Adaptive Trust-Based Framework for Securing and Reducing Cost in Low-Cost 6LoWPAN Wireless Sensor Networks

Abstract

Wireless Sensor Networks (WSNs) are the core of the Internet of Things (IoT) technology, as they will be used in various applications in the near future. The issue of security and power consumption is still one of the most important challenges facing this type of network. 6LoWPAN protocol was developed to meet these challenges in networks with limited power and resources. The 6LoWPAN uses a hierarchical topology and the traditional method of encryption and key management, keeping power consumption levels high. Therefore, in this paper, a technique has been developed that helps in balancing security and energy consumption by exploiting the Trust technique between low-cost WSN nodes called Trust-Cluster Head (Trust-CH). Trust between nodes is built by monitoring the behavior of packet transmission, the number of repetitions and the level of security. The Trust-CH model provides a dynamic multi-level encryption system that depends on the level of Trust between WSN nodes. It also proposes a dynamic clustering system based on the absolute-trust level in the mobile node environment to minimize power consumption. Along with a set of performance metrics (i.e., power consumption and network lifetime), the Cooja simulator was used to evaluate the Trust-CH model. The results were compared to a static symmetric encryption model together with various models from previous studies. It has been proven that the proposed model increases the network lifetime by 40% compared to previous studies, as well as saves as much as 28% power consumption in the case of using a static encryption model. While maintaining the proposed model’s resistance to many malicious attacks on the network.

1. Introduction

The Wireless Sensor Network (WSN) is considered as the infrastructure of the Internet of Things (IoT) technology [1], as it forms the basic core of data collection from different environments and processing within the Internet servers [2]. As wireless sensor networks are involved in various fields and industries, including military, civilian, surveillance, household and even inside the human body [3] known as Body Sensor Networks (BSN) [4]. Therefore, interest in this type of network has increased in the recent decade due to its economic feasibility for various applications [5].

ZigBee and 6LoWPAN are among the most popular protocols for managing this type of network [6]. They monitor the collection of data and then send it to the Access Point (AP) via the use of neighboring WSN nodes. The Internet Engineering Task Force (IETF) organization defined the protocols to transmit packets over IEEE 802.15.4 in a 2.4 GHz band [7]. The difference between them relates to addressing the nodes using IPv6, as in 6LoWPAN. It is also related to the topological type, where the ZigBee uses the mesh and the other uses the hierarchical. Moreover, the 6LoWPAN uses Low power and Lossy Networks (RPL) in the process of routing data to and from the AP and plots the topology. Both use standard Advanced Encryption Standard (AES) encryption [8]. Whereas, although AES is subject to a global standard, it is considered unsuitable for this type of network and its embedded devices [9]. It consumes more energy and power in encrypting and decrypting. The development and improved performance of these networks and sensors at present have an impact on the prices of these sensors. Inexpensive cost sensors are used in the majority of applications, including monitoring, humidity, rain, air, temperature, pedestrian monitoring, lighting, etc., due to their low cost and functional capabilities [10]. Therefore, several authors suggested modern encryption algorithms that are robust and reliable with lightweight in implementation, such as Speck128 [11], SIMON [11], Tiny Encryption Algorithm (TEA) [12] and FlexenTech [13]. Some of these lightweight encryption algorithms were analyzed in [14] to show that AES requires a high rate of power consumption. Furthermore, other works have proposed lightweight authentication algorithms for WSN [15] and others have used watermarking technology for the authentication process also [16] to reduce security costs and extend the WSN lifetime.

Hence, due to the increasing demand for this type of device (low-cost WSN node) included in future applications, it requires these devices to run for longer periods. It also requires the provision of minimal data protection at the stage of data collection or the stage of its transmission (data) to other WSN nodes until it reaches the AP. Therefore, given the importance of these two parts (power consumption and security) for WSN nodes, some mechanisms contribute to their improvement in parallel. One such method is to use the concept of clustering, in which a group of nearby WSN nodes selects the main WSN node called Cluster Head (CH) to use it as a means of sending data to the AP. This mechanism contributes to improving communication between nodes by exploiting short distances and a good signal for transmission. As the weak signal in communications contributes to increased power consumption [17]. Moreover, there are several factors that the studies have relied on to improve this performance (Clustering) to improve the efficiency and increase the Quality of Service (QoS) of WSN networks. Among these factors are the distances between the nodes, the signal strength, the nodes’ power level and more [17,18,19,20,21]. Concerning WSN mobile nodes, some studies have been conducted to improve the clustering method and raise its efficiency [22,23,24,25]. In addition, others have automated the process of clustering and the process of security complexity based on the state of the nodes [26]. Others use the principle of trust between nodes to facilitate communication processes and reduce costs while maintaining the security of the network [27]. Building trust is based on filtering each node to its neighbors using a combination of factors such as monitoring the messages sent and received, the node’s behavior and others [28]. Some studies have looked at improving trust by suggesting a lightweight trusted algorithm [29]. Others have been interested in using a complex set of parameters for security assurance using a trust model [30].

However, most of these studies discuss the partial issues related to optimizing one part (such as security and trust) and neglect its negatives over the rest of the parts (energy and CPU). Therefore, in this paper, we will discuss the issue of adapting the trust term to the situation in which the WSN node is to ensure security and maintain the network lifetime as long as possible. Our proposal is called “Trusted-CH” which initially builds trust between nodes based on monitoring the packet loss rate and acknowledgment messages generated at each connection establishment. After Trust is built, the choice of CH depends on several factors, including full trust. Then the nodes that contain the absolute trust communicate with a single master key and a low level of encryption. Conversely, CH communicates with the rest of the nodes that have not reached the degree of absolute trust using a lightweight security algorithm proposed in [26], without using its encryption automation. Trust contributes to ensuring the security of WSN nodes and reducing cryptographic complexity, which in turn extends the lifetime of the network while maintaining security. It is also important to note that the proposed algorithm achieves data integrity, data freshness and authentication.

