consumer. It calculates trust for service provider. It identifies trustworthy and untrustworthy IoT services. For accuracy and training time, the effectiveness of the approach is tested on a real-world dataset.

Caminha et al. [8] proposed trust management using machine learning. Author presented elastic sliding window technique for accessing IoT trust by evaluating service attributes. This approach helps in identifying malfunctioned nodes among misbehaving nodes. This method also worked for on off attacks and faulty nodes.

Energy efficient trust management system for resource constrained application is proposed by Khan [9]. This approach is used to detect malicious node behavior in IoT system. Author described three algorithms No listening for data forwarding, Listen own data forwarding and Listen to all transmissions for energy efficient IoT network. Energy efficiency is possible by measuring active listening time of IoT nodes. Zhang et al. [10] describe the issues in emerging information technology such as Internet of Things (IoT). Several issues associated with trust, security, and privacy in IoT are outlined in this article. The authors [11, 12] discussed methods developed by researchers to find the trusted device in IoT. Privacy, safety and privacy implementation challenges in IoT system are discussed in the study by Lee et al. [13]. Trust characteristics are discussed in the studies by Grandison and Sloman [14] and Pranata et al. [15]. A survey on trust computation methods and algorithms in IoT system presented in Wee and Banister [16]. Quality of Service based trust metric is used to evaluate trust value in Nitti et al. [17].

Trust computation security model based on experiences and recommendation is discussed in Hellaoui et al. [18]. Clustered sensor network in machine to machine communication uses trust metric for identifying malicious devices in the study by Yan et al. [19]. The authors [20, 21] proposed trust computation method for analyzing node behavior in sensor network. Trust computation model is designed to monitor sensor node interaction in sensor network discussed in Ganeriwal and Srivastava [22]. Trust parameters in community based Internet of things are proposed in the studies [23-25]. Among the trust parameters, honesty, cooperation, and community interest serve as indicators of a trusted social network. Direct and indirect trust parameters are measured to determine the value of a node's trust. Authors designed ascertain trust in social using trust based scheme in Nitti et al. [26]. The author node trust is evaluated by aggregating trust from common friends and direct experiences.

The numerous research is done on the access control methods to provide safe communication on IoT environment.

In Kaur et al. [27], capability-based strategies are illustrated. It provides a conventional capability-based approach and examines issues with access control from the viewpoint of the user. Additionally, it is said to be more efficient than conventional approaches. Authors suggested conducting a study on IoT context awareness from the viewpoint of the commercial market presented in Perera et al. [28]. The basis for access authorization in Kuhn et al. [29] is Roles, which can only be granted permissions by an administrator. An admin or user cannot alter the privileges provided to a user, and the user may delegate some, all, or none of their role to another user.

The aforementioned rule does not seem to apply when a user has to temporarily delegate part of their powers or obligations to a trusted user. It demonstrates the need for a safe access control system in an IoT context in Ferraiolo et al. [30]. Theoretically, an approach based on capabilities refers to an unforgeable ticket to obtain resources listed on the capability discussed in Hernández-Ramos et al. [31]. The concept of an organisation, which is defined as a well-organized collection of operational units, serves as the foundation for the Access Control in Organisations system. In order to play out various roles, subjects were used. While a group of things is viewed as a view, a collection of activities can always be conceived of as an activity presentd in Kalam et al. [32]. The primary problem with this paradigm is its centralised structure, which results in a host of complications. It has problems with large-scale implementation, and the absence of a simple instrument or procedure makes it challenging to implement.

I. Trust Computation Model

To identify the node legitimaty, trust computation method is useful. The following trust parameters are considered for trust based device identification.

a. Direct Trust (DTS)

Reliable packet (RPA) and device availability (DAA) parameters are considered for asseing direct trust.

Node nⁱ direct trust as computed by node n^j at time t is reflected in the DTS. This trust is calculated based on successful packet transmission.

$\operatorname{DTS}(\mathrm{n}(\mathrm{t}))=\left(\alpha^1 \times R P A\left(\mathrm{n}^{\mathrm{i}, \mathrm{j}}\right)\right)+\left(\alpha^2 \times \operatorname{DAA}\left(\mathrm{n}^{\mathrm{i}, \mathrm{j}}\right)\right)$                    (1)

b. Indirect Trust (IDS)

Packet Delivery Ratio (PD), Energy Available (EI), and Nearest Node Neighbor (NN) are the parameters used to generate indirect trust ratings. The EI is used to lower a node's energy consumption, the PD is used to estimate a node's packet reception rate, and NN is used to represent the degree of neighbor sensor nodes.

