A Multi-Tier Trust-Based Security Model (MTTS) for Resource-Constrained IoT Devices to Detect Blackhole Devices and ensure Authentication

In the Internet of Things based military environment, to achieve an assigned task without compromising security concern is a challenging task. Resource-constrained devices, Self-organized, disseminated nature of sensor nodes, unstable topology environment, open wireless medium, and shared network environment make IoT differ from other networks. All together leads to security threats to the IoT environment. The success of any IoT based application depends on effective communication among the devices that can be achieved by routing protocols. Typically, they designed to cope with fruitful communication but in effect, they might be targeted and affected by mischievous devices. Blackhole devices are such devices that aim to disturb the normal routing operations. But the identification of trusted devices that means authentication needs to be provided for successful task completion. At present, trust management plays an important part in prediction construction and appropriate for resource-strained devices. In this article, a multi-dimensional trust-based security model is proposed to guarantee authentication by avoiding blockhole devices based on direct, indirect, sense-making, social, and stereotypes. In addition, the q-learning mechanism is used to enhance the proposed method. The simulation study represents the applicability of the projected model.


Introduction
In telecommunications, modern military communications have access to an excess of real-time information therefore networking infrastructure might be in place on the battlefield situation. However, getting this information to the warfighter is still a problem due to the mobility of soldiers and unsupported network infrastructure. An alternative solution would be required to deliver information. Network operations on the battlefield aim to provide continuous access to all decision-makers and warfighters to form a clear and perfect view of the battlefield. Traditional mobile IP networks have fixed wired infrastructure. Applying such is impractical for a battlefield situation. To address this, the Internet of Things (IoT) can serve this purpose [1]. It defines as a collection of integrated physical objects that have equipped with sensors and embedded technologies. They can sense and interact with internal or external environments" [2]. The benefits of IoT for military environment benefits ranging from armed vehicles to personnel monitoring [3]. Typically, these types of networks might be manned or unmanned.
IoT based military applications differ from other kinds of applications. The first thing is, the military environment is not affected by protocols that have used. But it is affected by security threats. The second thing is loophole in security means the failure of IoT technology will lead to human life in question. Besides, sensor devices/ military equipment is working together with the assumption of all are operating together. But, achieving such collaboration amongst the nodes always a difficult mission. Because every sensor node should believe other nodes to complete a task. Such, blindness in communication will lead to security violations. Besides, IoT devices themselves have a complication in terms of their energy, processing, and computation and these features also lead to security violations. Alternatively, the wireless environment is reachable to both genuine users and malicious attackers. The From this, we can conclude that should not confine ourselves to enemies from external, but also consider the enemies thrown from inside of the network by compromised devices. Besides, the military environment is too self-motivated. The trust relationship among the trusted devices may also change because there is a chance for adding and removing of nodes at any time. So, ensuring identification becomes a question mark that means guarantee authentication. Due to the above reasons, a militarybased IoT environment is prone to various security violations in the form of various attacks. Among the attack, the black hole attack [11] is getting more attention. The nature of the attack is, drop all incoming packets that have to be forwarded to other devices by those nodes are trying to save their energy. Traditional safety mechanisms are used to address the security violations however it needs high processing, computational, and memory capabilities. Hence, it is not appropriate for resourceconstrained IoT devices. The applicability and success of any IoT based application depend on how it is secured in terms of its various functionalities. To ensure security, security requirements such as availability, non-repudiation, authorization, confidentiality, and authentication are needed such as authentication, authorization, confidentiality, integrity, non-repudiation, and availability [4]. In the security requirements, authentication is essential to provide an initial level of security by ensuring the right identity of the devices which are communicating with each other [5]. The main aim of this article is to ensure authentication by identifying a black hole attack based on trust management. The contribution of the proposed work is described as follows: • This work starts with explaining the basic things involved in the research work included Qlearning, sense-making, and stereotypes. • Then, this work discusses the related work and need of the proposed work.
• Then, the proposed MTTS model discusses. The main aim of the work is to identify the black hole attack over RPL routing protocol in IoT based military environment by the way authentication of the participating devices will be ensured.