Here is a summary of the main paper’s contributions:

• Propose an automatic security model for wireless sensor networks based on the level of trust between nodes. Building this trust depends on monitoring communications acts in a WSN environment. However, the main objective of this proposal is to provide security for the wireless sensor network as well as reduce its cost as much as possible.
• A new method for CH selection based on trust and other factors has been proposed.
• Suggest an adaptive encryption method based on the trust ratio of neighbouring nodes.
• The results were analysed based on the lifetime of the network and compared with existing algorithms.

The remainder of the paper is organized as follows; Section 2 presents an overview of the related work. Section 3 describes the proposed scheme in Trust-building, clustering and adaptive encryption techniques. A discussion of the performance evaluation, outcomes and security analysis is declared in Section 4. Finally, Section 5 includes the conclusion of the research paper.

2. Background and Related Works

In the concept of balancing power consumption and improving data security in wireless sensor networks, many studies have improved these two parts separately. Few studies have considered these two parts equally in terms of finding a technique that combines them [26,29]. One of these methods relates to the concept of Trust between WSN nodes. Therefore, to clarify these concepts, we first need to explain the WSNs’ infrastructure.

6LoWPAN is one of the most common protocols used to manage wireless sensor networks. The WSN nodes are distributed in the form of an Ad-hoc network and it relies on IPv6 in the process of adding IPs. In addition to adopting the hierarchical model in building the network topology. Moreover, RPL is used in the process of routing data to and from the Access Point (AP) [7]. AES is also used in the encryption process between data [8]. Figure 1 shows a standard 6LoWPAN architecture.

Therefore, in the 6LoWPAN standard, the protocol operates in different working environments, some of which are dangerous or miniature places inside the human body. Consequently, maintaining the lifetime of the network along with providing security is one of the necessary things in these networks. In this section, we review some studies that have improved the work of the wireless sensor network by saving energy and supporting data security.

The clustering method is one of the most common methods that save power to WSN nodes by maintaining the distance and signal strength between adjacent nodes when connected [31]. Processing the WSN nodes when the distance between them is as wide as possible leads to increased power consumption especially in mobile node environments [32]. In [19], the authors split the network into a set of layers until it ends in the root, which is the AP. Each layer is divided into clusters that are identical in structure. These groups share data using two heads (CHs) to balance the energy of each group. Moreover, in [20], the authors used various factors such as the node’s power level, the node’s motion angle and the distance between the node and the AP to choose the CH node. Continuously in the process of optimizing the process of clustering method, the [33] has implemented an improved routing protocol to determine the ideal CH cluster and the actual collaborative data. Their proposal always selects the CH node with the highest energy parameter without consideration for other important matters. In [26], the authors proposed a new clustering method for 6LoWPAN mobile networks based on various factors such as node power level, number of adjacent nodes in a particular area and distance between adjacent nodes to reduce nodes’ power consumption during communication. Next, the authors balance node security and power through an automated encryption level.

Regarding linking Trust models with clustering methods to develop secure and energy-efficient models. The authors in [29] proposed a lightweight trust algorithm (LTA). This algorithm was based on the mutual monitoring of nodes using several parameters such as data packet, control packet and event. The traditional aging factor was used to determine the trust score of each node. It gave good results in identifying some types of attacks, such as the black hole, but it is slow in responding to the attack. Moreover, authors in [34] calculated the reliability of the nodes (vehicles) using various parameters such as connection duration, security level and centrality. Sensors in vehicles adapt the level of security through the level of confidence in neighbouring cells and vehicles. However, the vehicle is central to these sensors (WSN nodes) and can perform complex operations on several security levels. In addition, work in [35] performed a machine learning method to calculate the trust level for each neighbouring sensor (vehicles). Furthermore, A quantified trust accessibility model was presented by [30]. According to the authors, the future of IoT is based on trust. To extract basic trust traits, the authors used a numerical technique. In addition, unattended machine learning methods were applied: k-meant clusters to categorize interactions as reliable and untrustworthy. The optimal size of the cluster was computed using the Elbow technique and the main component analysis was further employed to decrease the training matrix. Features have been standardized to provide sequence data in the range of 0 and 1. The radial function kernel is being used to reach the nonlinear limit that can differentiate reliable and unreliable events and learn the optimum strategy to integrate trust characteristics to acquire the last trust value.

In addition, the combination of WSN, different security and double locking systems was used and the Trusted and Secure Routing (TSRS) was proposed in [36]. TSRS has been able to ensure that the path takes place, combating some internal threats such as black hole attacks and selective redirection attacks based on the commitment trust. This technique may integrate force selection and give a secure path by applying the trust and cuckoo search algorithm to distinguish the trusted path. The system also guarantees the life of the network. Furthermore, authors in [37] developed an Energy Trust Model (ETM) to address the problems connected to relay nodes’ dependability and data sensing secrecy. Finding a trustworthy connection and secure data transmission have also been implemented by the trust-based Secure Directed Diffusion Routing (TSDDR) protocol. Furthermore, the proposed approach ensured the corresponding energy usage and data confidentiality. Likewise, the authors in [38] proposed a fuzzy-based hierarchical trust management algorithm. The nodes that are securely joining the cluster are established utilizing an intelligent and protected fuzzy clustering technique using a Balanced Weight Sub-cluster formation (BWS). The BWS procedure uses the scheme Fuzzy S-means for CH-identification. The CH is liable for trust-based pathways and the utilization of energy. The authors in [39] suggested a Trusted Moth Flame Optimization and Genetic Algorithm (TMFOGA) for CH selection. The algorithm depends on the density of a node, residual energies, packet forwarding progress, distance and transmission latency. However, with TMFOGA, there is a significant drawback that omits its suitability for hostile applications.