Using a weighting method, the total trust score of node n is calculated:

$\mathrm{IDS}=\alpha^1 \times \mathrm{PD}^{\mathrm{n}}+\alpha^2 \times \mathrm{EI}^{\mathrm{n}}+\alpha^3 \times \mathrm{NN}^{\mathrm{n}}$                     (2)

The values for α¹, α², & α³ are chosen such that sum is 1. The final trust score ranges from 0 to 1. A device is considered to be legitimate if it has a high trust score value.

Access permissions are assigned based on the trust score. The fuzzy inference system uses both direct and indirect trust score for assigning rights to devices shown in Figure 1.

1.jpg

Figure 1. Fuzzy inference system for access control

The fuzzy model mainly consists of four blocks as follows. Fuzzification: The two trust scores are fuzzified and divided into three linguistic variables depending on their values: low, average, and high.

• Base of Rules: Base of rules gives the node reliability based on the rules formed having input values, direct and indirect trust score.
• Inference System: Using the rules from rule base, fuzzy decision has to be taken.
• Defuzzification: This stage takes node reliability (NR) and performs defuzzification to determine the final prediction of node or device.

$F L(n)=\left\{\begin{array}{cc}1, & N R(n)==^{\prime} \text { highly trusted }^{\prime} \\ 2 & N R(n)==^{\prime} \text { trusted }^{\prime} \\ 3 & N R(n)==^{\prime} \text { rather trusted }^{\prime}\\else & 4\end{array}\right.$                   (3)

• Access Rights:

Access rights are given to the devices based on the type of node reliability.

The suggested method ensures the secured path through which only legitimate devices will communicate with their access permission.

4.1 Publication trends

The article with Internet of Things security and trust computation technique keywords were searched from the year 2014.

It shows improvement in number of publications till year 2020. In year 2020, significant research has done in trust computation techniques (Figure 2).

4.2 Document type analysis

Document type analysis shows documents of type article, review and conference papers.

2.png

Figure 2. Year-wise publication trends

Source: http://www.scopus.com (accessed on 05-08-2021)

3.jpg

Figure 3. Documents by type analysis in Pie chart representation

Table 1. Year wise publication count

There are 79.2% of publications published by Scopus that are article-type documents. Article type is followed by conference papers with 23.8% and review with 3.0%. The document type distribution has shown by pie chart in Figure 3 and year wise count is shown in Table 1.

4.3 Analysis by subject area

Documents with keywords Internet of Things security using trust computation technique are not only found in engineering and computer science but also in material science, decision science, mathematics, physics, chemistry, and chemical engineering. Figure 4 shows Document analysis by subject areas in pie chart representation.

4.jpg

Figure 4. Documents by subject area in Pie chart representation

5.jpg

Figure 5. Publications by country analysis

4.4 Geographical analysis

Figure 5 presents details of countries involved for trust computation research. It shows China carried out detailed research on IoT security using trust computation. The research in Internet of Things security is carried by various countries all over world. The top 10 countries research in same area is considered for analysis.

4.5 Document analysis by affiliations

This type of analysis shows research work carried out in Internet of Things security by different organizations. Figure 6 shows document analysis by affiliations using Scopus database. Top 15 organizations contributed to research are shown along with count of documents. King Saud University has done most of the work in this research area.

6.jpg

Figure 6. Document analysis by affiliations

4.6 Source type analysis

In this section, Figure 7 presented analysis of number of research article published in renowned journal. It is observed that maximum numbers of articles are published in IEEE Access followed by Computers Materials and Continua, Sensors, Wireless Communications and Mobile Computing and Advances in Intelligent Systems and Computing. Table gives documents per year by source.

7.png

Figure 7. Document per year by source

4.7 Analysis by funding agencies

Table 2 shows research analysis done by funding agencies. According to National Natural Science Foundation of China, the maximum funds have been allocated for IoT security research.

European commission, horizon 2020 Framework programme and many more funding sponsors as shown in Figure 8 have provided funds to carry out research.

Table 2. Per year document by source (Scopus database accessed on 05-08-2021)

8.jpg

Figure 8. Documents by funding agencies

4.8 Analysis on number of documents by author

Figure 9 below shows top 15 authors’ research work. Kurdi, H. has maximum number of publications in this research area.

9.png

Figure 9. Number of documents by author