Authentication, Trust and its properties, Stereotypes, Sensemaking and Q-learning
The proposed MTTS, involving various fields such as trust management, sense-making, stereotypes, and Q-learning. Figure 1 illustrates the backbone of the MTTS model. The following section describes the contribution of each field.
As discussed earlier, once authentication is achieved, the remaining security requirements can be achieved easily. Though authentication assures the right identity of soldiers, it does not guarantee that they cooperate and behave as expected because soldiers may transform their behavior and misconduct purposely as a lack of resources that are having [6]. From this, we derive that a proper mechanism is expected to distinguish the correct objects and adversaries to secure the mission and improve its performance. Trust computation shows an important part in taking decision [7], it is utilized by many technologies such as wireless sensor networks and MANET [8], peer to peer networks [9] and ecommerce [10], etc. To achieve authentication, various cryptographic techniques such as digital signature and certificate [11], public key infrastructure [PKI], hash functions and etc are used but these techniques always depend on a third party and they need the centralized object to monitor [12]. It requires more computation power, memory, bandwidth, and battery-operated control. To overcome such issues, trust management comes into play for providing security. The field of social science is the ancestor of the term 'Trust'. The definition is "one object/trustor is prepared to be contingent on another object/trustee [13] [14]". The network and communication domain give an alternative definition for the trust that is "a cluster of relatives among objects and their contribution to building on the earlier communication of objects within the protocol" [15] [16]. Trust management has the following properties. Trust is dynamic [17], subjective [18], incomplete and transitive [19], asymmetric [20] and context-dependent [21]. In this research work, we focus on subjective and context-dependent properties while evaluating trustworthiness. Figure 2 shows the direct trust representation where 'A' trusts 'B' likewise 'B' trusts 'A' whereas in figure 3 represents the    . Stereotypes (www.dweebist.com) Another thing has used in this concept is stereotypes. It is defined as outlooks completed around a collection of people and it will be applied to persons irrespective of their characteristics since of their association with a certain group. Stereotypes can be classified into three category such as positive, negative or neutral [22] [23]. Figure 4 illustrates the stereotypes model where people who interest in finding something, they always look for google site and people who interest in music always look for YouTube. Another concept is sensemaking and it is defined as what persons ensure to choose how to turn in the circumstances they encounter [23] [24]. Figure 5 illustrates the function of sensemaking.
In MTTS, the Q-learning algorithm is used to enrich the proposed model that discussed later. The concept of Q-Learning algorithm has introduced in the year 1989 by Watkins [25]. The algorithm is defined as follows: This algorithm compromises of an agent, states 'S' and a set of activities per state 'A'. An action 'A' has received after the change of an environment. A reward can be assigned after executing a specific action. The objective is to attempt to discovery an optimum strategy that inspires the agent to get the total reward during the entire process [26], and based on the total reward decision will be taken. The algorithm is defined as, where r (S, A) is a direct reward, the discount factor represents as γ. This factor used to govern the status of the forthcoming reward. The value of discount between 0 and 1 range, S' represents the new state

3.Related Work
This section discusses the mechanisms that are available to ensure security in the IoT environment.
Liang Liu et al, 2019 [27] proposed the trust mechanism which is based on trust and k-means. In this work, trust is calculated based on the perceptron. This work deals with replay, drop, and tampering attacks. Also, an enhanced learning process is used to increase the detection accuracy of those attacks. Noshina Tariq et al, 2019 [28] proposed a mechanism to detect internal attacks that are based on codedriven. The factor called forward behavior is used to detect the internal attacks such as black and grey hole.
F.Ahmed et al, 2016 [29] proposed a trust-based mechanism based on the forwarding behavior of the sensor nodes. The main objective of the paper is to detect black hole devices from the IoT environment. Chen et al, 2011 [30] implemented a fuzzy repudiation-based trust organization to mitigate internal attacks. They have mainly considered the Quality of Service factors for instance delivery and forwarding capabilities of packets and utilization of energy.
Bao F et al, 2013 [31] proposed a model based on direct and indirect recommendations based on the following properties of trust such as the interest of community, honesty, and cooperativeness. The ultimate objective is to detect and eliminate attacks that occur in the routing process. Similarly, based on the forwarding behavior of participating devices, David et al, 2017 [32] proposed a mechanism that is incorporated with the RPL routing protocol. Attacks handled in this paper are selective forwarding and blackhole attack. The above techniques are effective in detecting the black hole attack however the detection ratio needs to be improved to improve the performance of the entire network. This can be achieved by the proposed MTTS model.