To make it easier to understand and clarify what we discussed earlier, we have created

Table 1. Review of the trusted-based methods.

Refs	Technique	Key Idea	Strengths	Weaknesses
[19]	Clustering method	Uses two heads in order to balance energy between nodes	Reducing the amount of intra-cluster communication	Increase complexity of routing filtering
[20]	Clustering method	CH is selected based on various factors	Extend the network lifetime	High complexity
[33]	Clustering method	CH is selected based on the highest residual energy	Reduce nodes power consumption	confidentiality and integrity are not supported
[26]	Automated encryption	Balance nodes security and power	Reduce power consumption Support different types of internal and external attacks	Priority data is not considered
[29]	lightweight trust algorithm	Monitoring WSN node’s parameters	Support different types of internal attacks	Slow response time
[34]	Adaptive trust-based	Adjust the security level based on the trust level of nearby vehicles	Improve the QoS for networks	Not suitable for ad-hoc networks
[35]	Centric trust framework	Using decision tree classification and artificial neural networks	Improve the accuracy of the trust model	Costly
[30]	Trust metrics framework	k-meant clusters to categorize interactions as reliable and untrustworthy	Improve the accuracy of the trust model	Costly
[36]	Trusted and Secure Routing	Cuckoo search algorithm	Provides the secure routing path Minimizes the probability of malicious nodes	Packet delivery ratio is lower
[37]	Energy Trust Model	establishing a trustworthy communication pathway	Balance power consumption between privacy and energy Support end-to-end communication	Higher computational complexity. Higher energy overhead
[38]	Hierarchical trust management	Balanced weight subcluster formation uses Fuzzy S-means to the CH-identification scheme	Cover a high variety of malicious attacks.	Complexity is high
[39]	Trust clustering model	Using GA to choose the appropriate CH	Improve the power consumption	Costly

which includes an analysis of all the techniques discussed above. It highlights technology, its main idea, strengths and weaknesses. Based on the studies presented previously, most focus on calculating confidence and the amount of improvement in one part without the other (security or energy). In addition, most did not mention the authentication methods used or even the key management processes. Therefore, in this paper, we will detail the trust calculation process, the CH selection process, authentication methods, key generation based on the trust level, the encryption algorithms used and the adjustment mechanism between trust and the level of encryption that will be used.

3. Proposed Trust-CH Framework

Trust-CH’s proposal consists of three main phases. The trust-building phase between WSN nodes, the CH selection phase and the security adaptation phase. In the trust-building phase, the CH nodes monitor the trust with the nodes associated with them and then pass the trust file between them (CH nodes). When a node reaches Absolute-Trust, it is nominated to be among the options for selecting the main nodes (CHs) with other factors. The main objective of rotating the clustering heads is to maintain a balanced power consumption and ensure the QoS. The Security adaptation phase also aims to reduce the cost of encoders for power, based on the amount of nodes trust. Figure 2 shows the overall diagram of the Trust-CH.

From Figure 2, the communication between nodes based on two security frameworks. The first is the proposed model (Trust-CH), through which the CHs communicate with each other and with nodes with a high Trust value. The other model is FlexCrypt [26], which we modified by fixing the number of Rounds and Block size for each encrypted packet.

Table 2. Notations and their description.

Notations	Description
xi	Node number i
$x^j$	Node number j
$t$	Time
$w$	The weight of the node
$\lambda$	Constant values
$b$	Block size
$P_{TX}$	The power consumed when the node sends the message
$q$	Message length
$d$	The distance
$P_{TX-amp}$	The node transmitter circuit consumption based on signal amplification
βfsm	Free space model
βtrm	Two-ray grounded propagation models
$P_{RX}$	The power consumed when the node receives the message
$P_{RX-elec}$	The power dissipated by the receiving node$\mathrm{in} \mathrm{the} \mathrm{reception} \mathrm{of} q$
$Th\left(n\right)$	The power threshold of a node
$N_{adj}$	The number of neighboring nodes
$P_{elec}$	Power expanded by transmitter and receiver
$\sigma$	The average of interactive factors
$n$	Total number of nodes
$Tn$	Candidate node
$\partial \left(Tn\right)$	Average distance
$\mathsf{\omega }$	The average weights
$a$	The number of adjacent nodes
$P_r$	Residual energy
$\partial$	The average distances between the candidate node (Trusted node) and the existing CH nodes
$c$	Minimum number of host nodes
$d_{Th}$	Distance between CH nodes and the host nodes
$m$	Total number of CHs operating at time (t)

shows the notations and their description used in the scheme.

In trust-based adoption operations, the trust level of a WSN node increases based on the definition of its neighbors. Therefore, in our proposal, a table (T) is used to store the trusted WSN nodes and their respective level. Furthermore, each WSN node shares its table between other WSN nodes in each update by joining, modifying, exiting any WSN node. The CH node updates its table, building the information of each WSN node associated with it and how often it is specified in the other CHs. If a WSN node is active and passes through more CHs or creates new CH groups, the level of trust increases and if its activity decreases, the level of trust decreases. Likewise, if it is missing from the update for extended periods, it will be deleted from the table. Figure 3 shows the composition for the table used.

