Enhanced security using multiple paths routine scheme in cloud-MANETs

The clustering process on MANET nodes allows adaptation to topology changes in the network, thus ensuring the efficient utilization of the available bandwidth and high network throughput. A variety of clustering schemes were developed for making the network energy-efficient and improving the network lifetime. However, for the purpose of data forwarding, and with the increase in the cluster number, energy consumption tends to shoot up. Existing clustering approaches have several drawbacks which include: increased control overhead for the purpose of retaining the backup routes, highly complex approaches, and the recurrent beaconing.

The devices in MANETs operate on limited energy resource i.e., batteries [8]. The uncontrolled mobility of nodes leads to the formation of a network that is highly dynamic and changes its topology rapidly. Thiscan all result in the frequent failure of routes. For nodes to remain connected and be able to communicate for longer times, low delay and high throughput is required from the routing approach used. Existing routing approaches do not provide desirable results under such cases.

The heterogeneous devices forming the MANETs have inconsistent data rates and different processing powers, which may lead to problems in communication. Majority of the existing works on clustering focus on issues such as reducing the energy consumption of mobile devices, reducing the maintenance overhead, and increasing the network lifetime.

In a clustered MANET, all routing from outside the network or internet is directed through the Cluster Head (CH). There may be a difference in the rates at which data is received from the outside service provider and forwarded to the member node. Thus, the communication process experiences unnecessary delay because of the difference in the reception and transmission capacity of the CH. It may also lead to increased packet loss due to buffer overflow at the nodes.

With the node movement and clustering, reorganization of nodes and changes in the network topology occur frequently in MANETs. The rapid dissipation of energy and changes in the nodes’ topology may lead to the failure of links. An optimal solution to this problem is missing in existing research. Additionally, the dynamic nature, open medium, and decentralized control of MANETs makes it highly susceptible to attacks. In the traditional works, a trust-based framework is offered by the central parties like Trusted Third Parties (TTP). In the clustering-based frameworks, the cluster head acts as a TTP for providing a secure communication in inter and intra cluster routing. In order to manage the cluster members in the group, the cluster head requires lightweight key management algorithms. To satisfy such objectives, many traditional works proposed different algorithms, but the problems of increased computational complexity, network overhead, inefficient security, unreliability, and reduced network throughput rendered them less useful.

This work aims at designing a routing protocol for MANETs which is energy-aware, provides security and stability, and enhances the chances of successful data transmission in disaster response and rescue applications. The major contribution of the proposed work is the development of an energy-efficient routing protocol that allows a reliable and secure exchange of information between modern MANETs. For achieving energy-efficiency, increased network lifetime and scalability in dense areas, the clustering approach is applied. Also, the multiple node disjoint paths help in accomplishing reliability, stability and network robustness. We attempt to create a maximal route between a pair of nodes, and to ensure the appropriate and accurate delivery of the packet.

The most commonly used cluster-based routing protocols [95] for MANETs are energy-based, connectivity-based, mobility-based, and weighted. Some of the existing works based on clustering and cluster-based routing are discussed below:

In [9], SYN, an energy-efficient clustering protocol was proposed. It used the Max-heap tree to model varying levels of energy of the nodes. In [6], the authorsintroduced a Particle Swarm Optimization (PSO)-based protocol for optimizing clusters and conserving energy in ad-hoc networks. It discussed the Zone-based protocol for routing with a corresponding collision-guided broadcasting algorithm that used one-hop clustering [10]. In the MBHC method, the clusters are formed according to the node mobility pattern [11]. The cluster head is selected considering the lowest stability value amongst neighbors. Here, the distance travelled by any two nodes estimate the relative mobility between them over time. It allows the immediate merging of clusters with similar mobility patterns. The mobility pattern is considered an essential measure for cluster formation as two nodes with similar velocity and direction have lower mobility difference and are grouped together in the same cluster. The major advantage of this scheme is the minimization of the cluster number by taking into account the group mobility pattern, and addressing the issue of scalability for routing protocols used in large MANETs. Cluster formation in the EPAC method is done on the basis of the energy requirement for the completion of tasks assigned [12]. The node with the lowest ID, which is dependent on the maximum energy left, is chosen as the cluster head. The algorithm tries minimizing the energy consumed in a cluster and the CH selects transmission loads based on the energy available at the cluster nodes. The major advantage is the formation of energy-efficient and stable clusters. This also reduces the energy drainage and chances of re-clustering. WBDC is a weight-based clustering based on the collective weight metrics: ideal node-degree, mobility, and energy of the nodes [13]. The weight of every node is estimated considering the weighted metrics. The node with the maximum weight amongst the neighbors is made the cluster head. The threshold value determines the re-affiliation process. When the mobility metric value of a node is higher than the threshold value, the node starts to move out of the cluster. The algorithm considers realistic network parameters and includes the flexibility of weighing factor adjustments. Neethu and Singh [14] proposed an approach for re-clustering in MANET routing protocols where security can be a challenge. The process involves the generation and sharing of different keys to all the member nodes. Sung et al. [15] designed a distributed cluster formation scheme which involves algorithms for the creation and maintenance of clusters in high mobility networks. Loutfi [16] discussed the fault tolerant multipath that is mobility-aware using multicast routing protocol for MANETs. The main objective was to enhance the reliability of the routing protocol in spite of the continuous movement of the nodes. The experimental result of the proposed method attained better fault tolerance by means of an enhanced packet delivery ratio and decreased delay. In [17], a new weight-based clustering scheme for improving security of MANETs was designed. In [18], the authors proposed an energy-efficient cloud-based IoE framework to integrate cloud computing with the Internet of Everything and provide valuable services to consumers. The framework is optimized by clustering the IoT networks using a wind driven algorithm.

