Enhancing IoT Security with Trust-Based Mechanism for Mitigating Black Hole Attacks

Mobile Ad Hoc Networks (MANETs) are rapidly evolving, self-organizing networks that exhibit a remarkable capability for swift deployment. Their adaptability has proven beneficial in numerous real-world applications [1]. With a surge in the number of portable devices and advancements in wireless communication, the popularity of ad hoc networking continues to grow. This networking paradigm's almost infrastructure-less nature allows for its use anytime and anywhere. The applications of MANETs span a wide range, from small, static, power-constrained networks to large, mobile, dynamic networks. As legacy systems adapt to such ad hoc environments, it is expected that the number of devices and applications will increase [2].

The inherent properties of MANETs, such as self-organization and adaptability, require devices within these networks to possess the ability to detect other existing devices. This is essential for performing the necessary setup to facilitate communication, data sharing, and service provision. In addition to maintaining network connections, these devices should be capable of adding and removing devices from the network. The highly mobile nature of the nodes can lead to unpredictable and swift changes in network topology over time. Given the decentralized nature of these networks, the nodes must ensure that both data delivery and network organization are effectively carried out.

In MANETs, users can access and exchange information irrespective of geographical location or the presence of traditional network infrastructure. The absence of a fixed infrastructure, a unique characteristic distinguishing MANETs from other mobile networks, and dynamic connections lend these networks a distinct advantage in terms of robustness and flexibility [3]. The primary objective of MANETs is to extend this flexibility into self-sustaining, mobile, and wireless domains where nodes can function as both routers and hosts, forming the infrastructure to route the network in an ad hoc manner. The dynamic nature of wireless networks makes routing a significant challenge. An ad hoc network, in a typical setting, is characterized by a set of constantly moving nodes that form a transient network dynamically in the absence of any network infrastructure. Within such a network, mobile nodes coordinate to exchange information for specific durations. Given that these nodes remain mobile during data exchange, networks must continually adapt to establish paths without external support.

Various ad hoc networks have been developed with different routing protocols, which can be categorized based on several conditions. Route discovery forms a critical criterion for categorization. Based on this, routing protocols can be either reactive or proactive. In the former, such as Dynamic Source Routing (DSR) and Ad hoc On-demand Distance Vector routing (AODV), route requests are initiated on-demand. If a node wishes to establish communication with another, it sends a route request and awaits a response from the destination [4]. Contrarily, proactive protocols such as Optimized Link State Routing (OLSR) continuously update routing information to maintain a comprehensive overview of the network topology.

The Ad hoc On-demand Distance Vector (AODV) protocol is among the most commonly utilized within the realm of MANETs. Its distinguishing feature is the on-demand routing mechanism, which is initiated only when paths to other nodes need to be discovered or verified, thereby enhancing the routing efficiency. The two fundamental operations of the AODV protocol are route discovery and route maintenance. Route discovery is initiated when a node intends to communicate with a destination for which it lacks a valid route entry [5]. Route maintenance can be achieved in two ways. One approach involves a node broadcasting connectivity information via local hello messages, enabling nearby nodes to listen to these packets and determine connectivity. An alternative approach utilizes a link or network layer mechanism, such as the IEEE802.11 Media Access Control (MAC) protocol [6].

The Ad hoc On-demand Multipath Distance Vector (AOMDV) protocol represents an evolution from the AODV protocol. This reactive protocol is designed to identify multiple node-disjoint and link-disjoint paths [7], evaluating multiple loop-free paths. A drawback of this approach is the increased message overhead due to extensive flooding during route discovery. As a multipath routing protocol, AOMDV might encounter substantial control overhead due to the necessity for the destination node to respond to multiple Route REQuests (RREQs). This results in the generation of larger overhead packets in response to a single RREQ packet [8].

In recent times, trust-based routing has been considered a promising security solution. To protect the routing protocol, a trust-based scheme is employed. Each network node independently implements a trust model to estimate its trust level in other network nodes, which can then be incorporated into routing decisions. Unlike traditional routing protocols which prioritize establishing the shortest route, trust models aim to establish the most trusted routes [9].