3.1. Trust-Based Calculation

This trust-table is passed to all WSN nodes associated with the CH as well as to other CHs, which in turn pass it on to the rest of the WSN nodes. Updates to those tables are sent in encrypted form to the rest of the nodes, as will be explained later. In this process, we ensure that active and trusted WSN nodes are not subject to trust calculations more than once, which in turn contributes to energy savings. However, if any new WSN node joins the existing set of WSN nodes, then a computation for that WSN node is carried out using a set of criteria. These criteria are discussed in detail as follows:

• Packet of Transmission Behavior

When two WSN nodes are within one hop distance, the direct trust of $x^i$is calculated by $x^j$ at time t. Each cyclic trust evaluation round includes WSN node $x^j$observing WSN node $x^i$directly. Since aggressive users are primarily concerned with packet loss and fake acknowledgment, the packet of transmission behavior (TB) parameter allows for accurate estimation of the WSN node through x-rate estimation [40].

$$TB\left(x^i\left(t\right)\right)=\left(PF\left(x^{i,j}\right)\times w_1\right)+\left(AF\left(x^{i,j}\right)\times w_2\right)$$

(1)

where $PF\left(x^{i,j}\right)$ denotes the successful packet forwarding ratio $x^i$by $x^j$, AF($x^i$) denotes the availability factor of $x^i$by $x^j$, $x^j$, $w_1$ and $w_2$donate the weight parameters some of them is 1. Weight parameters represent the Trust-coefficient between the nodes. The PF is calculated as shown in Equation (2) [40].

$$PF\left(x^{i,j}\left(t\right)\right)=\frac{f\left(x^{i,j}\left(t\right)\right)-f\left(x^{i,j}\left(t-1\right)\right)}{f\left(x^{i,j}\left(t\right)\right)+f\left(x^{i,j}\left(t-1\right)\right)}$$

(2)

where, $f\left(x^{i,j}\left(t\right)\right)$ is the number of successful packets sent by $x^i$ at time 𝑡. This parameter effectively protects the network from cyber threats and identifies the malicious behavior of WSN nodes in the network by calculating $x^i$ ACK packets. For further explanation, suppose that node $x^i$ forwards data packets from another node and broadcasts 4 ACK packets. During this time, the node $x^j$ collects such ACK packets of node $x^i$to estimate the number of sensing packets. Calculated data transmission packet rate in time 𝑡 is illustrated in Equation (2). In addition, the AF is also used to identify malicious nodes by calculating the Hello messages sent. The “Hello” packet is transmitted by $x^j$ to check whether $x^i$is available to receive this packet. Once ACK received for Hello packet from $x^i$at $x^j$, it defines the availability of $x^i$. The AF is calculated in Equation (3) [40].

$$AF\left(x^{i,j}\left(t\right)\right)=\frac{A\left(x^{i,j}\left(t\right)\right)}{A\left(x^{i,j}\left(t\right)\right)+AH\left(x^{i,j}\left(t\right)\right)}$$

(3)

where, $A\left(x^{i,j}\left(t\right)\right)$denotes the number of acknowledged Hello packets and $AH\left(x^{i,j}\left(t\right)\right)$ denotes the number of non-acknowledged Hello packets.

However, a higher $TB\left(x^i\right)$value indicates greater chances of $x^i$as a legitimate WSN node. Based on the output of the $TB$, it is determined whether the nodule is malicious, suspicious, or benign. If the $TB$ value is within the range of the malicious node, the $x^i$node is rejected and the details of that node are reported to the rest of the CHs. Otherwise, the Trust Level (TL) value in Table (T) will be populated with the TB value only for the first time (The $TL$ value is selected based on the $TB$ value). The scale variable for the TL selection process is used to determine which one is in Equation (4).

$$TL=\{\begin{array}{c}Malicious, TB<\lambda \\ Suspicious, \lambda \le TB<{\lambda }_1 \\ Benign, TB>{\lambda }_1\end{array}$$

(4)

where $\lambda \mathrm{and} {\lambda }_1$ are constant values. The $TL$ value is selected based on $TB$ value.

2.

Number of Repetitions

If a new node is classified as benign or suspected, it remains under trial and measurement until trust is confirmed. Each time the WSN node goes to the link in another CH, it is also monitored and $TB$ is calculated. The new $TB$ output is added to the TL value in Table T. This operation will continue in more than one CH or the same CH. Even if the TL value reaches 1 for any of the WSN nodes, the WSN node obtains the absolute trust criterion in the cluster and processes it reliably without any future trust operations. Conversely, if the value of $TB$ decreases, it is subtracted from the TL and the node becomes suspect until it is eventually rejected. These operations appear in Equation (5).

$$TL\left[x^i\right]=\{\begin{array}{l}TL\left[x^i\left]-TB\right[x^i\left], TB\right[x^i\right]<\lambda \\ TL\left[x^i\left]+TB\right[x^i\left], \lambda \le TB\right[x^i\right]<{\lambda }_1 \\ TL\left[x^i\left]+TB\right[x^i\right], TB\left[x^i\right]>{\lambda }_1\end{array}$$

(5)

Likewise, it should be noted that WSN nodes with a trust value (TL) less than the threshold value ($\lambda$) will not exchange the T-table with them. AP is also aware of this table (T) because it will decrypt the encryptions based on the trust values of those WSN nodes. Algorithm 1 provides a summary of the Trust-building steps we discussed earlier for each WSN node.