Table 1 summarizes the cluster-based protocols for routing in MANETs discussed above:

The issues connected with the existing clustering algorithms are: loss of energy when a node continues to be a CH for a long duration of time, and increasing communication overhead as the cluster number increases [95]. Also, the throughput of the network is reduced and energy usage is increased due to the CH re-election.

The major disadvantage of connectivity-based clustering algorithms is throughput reduction as the number of cluster nodes starts increasing. The mobility-based clustering algorithms undergo degradation in performance because of the randomly moving network nodes. The shortcoming of the weighted clustering algorithms is that for clustering to begin, it requires all the member nodes’ weight parameter. The energy-based category faces the issue of requirement of increased overhead to select the CHs.

Routing in MANETs can be quite challenging owing to the node mobility, dynamic topology and constrained resources. To improve end-to-end throughput and achieve balancing of load in the limited resources environment, multipath routing should be adopted [23]. Multiple paths can either be link disjointed or node disjointed. Node-disjoint multipath have an edge over link-disjoint as no matching node exists between any route, which ensures that at least one route remains available whenever node or link failure occurs.

A lot of research has been done on clustered networks applying multipath routing in MANETs. AOMDV (Ad-hoc On-demand Multipath Distance Vector) [19] showcases an extended multipath version of AODV (Ad-hoc On-demand Distance Vector) [24]. It is a simple routing protocol with link disjoint characteristics which aims at discovering loop-free routes. SMR (Split Multipath Routing) protocol uses multipaths that are based on maximum disjointness [20]. As a result, load balancing is required over multiple paths. A new weight metric is used by EIDM for path selection, which combines the remaining energy, interference and packet drop rate, along with mobility of the nodes [25]. An energy-efficient multipath routing protocol MEER (Multipath Energy-efficient Routing) uses an approach of energy control to help the network extend its lifetime [26]. Cluster-based multipath routing (CBMRP) distributes data along multiple paths and helps in reducing traffic congestion [27]. The major objective of routing is to determine paths that have less interference for smooth data transmission. Another multipath routing protocol which is also energy-efficient is ES-CMR (Energy-aware and Stable Clustered-based Multipath Routing protocol) [28]. The protocol works on keeping the energy of the nodes at optimum value, thereby keeping the network stable and enhancing its lifetime.One of the major issues connected with the existing algorithms is that some protocols don’t take energy into account or manage energy levels optimally, resulting in the death of nodes. Some approaches don’t use multiple paths to distribute network load, thus not utilizing network resources properly. This results in a decrease in the network efficiency,especially during conditions like congestion and link failure. Others don’t apply load balancing approach well, creating issues in big MANETs that are energy and power-constrained.

The proposed work is divided into four phases. First is the increasing of the network stability and energy efficiency in MANETs. This is done using a weight-based clustering scheme that focuses on electing a Cluster Head (CH) by using a weighted approach based on energy and node mobility. In the second part, efficient data traffic management is provided in heterogeneous networks when MANETs are connected to external networks. This is achieved using the data offloading mechanism. The data is offloaded in cases where the nodes in the network either have low data rate or low energy. This results in reduced energy consumption, traffic and better throughput in heterogeneous MANETs. Third is the increasing of the network lifetime and enhancing robustness to failure. This is done by designing a node-disjoint weighted multipath routing protocol where optimal paths are selected from the multiple paths chosen on the basis of a weighted scheme. The final phase issecuring the network against malicious attacks. This is achieved by designing a secure cryptography-based mechanism for MANETs that employs an Elliptic Curve Cryptography (ECC) encryption mechanism that provides security against malicious attacks and gives high throughput. Throughput in MANETs is increased considerably if protected against malicious attacks.

Firstly, a clustering approach is designed which enables energy conservation, low communication traffic and overhead, and provides high network stability [47]. It is designed to set threshold valuesfor various parameters in dynamic ad-hoc environment. Environments like these haveunrestrained node movement and low bandwidth wireless channels along with energy limitations. The mechanism proposes the use of two types of CHs:primary and secondary for stable communication in spite of mobility of the nodes. It is an energy-aware and stable clustering mechanism. Here, the CH is chosenusing a weight-based approach that uses ideal energy and mobility parameters. The proposed scheme operates in the following phases:

The following sub-sections elaborate these three phases.