.2 Tier 1: Direct Trust evaluation
As discussed earlier, all the devices are collaborating healthy and reliable initially. Because of the environment changes, devices can act as black hole devices hence poor mission performance. Consequently, each device in a condition to validate its interactive militaries by accomplishing the MTTS model to leads the mission successfully. Various factors involved in the MTTS model included direct, indirect, sense-making, and social trusts. The reason is, sometimes a single trust value is not sufficient to assess the reliability of a particular device hence various factors of trust values have used herein. For instance, a device 'A' would cooperate with device 'B' therefore the cooperation between them is good in terms of providing services. At the same time, device 'A' might not with device 'C' due to surprising circumstances. This concludes a single direct trust is not sufficient to assess the trustworthiness [23]. The MTTS model has described as follows; Every device ensures their one-hop neighboring nodes by sending a special packet called "HELLO". Based on the response for this packet, every object can know its neighboring nodes who remain in a similar communication range. By the way, every device accomplishes how many devices are continuing as one-hop neighbors. After that, each device executes the MTTS model. In MTTS, the technique called "passive acknowledgment" [23] has used to screen the forwarding behavior of participating devices. The following equation 1, calculates the direct trust of soldier 'j' for soldier 'i' and figure 7 shows the direct trust evaluation.
In equation 1, 2 ( ) denotes direct trust value, ( ) represents the control packets forwarding ratio of solider 'j' for solider 'i' over n period. Similarly, ( ) denotes the data packets. μ1 and μ2 are weighting factors. In networking, each routing protocol contains the subsequent packets such as control and data packets. These packets are using while routing discovery and route maintenance. In the route discovery process, control packets such as RREQ-Route Request, RPLY-Route Reply have used. During the route maintenance, RERR-Route Error packets have used. Trust evaluation is also involved with these packets as they are providing significant effort towards continuous network operation. Blackhole devices will process those packets though compare with legitimate devices the process of such packets is low. From this idea, we calculate the forwarding and responding behavior of the control packet. The following equation 2 used to calculate the same.
ℎ , , , = 1,2,3 … ≠ denotes no. of data packets forwarded by soldier j with respect to soldier and denotes the no. of data packets received by soldier 'j' from soldier 'i'. Likewise, every device could calculate the control packet and data packet ratio and applied in equation.1 by the way trust value their members could be calculated by other soldiers and update their trust table.

Tier 2: Indirect Trust evaluation
In tier 2, solider 'i' will calculate the indirect/recommendation trust of solider 'j' and the recommendations from solider 't'. Before this, as tier 1 executed, everyone knows the direct trust value of others. Equation.4 will calculate indirect trust. The following equation 4 is used to calculate the indirect trust value through recommendation collection from others and figure 8 shows the indirect trust evaluation. In equation 4, 2 _ ( ) represents indirect trust, 2 represents direct trust received from soldiers t about soldier j and 2 ( ) represents its own trust value.

Tier 3: Sensemaking trust evaluation
As the definition stated earlier, the sense-making trust will be calculated based on the soldier's responses to the tasks assigned to them. They will respond immediately if they understand the objective of the mission clearly. Otherwise, they will not, as they may be in trouble during the mission execution. With this idea in mind, sense-making trust has calculated. Based on the equation 5, the sense-making trust has calculated. In the above equation, 2 ( ) denotes social trust, denotes average response by a solider and denotes total requested response to a soldier.

Tier 5: Overall trust or aggregated reward computation
After the trust aggregation tier, the aggregated trust values obtained for every soldier. An evaluating soldier assigns a reward for each interaction that had with its one-hop soldiers based on the stereotypes model. As discussed earlier in [16], they are three categories of reward will be awarded under a team such as positive, negative, and moderate. The reward value ranges between 0 and 1. 1 specifying the maximum, 0 specifying no reward, and 0.5 specifying moderate reward. The specification mentioned below. Thereafter, evaluating soldiers utilizes the Q-Learning algorithm to evaluate the overall performance of its neighbor soldiers because a soldier can get a high reward for some actions and vice versa. Based on equation.7, the aggregated reward also called Overall Trust (OT) has calculated. denotes the maximum reward. The maximum reward can be assigned based on equation 8, with the help of aggregated trust that has calculated in equation.6. Then, the immediate reward for all the neighbors has calculated based on social trust. It can be calculated based on the observation of soldiers/devices about other soldiers/devices in terms of various properties such as friendship, honesty, friendship, and cooperativeness. In addition, each manned device has a different ability of cooperativeness and energy. Hence, this work considers those factors as an immediate reward and it can change over the period of time. For illustration, the device with the highest capacity will give less reliability rank for the device with the lowest capacity and at the same time will give more reliability rank for high capacity devices [28]. With this idea in mind, social trust is considered and applied in aggregated reward calculation. Equation 9, is used to calculate the social trust.