Algorithm 1: The WSN trust-based calculation.

1.  for each i∊ $n$do // all nodes

2.   $x^i$ represent the normal node

3.   CH represents the Cluster-head node

4.   for each xi ∊ CH do

5.     search xi in CH.T // CH table

6.     Find P$x^i$F ()

7.     Find AF($x^i$)

8.     Calculate (TB($x^i$))

9.     Execute Equation (4)

10.    if T.xi.TL is malicious do

11.     Broadcast a warning message to all WSN trusted nodes

12.     Block the malicious node

13.     goto end // line 21

14.    elseif T.xi.TL <> Null do

15.     Execute Equation (5)

16.     Send a message to all trusted WSN nodes and CHs

17.     Do private authentication and encryption

18.     else

19.     T.xi.TL= $TB$

20.     Send a message to all trusted WSN nodes and CHs

21.     Do private authentication and encryption

22.    endif

23.   endfor

24.  endfor

3.

Security Level

Security and the Quality of Service (QoS) are known to operate in contradiction to each other. Furthermore, because we are dealing with WSN networks with limited capabilities in terms of CPU and energy. Therefore, finding a solution that balances achieving security and raising service quality is one of the most important contributions of this paper. Reference [26] will be used for authentication and encryption (FlexCrypt) between CH nodes and WSN nodes that are not fully trusted. In addition, if the WSN nodes have sufficient trust, the security factor will be lowered to conserve energy consumption. This is carried out through the use of a unified secret key for all WSN nodes that receive the ‘trust’ address and also by using less complex encryption of the data sent and received. Equation (6) shows the block sizes (b) in the encryption algorithm based on the TL value.

$$b=\{\begin{array}{c}8, TL\left[x^i\right]\ge 0.7 \\ 32, TL\left[x^i\right]<0.7\end{array}$$

(6)

The process of building trust begins gradually between WSN nodes, which means that the process of authentication and key generation is carried out using the FlexCrypt algorithm and after the WSN node reaches the full trust, it can then receive the T table, which contains only all the trusted nodes. This table also contains the master key shared between those WSN nodes. Next, the WSN node detaches from the FlexCrypt algorithm’s authentication system and starts using the table (T) information to exchange data. However, if one of those WSN nodes (Trusted) turns into the CH, it uses the two systems based on the FlexCrypt and the table (T) to handle both trusted and untrusted WSN nodes.

Furthermore, all the information about both techniques (e.g., keys, nonce) should be present in the AP. If the CH transmits data from the child’s CH to the parent’s CH, the incoming data is encapsulated using the trust technique. Untrusted nodes are initially encapsulated by the associated CH and then passed between the parent CHs using the trust technique. Therefore, no CH node can be fully trusted. Figure 4 shows a summary of the previous steps in the process of determining the security level and selecting the security algorithm.

3.2. Dynamic Selection of Cluster Heads

Based on what was explained at the beginning of this section, what the 6LoWPAN protocol does is connect WSN nodes in the network to each other and send data to an AP using the RPL routing protocol [8]. This protocol (RPL) divides the network topology on a hierarchical basis, where the AP is the top of the hierarchy and the level of the WSN nodes continues to decrease until it reaches the final nodes (maximum). All WSN nodes in Standard 6LoWPAN have the same tasks and priority, but they differ in the level of permission to transfer data. As WSN nodes near the AP allow data from all lower levels, as well as their own data, to be passed to the AP. Moreover, since we propose to improve QoS, save energy and maintain WSN security, we will use cluster technology between WSN nodes.

In addition, we will modify the RPL routing protocol to cover only interconnected CH nodes. Furthermore, the choice of CH will depend on several different factors, the most important of which is that the candidate WSN node must belong to the trusted nodes. Other factors are the residual power in the node, the number of adjacent nodes and the distance between the adjacent CH nodes. The proposed process of selecting the CH node is as follows:

1-

The radio frequency computing power (P) model for the proposed is based on [21]. As most of the energy consumption of the WSN nodes is caused by the correspondence between them. Moreover, distance affects how much power is consumed for both WSN nodes (transmitter and receiver). The relationship based on the power consumption of the transmitter and receiver with distance and message size is shown in Equation (7) [17].

$$P_{TX}\left(q,d\right)=P_{TX-elec}\left(q\right)\times P_{TX-amp}\left(q,d\right)$$

(7)

where $q$ is the message size, $d$ is the distance between transmitter and receiver, $P_{TX}$ is the total power consumption through sending $q$ over distance $\left(d\right)$, $P_{TX-elec}$ is the node transmitter circuit consumption based on $q$ size and $P_{TX-amp}$ is the node transmitter circuit consumption based on signal amplification. Moreover, the $P_{TX-amp}$ will be equal to either β_fsm for a free space model or to β_trm for the two-ray grounded propagation models depending on the $d$ between the transmitter and receiver. Moreover, the WSN node is responsible for receiving data messages from other nodes. Equation (8) [17] can be used to compute the power $P_{RX}$ needed to receive $q$ the message:

$$P_{RX}\left(q\right)=P_{RX-elec}\left(q\right)=P_{elec}\times q$$

(8)

where $P_{RX}$ is the power consumed when the node receives the message, $P_{elec}$ is the power expanded by transmitter and receiver in nJ/bit and $P_{RX-elec}$ is the power dissipated by the receiving node in the reception of $q$.