For design and the deployment of the security systems, trust is a crucial factor. Trust evaluation in MANET can be used in trust routing, node authentication and access control. Evaluating the trustworthiness of the related nodes results in enhanced security and also better routing performance. Trust refers to the belief that is held with respect to factors like dependability, reliability, honesty, competence etc. in reference to a particular context.

The evaluation of the direct trust value, which may be either positive or negative, is predicated on the direct experience of the trustor node with the trustee node. In instances where the trustor lacks sufficient trust in the trustee node, the recommendation of a third node may be necessitated. Once both direct and recommendation trust values are acquired by the trustor node, a combination formula may be required to strike a balance between the direct and recommendation trust [10, 11].

This paper proposes an augmentation of the AOMDV protocol with a trust model, aimed at enhancing network performance. The second section provides a review of the relevant literature. The methodology is elucidated in the third section, while the fourth section presents the empirical findings. The paper concludes with the fifth section.

In the context of Mobile Ad hoc Networks (MANETs), blackhole attacks manifest when data packets are enticed by malicious nodes through the utilization of fresh routes, followed by the abandonment of these packets [7]. This work introduces a proposal for a trust-based routing method intended to mitigate blackhole attacks.

The trust metric contains the following properties [11]:

• Dependency on the context: Only in certain contexts, the trust relationship have a certain meaning.
• Function of uncertainty: Evaluating the possibility whether an entity performs or not is referred to as trust.
• Whether a given entity will perform the action or not is evaluated by Trust.
• Quantitative values: either continuous or discrete numeric values can be used for representing trust.
• Asymmetric: Trust need not be two way process. That is, if AX trusts Y, it is redundant to clutch that Y trusts X.
• Transitive: Trust is transitive, i.e., Suppose node B is trusted by A and node C is trusted by B, then C must also be trusted by A.
• Personalized: The matter of trust is highly subjective and the trustworthiness about the same entity may be evaluated differently by two people.

The suggested trust based ad hoc on demand multipath distance vector protocol is explained in the current section.

3.1 Ad hoc On-demand Multipath Distance Vector

A multipath extension on top of AODV is the ad hoc on demand multipath distance vector routing. For enabling multiple paths, the process of route discovery has been altered. The disjointedness of the links is considered for multiple paths so that the nodes only and not the edges are shared. Also, using the sequence number of the nodes assures loop freedom. Due to the need to address the problems, split the traffic along each path and packet reorder at the destination, a single path is chosen. Another feature that distinguishes the AOMDV is the usage of periodic HELLO messages for detecting the sales path [25].

A list of subsequent hops plus the hop counts for these are included in the routing entries, for keeping a track of multiple paths. Same sequence number is possessed by all the next hops. Hop count that is promoted is defined as the maximum hop count for all the paths. This, in turn, is maintained by a node for every destination. For sending route promotions to the receiving nodes, this is the hop count that is used. Each duplicate route promotion that a node receives defines an alternate path to the destination.

If and only if the hop count for a path is lesser than the promoted one for the destination does a node accept that alternate path. This is to ensure loop freedom. As the greatest value of the hop count is used, the promoted hop count remains the same for the same sequence number [26]. The subsequent hop list along with the advertised hop count will be initialized again when the route promotion is obtained for a destination that has a higher sequence number. For finding node-disjoint or link disjoint routes, AOMDC can be used and for finding node-disjoint routes, the duplicate RREQs are not instantly rejected by every node.

A node- disjoint path is defined by the source with different neighbor arriving by each RREQ. Any two Route requests arriving at an intermediate node with neighbor distinct from that as the source wouldn’t have crossed the same node as the nodes cannot send repeated route requests. For obtaining multiple link-disjoint routes, the sink node sends responses to duplicate Route requests. The response of the destination is to the RREQs that have unique neighbors. The RREPs after the first hop, follow the reverse path, which are both node as well as link disjoint. Even though the trajectories of every RREP may traverse the intermediate node, every RREP takes a distinct reverse path to the source for ensuring link disjointedness [27].

The AOMDV, while still choosing the disjoint paths, enables the intermediate nodes to respond to the RREQs and this is the main benefit it offers. AOMDV has a lot of message overheads, as there is increased flooding during route discovery. Also, being multipath routing protocol, the destination responds to the multiple RREQs whose results are in longer overhead.