Tier 6: Identifying black hole nodes
By applying equation.8 and equation.9 in the equation.7, the aggregated trust value is calculated. Then, based on the aggregated value, the decision will be done. In table 2, the soldier's trust levels have presented, and their reflection. Figure.9, shows the soldier's interpretation in the presence of aggregated reward. These values have classified into three categories first one is trusted soldiers; we can allow those soldiers in mission activities. Second is partially trusted soldiers; this work allowed those soldiers to take part in the routing operation but they will not be involved in mission execution activities. Finally, misbehaving soldiers also called black hole soldiers, and those soldiers are isolated from the mission, and information about the misbehaving soldiers can be broadcast by commander therefore those soldiers will be deleted from all the soldiers' trust tables. By the way, such soldiers isolated from the mission, and authentication has ensured. interactions. Then, the soldier's satisfaction degree is calculated based on the neighbor soldiers' aggregated reward over time.

Results and Discussion
The Contiki/Cooja 3.0 has used to implement the proposed work. The simulation area is 500m x 1000m. The maximum speed of nodes are restricted to 2 m/s and used TMote Sky mote as mote type. The core routing protocol is here the Routing Protocol for Low Power Lossy Networks (RPL) Simulation parameters represented in table 3. The table 4 represents the threshold values setting. The evaluation of the proposed work has done for the following cases.
1. The aim is to identify the black hole nodes so that it is necessary to know the effect of black hole nodes. In this regard, including the black hole nodes by increasing percentage and observed the packet dropping ratio. The analysis has executed with changing blackhole nodes under normal RPL routing protocol. From the simulation results, we observed that whenever the blackhole nodes increase activity like packet dropping also increases proportionally. This is an initial experiment. It shows the impact of blackhole under traditional RPL routing protocol in a military scenario figure 10 shows the probability of misbehaving activities.  The table 5 showed the dataset of 20% of black hole nodes and its detection under the MTTS model and plots the graph. Likewise, we made the dataset for 40% and 60%. For better understanding the figure. 11 showed the trust evolution analysis and based on the figure, the figure. 12 and figure. 13 demonstration the finding of blackhole nodes for the dataset with 20% and 40% of adversaries.    [32] when the number of black holes has risen in MTTS model, the probability of finding also intensely improved compare with [32] model. The reason is in MTTS, we make use of multiple trust metrics to analyze the overall trust. Whereas in [32] model, they have only considered the packetforwarding behavior because of that reason the detection ratio is also low compare with MTTS. In general, traditional RPL cannot detect black hole nodes hence we did not take it into account.  [32]. The reason is, black hole nodes are detected and isolated therefore they avoided. In MTTS, the trust evaluation mechanism is involved with various factors whereas in [32] they have considered only forwarding factors. Hence, the improved packet delivery ratio in MTTS. Then, the RPL routing protocol does not have any detection mechanism of black hole nodes therefore result in a less packet delivery ratio. End to End delay: This is calculated by the average time taken by a packet from the source to the destination. The end to end delay of MTTS model with [32] compared in figure 16. From figure 16, end to end delay is low for proposed MTTS model as black hole nodes are avoided. As RPL has no ability to detect black hole, high end to end delay compare with [32]. Due to the weakest measurement of trust in [32], end to end delay is high compare with proposed MTTS model.

6.Conclusions
As applications of IoT are increasing, weakness insecurity is also increasing proportionally. The success rate of IoT deployed military depends on how it is security is defined. Authentication is a primary security requirement in the military scenario to ensure the correct identity of soldiers. As this focus in mind in this paper, Multi-tier Trust-Based Security model (MTTS) to identify black hole nodes to ensure the authentication for military scenario has proposed. In this model, the proposed model makes use of each control packet to calculate the trust evaluation because we cannot say that all the packets have well processed in this distinct environment. The sensemaking concept is also to ensure whether the soldiers are behaving as per the commands or not. Q-learning algorithm has used to enrich the trust calculation by giving an immediate reward for each soldier by assessing their current behavior and posing of resources and maximum rewards for aggregated trust values. It helps to make the correct decision on soldiers by the way authentication has ensured by detecting black hole nodes. Besides, this work does not avoid partially trusted soldiers and gives a chance to take part in the routing operation so that coordination has increased. In this model, communication overhead and memory overhead are low because each soldier will invoke the MTTS model only the satisfaction degree of the mission is low so that our model does not involve to store too much trust values and do not communicate unnecessary with other soldiers while the mission is in progress. In the future, the proposed work would be concentrating on other routing protocols over other IoT applications like healthcare, agriculture and connected vehicles.