2-

The WSN node energy threshold $\left(Th\right)$ Since the CH node works centrally to the neighboring WSN nodes, the power consumption and computations will be higher, thus we must keep the node from collapsing by choosing the $Th$ value. If the node energy reaches that value, it will be excluded from running for the CH position. Moreover, if it is also in this position and the amount of energy falls below the $Th$ value, it is also replaced. This factor balances energy consumption between the nodes and maintains the lifetime of the entire network. The $\left(Th\right)$ for each WSN node is calculated in Equation (9) [17].

$$Th\left(n\right)=N_{adj}\times q\times P_{elec}+N_{adj}\times q\times \left[P_{elec}+{\beta }_{fsm}\right]\times d^2$$

(9)

where $Th\left(n\right)$ is the power threshold of a node, $N_{adj}$ is the number of neighbouring nodes and ${\beta }_{fsm}$ is a free space model. Furthermore, each node (N) sends the identifier (ID) and power level to all trusted nodes by data_message which is used to compute and update the table of trusted nodes in each session.

3-

The average of the set of interactive factors. The number of adjacent nodes $\left(\mathsf{\alpha }\right)$ for each trusted node. The greater the number of WSN nodes close to it (nominated trusted node), the greater the chance of it becoming a CH node. Moreover, the average distances between the candidate node (CH) and the adjacent node will be taken into account, as shown in Equation (10) [17].

$$\sigma \left(N\right)=\frac{{{\displaystyle \sum }}_{j=1, j\ne n}^{\mathsf{\alpha }}{d}_{\mathrm{n}⇢\mathrm{j}}}{a}$$

(10)

On the other hand, we also need to set the threshold ($d_{Th}$) value on the distance between CH nodes and the host nodes. Since increased distance and weak signal lead to increased unwanted power consumption. Thus, we will determine this value ($d_{Th}$) based on Equation (11) [17].

$$d_{Th}=\left(\sqrt{\frac{{\beta }_{fsm}}{{\beta }_{trm}}}\right)$$

(11)

In addition, we need to know the average distances between the candidate node (Trusted node) and the existing CH nodes to keep the RPL working to transfer data to the AP. Equation (12) will represent this process.

$$\partial \left(Tn\right)=\frac{{{\displaystyle \sum }}_{j=1, j\ne n}^m{d}_{n⇢j}}{m}$$

(12)

where $\partial \left(Tn\right)$ is the average distance, m is the total number of CHs operating at that time and $Tn$ is the candidate node. In this work, GPS will be used in real application to determine distances between nodes [19], but Cartesian coordinates (X, Y) will now be implemented for each node.

Finally, the residual energy level ($P_r$) will enter into the equation for choosing the best CH node. The node with higher $P_r$ value is nominated to be the CH. However, based on the average set of interaction factors that we proposed, the weight of each candidate node from the trusted nodes will be determined to be the CH. Equation (13) illustrates the average weights of these factors.

$$\mathsf{\omega }=\frac{a+2\times P_r+\sigma /10}{d_{Th}\times \partial }$$

(13)

We find that the values of the coefficient of multiplication between the interactive factors are not convergent, thus we take an approach by increasing the effect of the residual energy and decreasing the effect of distance by dividing $\sigma$ by 10. However, based on Equation (13), the selection of the optimal node (CH) depends on that it mediates the largest number of nodes, is close to the other CH nodes and has high residual energy. Furthermore, do not forget that it has absolute confidence (full trust).

Since the proposal depends on the scenario of the moving nodes, it is possible for nodes to move away from their associated node (CH). Thus, when the threshold value ($Th$) passes between it and its (CH), the WSN node either searches for the nearest CH or forms a new cluster based on its adjacent nodes. This process is carried out in the same way as reference [26]. However, keep in mind that the new CH has a trust feature.

Figure 5 explains the mechanism of the CH node creation and rotation between mobile nodes. The algorithm begins by tracking the distance between the WSN node and its CH, as the nodes of this type of network move at a slow speed. Assuming that the nodes are moving at a speed of about 5 m/s and the distance between the node and its CH is checked once every second (s) [41].

Once the distance between a node and its CH reaches the assumed limit ($d_{Th}$), the node searches for any adjacent CH nodes. If a node can find a variety of neighboring CH nodes, it will link to the nearest one and start the correspondence process between the node, the CH node and the target-CH node to handoff to the target-CH node. The Join_Request and Release_Request messages are used for these purposes. Next, the table (ns) of the CH parent nodes is modified, which is owned by each CH individually and called (ns). This table (ns) has all IDs of the associated nodes. The identifier is deleted from the CH parent node and added to the table of the target-CH node. If the node is a CH and it has moved far enough from the host nodes (children) and the distance (d) is closed to the $d_{Th}$, CH will also detach from its host nodes and return to the “normal node” in search of a link to the nearest CH. In addition, if a node’s trust rate is not enough to give it the ability to create a cluster, it sends a message to the nearest trusted node, which in turn (the trusted node) breaks off from its CH and creates a new cluster. As a result, a node that does not have an absolute trust will join the new cluster along with its closest nodes. However, when trusted nodes create a cluster, they compute the value of ω for each trusted node and share its results with other trusted nodes, then the most suitable node is chosen based on the factors we discussed earlier. Finally, the created CH node sends an announcement message (CH_Announce) to the rest of the nodes that a new CH node has been created to link with the other CH nodes and activate the RPL protocol.