3.2 Proposed trust model

Trust derivation, its computation and evaluation are performed by the trust model. It is from the packet forwarding ratio that every node derived trust factor. A linear aggregate technique has been used to estimate the trust of the entire node trust as per the trust factors and a path’s trust can be evaluated least value method. Trust based discovery of route and its selection are included in the trust application.

The firsthand information for the neighbors is Direct trust which can be easily achieved. Using the history of the direct interactions among the nodes for trust computation, the trust model can be simplified. After obtaining the service of a forwarding node, the procedure of trust evaluation is a routing process which is a remark of the sender. Due to gray hole, black hole attack, heavy traffic and poor wireless communication are the factors for packet dropping. Thus we can say that the quality of forwarding is evaluated using packet forwarding ratio.

The cumulative count of accurate forwards is denoted as N_C(t), and the total count of all requests made up to time t is represented as N_A(t). Within a given timeframe from time t_i-w to t_i, the correct forwarding count is calculated as N_C(t_i) - N_C(t_i-w), with w signifying the duration of the time window. The packet forwarding ratio within the i-th window, denoted as FR(t_i), is established according to the following Eq. (1) [28]:

$\mathrm{FR}\left(\mathrm{t}_i\right)= \begin{cases}\frac{N_C\left(t_i\right)-N_C\left(t_i-\mathrm{w}\right)}{N_A\left(t_i\right)-N_A\left(t_i-\mathrm{w}\right)}, & t_i>w \\ \frac{N_C\left(t_i\right)}{N_A\left(t_i\right)}, & t_i \leq w\end{cases}$             (1)

where, i=1,2,3.

The concept of verifying whether packets dispatched from node j to node k for forwarding have indeed been forwarded by node k is encapsulated by the notion of trust between the two nodes. This trust factor is referred to as the trust of node j towards node k. To ascertain the overall trustworthiness of a specific node, the trust values from two trust factors, namely the Control packet Forwarding Ratio (CFR) and the Data packet Forwarding Ratio (DFR), are assigned relative weights. The direct trust of node j in node k, denoted as T_jk, is quantified using the subsequent Eq. (2):

$T_{j k}\left(t_i\right)=w_1 \times C F R_{j k}\left(t_i\right)+w_2 \times D F R_{j k}\left(t_i\right)$            (2)

In this equation, CFR_jk(t_i) corresponds to the control packet forwarding ratio, and DFR_jk(t_i) indicates the data packet forwarding ratio witnessed by node j when forwarding to node k at time t_i. The parameters w₁ symbolize the weights allocated to the Control packet Forwarding Ratio (CFR), while w₂ symbolize the weights assigned to the Data packet Forwarding Ratio (DFR).

Within the phase of route discovery, the RRQ flooding invasion happens. It attacks the packet in route that invalid network destination. An IDS observe the cohort of the packet from the source and the source’ trust value is allocated. Initially every node allocates the trust value like one. The original trust value is measured by an IDS; then, as per the route discovery of these nodes, the network is apprised on this by the IDS. When IDS check the trust value for all the neighboring nodes [29]. Every node checks for the source’ trust value when the source node sends the route request packets to the neighbors. If the source’ trust value <the threshold, the source is thereby dropped, thwarting the attack. In case the source’ trust value is distinct from all the nodes, its trust value will be lesser than the threshold because of which it cannot forward the route request to the destination node.

For establishing a trusted routing path, the trust model carries out the following steps [30].

1. Initializing the statistics that are concerned with network trust and that of all the network nodes.
2. After the operations like dropping and forwarding of packets etc. of every node, updating the network statistics related to trust.
3. Broadcasting route request for the destination when it is established.
4. Upon reception of a route request by a destination, a route reply is sent and the route’s trust is initialized to 0.
5.  The intermediate nodes compute their trust value using network statistics when receiving route reply; this is added to the path’s trust in the route reply packet.
6. The source node stores the path when it obtains a reply for the first time. The data packets are sent along that path. The source node, on obtaining many path replies will compare the New Andy received path’s trust with that of the current route and the path having the maximum trust is stored.

The routing path is updated to the ones with most trust, periodically.