4. Implementation and Evaluation

A discussion of the working environment in which the proposed algorithm will be implemented is presented in this section. Then, we will discuss the effect of this algorithm on power consumption, network life and packet loss. Furthermore, the proposed algorithm will be compared with other trust algorithms and automation lightweight cryptography

It should be noted that the proposed algorithm (Trust-CH) uses the reference method ([26]) in the process of key generation and authentication if the Trust level in the WSN node does not reach 0.9, if it reaches that point, it will use a unified key between the trusted nodes.

4.1. Simulation and Performance Metrics

In this work, a Cooja simulator [42] will be used due to its support for the 6LoWPAN protocol. Moreover, the Cooja works within an operating system called the Contiki [43], designed to control the Internet of Things. Both systems are running on the Oracle VirtualBox in a machine with a Core i5 processor clocked at 1.8 GHz, 8 GB cache and 12 GB RAM.

Table 3. Simulation parameters used.

Parameter	Value
WSN size	60 m × 120 m
AP location	X = 30, Y = 90
n	200
Node speed	5 meter/second
q size	64 bytes
Control message size	25 bytes
P initial	0.5 J (joule)
βtrm	0.0013 PJ/bit/m4
βfsm	10 PJ/bit/m2
Pelec	50 nJ/bit
Pr	0.005 J
Pb	0.005 for 16 bits
dT	87 m

shows the parameters that will be calibrated based on simulation requirements, part of which was taken from Darabkh et al. [17].

The simulation system randomly distributes the nodes in the dimensions of 1000 by 1000 of the terrain connected to nine CHs in the initial values as illustrated in Figure 6, then they start to move randomly at a rate of 5 m per second [44]. In our scenario the UDP clients (normal nodes) send Constant Bit Rate (CBR) traffic of around 64 bytes with the periodicity of 0.01 s.

Related to performance metrics that will be used in this simulation, the network lifetime, node power consumption and packet loss. Each one of these metrics is explained as below:

• The energy consumed and the remaining energy of the WSN nodes: It depends on the energy difference between different times to know the degree of impact of the proposed algorithm on the energy consumed.
• Network lifetime: The life of the network is based on a measurement of the time period for the first node to reach zero in its amount of remaining power.
• Packet loss: This metric measures the rate at which packets fail to reach their destination when traveling over the network. Packet loss has several factors for its occurrence, but this proposal will refer to link the effect of attacks on its increase and energy consumption.

4.2. Experimental Results

In the configuration of the proposed Trust-Based algorithm, Speak-128 [11] will be used for encryption. Details of this algorithm and its impact on WSNs are explained in [14,45]. Details of the effect of block size and the number of rounds on encryption complexity, as well as the amount of power consumed, are also described in [26]. Furthermore, the scenario used in the simulation does not have the effect of sleep mode on the sensors CPU, as they are all in working mode for the duration of the execution. The relationship between security and trust to be used is shown in Figure 7. The Figure shows the relationship between increased trust between WSN nodes and reduced security complexity in generating different keys and the effect of reducing block size on Speak-128 algorithm. If the Trust-level reaches 50%, the security complexity will decrease by 10%. When Trust-level reaches 70%, security will decrease by 30%. When Trust-level reaches 90%, the security will drop to 30%. Finally, if the Trust-level reaches one value, the security complexity is reduced to 20%. As we also mentioned, the authentication process will use only one encrypted key management.

In order to measure the effect of block size automation in cryptography on the proposed method and its effect on reducing power consumption, Figure 8 and Figure 9 illustrate this relationship. In the first scenario, we will examine the power consumption rate of the proposed algorithm (Trust-CH) with the effect of the encryption algorithm (Speck-128) if its encryption is used fully or dynamically. In this scenario, the amount of energy at each node is set at 1 joule and the number of nodes is set at 200 in all scenarios. Execution times were also chosen between 10 s and 40 s to show the effect of increased Trust-level on WSN node power consumption. Figure 8 illustrates this scenario.

From Figure 8, we notice that at the beginning of the execution, the Trust level is close to 0, so in 10 s we notice a convergence of power consumption between full (static) and dynamic encryption. Furthermore, because the encryption process depends on the second type (cyprot1) for the most part. After that, the Trust-level increases and the complexity of the encryption decreases, so the differences between static and dynamic encryption increase, reaching a maximum of 40 s, where the difference is 28%. It is also known that with the increase of time the Trust-level increases to reach in most of them (WSN nodes) 100% and this means more savings in the rate of depreciation.

Moreover, Figure 9 presents an analysis of the Trust-CH network lifetime. Increasing the initial power of the network nodes in a time period of about 100 s increases the level of trust between the nodes, resulting in greater energy savings compared to static Trust-CH encryption.

In defense operations against malicious attacks, the proposal used the technology of symmetric encryption and symmetric authentication (FlexCrypt) in the event of new WSN nodes join to the CH. Eavesdropping, replay and brute force attacks are subject to protection from these attacks. Other attacks such as black holes are also monitored by monitoring incoming and outgoing messages to and from the CH. Therefore, the suspicion of the WSN node results in a decrease in Trust-level and an increase in the rate of power consumption, which is also the result of the increase in the re-transmission of the packets between the two communication ends (nodes). Figure 10 shows the effect of different percentages of attacks on the network, as the increase in attacks results in a decrease in the life of the network.

Figure 10 shows the percentages of WSN nodes under attack between 5% and up to 30% of the total WSN nodes. Trust-CH network lifetime decreases by 12% when malicious WSN nodes increase from 5% to 30%. Furthermore, to compare the Trust-CH algorithm with existing works, most of the previous work revolved around Trust regarding CH selection without using any of the cryptographic algorithms. Thus, on the security side, we are going to compare the Trust-CH to the FlexCrypt algorithm [26]. One of the weaknesses of the FlexCrypt algorithm is its absolute dependence on the state of the WSN node without paying attention to the importance of the data transmitted by this node. Therefore, the concept of trust provides a typical solution for data security without paying attention to the state of the WSN node while extending the lifetime of the network.

Furthermore, to compare the Turst-CH algorithm with the current works, we chose the Flex-Crypt and TMFOGA algorithms. Figure 11 and Figure 12 show these comparisons in two separate figures due to the technical difference in both algorithms. Each of them has its own settings. Figure 11 shows a comparison of Trust-CH and FlexCrypt algorithms for a variety of WSN node initial power. Since the FlexCrypt algorithm is also used in the Trust-CH algorithm when the trust level of each node is less than 50%. The network lifetime level is close between the two algorithms when the initial power value of the nodes is low. As the initial power level of WSN nodes increases, the level of Trust-CH increases and the impact becomes larger, thus widening the difference in network lifetime. This is due to the use of Speck-128 in Trust-CH, which is more light and less complex compared to FlexenTech [13] used in FlexCrypt. Moreover, the dependence of CH filtering on the level of Absolute-Trust gave lower CH candidacy compared to the total number of nodes. When the WSN nodes’ initial power is 1.5 J, the network lifetime difference between the Trust-CH and the FlexCrypt algorithms is 40%, in favor of the Trust-CH algorithm.

As for comparing the Trust-CH algorithm with the existing works using Trust-model in selecting the appropriate CH, we have canceled the encryption part of the algorithm, to remove the energy consumed in those operations. Next, we compared it with the TMFOGA algorithm [39]. We also adjusted the simulation time by increasing it to 400 s to allow the nodes to reach zero energy. Moreover, we reduced the initial energy of the WSN nodes. Since after deleting the encryption processes, the power consumption becomes much less and therefore we need more simulation time.

Figure 12 shows a comparison between the Trust-CH and the TMFOGA for a different set of WSN nodes’ initial power. It can be seen from Figure 12 that increasing the initial power of the nodes increases the lifetime of the network systematically since trust is adopted as a major part of the CH selection factors. Moreover, we also note that TMFOGA increases network lifetime in a semi-regular manner as well. The difference in network lifetime between the two algorithms increases the initial power of the WSN nodes increases. On one hand, the Trust-CH algorithm is quick to choose the appropriate CH and on the other hand, the complexity of CH selection in the TMFOGA remains constant. TMFOGA uses a Genetic Algorithm and a variety of factors for each CH selection. Therefore, the Genetic Algorithm needs a data set to investigate and the TMFOGA algorithm naturally consumes more time and energy, while the Trust-CH algorithm is easier and faster to choose CH. The difference in network lifetime between the two algorithms (Trust-CH and TMFOGA) when the WSN nodes’ initial power is 1 J is 42% in favor of the Trust-CH algorithm.

4.3. Threats Resistance

In terms of security analysis, we used a Speck128 encryption algorithm that is known to be lightweight, fast and secure. As for authentication, it is also analyzed in [26]. Regarding attack types, we used eavesdropping, man-in-the-middle and replay attacks. The FlexCrypt algorithm is used initially in the Trust-CH technique in an effort to lower the cost of authentication, key generation and integrity. Moreover, in an effort to cut costs by depending on the nodes’ shared confidence. After being designated as a trusted node, the Trust-CH mechanism may be used to communicate a unified key (k) to the interaction between the nodes and the CH. So even if the attacker utilized brute force, it was impossible for the adversary to find the Consolidated Key (k). This is due to the fact that k is transmitted in encrypted form based on the first algorithm’s authentication (FlexCrypt). Moreover, By flooding the network with redundant bogus data, posing as a CH or other node using that node’s IDs, or otherwise disrupting it, an attacker might reduce network performance or drain the nodes’ power. In our work, we guard against this kind of attack by adding a sequential number to the session key (Sk), guaranteeing the validity of sent data and preventing data repetition.

5. Conclusions and Future Work

In this paper, we proposed the Trust-CH framework as simultaneous security and power-saving scheme for a wireless sensor network. In the proposed scheme, a mechanism for building trust between the WSN nodes was proposed, depending on the behavior of the nodes and monitoring the messages they send and receive. These messages are represented by Hello, ACK and lost packets. After the trust is built between nodes, the clustering mechanism is activated based on trust technology to reduce communication distances between nodes and save energy. After building trust between nodes and selecting CH, the scheme proposes to reduce cryptographic limitations by using a lightweight algorithm and automating coding parameters based on the level of trust between nodes. Trusted nodes also use a shared key and lightweight authentication, while untrusted nodes use other encryption and authentication algorithms.

The Cooja simulator was used to simulate the Trust-CH algorithm and analyzed for several metrics such as power consumption, network lifetime and degree of impact on malicious nodes. The achieved results of the proposed scheme show that it can effectively save power consumption by 28% compared to Speck128 static encryption. Furthermore, it can enhance the network lifetime by 53%. Moreover, it extends the network lifetime by 40% and 42% compared to the FlexCrypt and TMFOGA, respectively. The proposal supports clustering and dynamic encryption with the same technology.

In future work, we are planning to study the viability of extending the proposed method to deploy and test it for a variety of practical IoT applications due to the rapid growth of IoT technology which has common security problems with wireless networks.