The network is broken down into clusters which are made of CHs and cluster member nodes. All the cluster nodes have a unique number called the cluster ID. Every node stores its updated status value i.e., energy and mobility values. At the time of initialization, all the network nodes send HELLO messages. This message is a broadcast message. The objective of the HELLO message exchange is the identification of the transmission range of nodes. The strength of signal obtained between any node pair estimates the distance between them. It allows every node to determine its neighboring nodes, which are within the defined transmission range. It also defines the various messages exchanged during the process. The various steps involved in the initialization phase are as follows:

In the proposed approach, every node in the cluster is at one-hop distance from the CH. All the nodes that fall under a particular range (predefined) of the selected CH become members of that cluster. The nodes lying outside initiate an iterative CH election process which continues till all the nodes either become part of a cluster or become the Cluster Head (CH) themselves. A threshold is also predefined for residual energy of the CH. Once this energy falls below the threshold, a new CH is assigned. Figure 2 below illustrates the flow of CH selection.

Figure 3 illustrates the proposed data offloading mechanism with energy efficient clustering. Initially, once the network has been created, the neighbors exchange messages and a neighbor table is created. Clusters are created after the CH selection using the weighted scheme. Single-hop clusters with maximum capacity cluster head are designated to transform data received according to the data rate specifications.

To decide if offloading is needed, the following conditions are checked:

The CH after receiving data from the external cellular network, has to convert the processing rate of data correspondence to the value of the data rates of the nodes that are a part of the cluster. The data is brokendown on the basis ofmember nodes’ data requirements and the MTU size of the connecting link. Before any data transmission, the CH checks if the value of the available link data forwarding rate matches the defined data rate specifications. The data continues until there is no data left at the CH in the offloading queue.

The algorithm for the proposed scheme can be illustrated through the steps below:

Or

If either Eq. (10) or Eq. (11) are true, i.e., the nodes have lower data rate than their CHs or if the node’s energy is getting low, the process of data offloading begins.

Next, the management of the network traffic with data forwarding is done utilizing the data offloading mechanism.

Figure 4 illustrates data traffic management in MANETs. The cluster includes different types of hostswith different data processing rate requirements. In a cluster, a member initializes communication by requesting data coming from the server host passing via the CH. The CH will then be able to receive data coming from the server by utilizing a 4G connection of about 30 Mbps. A mobile device processing capacity is between 2 and 7 megabits per second. In such case, the cluster must perform the data transmission based on the node’s requirements as well as the processing power.

To conserve energy and bandwidth of the CH, the nodes with similar requests are grouped together by the CH in order to retrieve data from the server. It can thus avoid wastage of bandwidth and data traffic.

With the movement, clustering and reorganization of nodes in MANETs, topology changes are inevitable. The rapid dissipation of energy and changes in nodes’ topology leads to breakdown of paths. An optimal solution to this problem is missing in the existing schemes as explored in the last section.

The proposed multipath process uses a weighted approach for selection of paths in networks with clusters, and has the following features:

In a clustered network, for transmission of data, the cluster member passes the packet it wants to transmit to some other node, to the CH first. If both the source and destination nodesare in the same cluster, intra-cluster routing is applied. If it is part of another cluster, then inter-cluster routing is used. It uses three types of messages, like in other multipath routing frameworks: Route Request Message (RREQ), Route Error Message (RERR), and Route Reply Message (RREP) and [19].

In cases where both the destination and source are a part of the same cluster, the CH which receivesthe RREQ message from the sender node uses its neighbor table to get the destination details. It then sends back a reply RREP message to the sender node. A list of intermediate nodes between the sender and the receiver is maintained by the RREP message. Therefore, the sender gets the path to the destination which passes through the CH. Inintra-cluster routing, multipath routing is not feasible due to the one-hop distance between the CH and the member nodes.

In cases of the source and destination being parts of different clusters, the CH that receivesthe RREQ message from the source passes it to every neighbor CH. It is then further passed to neighbor CHs. In the process, addresses are collated, and paths are established following the information logged in at the previous node. On the route, links are set up using the route information of the previous path. The Cluster Head that gets the RREQ message looks for the destination address in its parent cluster. Once any CH finds it, the complete path is formed to the destination and stored. Finally, the RREP data packet at the destination CH uses the reverse path established to reach the source CH. This route goes through the previous neighboring CH on the way back.

In situations where the RREQ packet gets lost or dropped at some point in the network, the route discovery is redone a fixed number of times (depending on the network conditions). If it is still unsuccessful, the process is not repeated and the destination is stored as unreachable [2, 4, 54]. The steps involved used in finding multiple paths and selection of optimal paths during route discovery are described in detail below

In a fixed time period, all the RREPs received are categorized into either shortest or optimal paths. The shortest paths are those which have the minimum number of hop counts. The optimal paths are generated using a weighted metric-based on energy value and mobility coefficient as described by Eqs. (1) and (2).

Parameters (P1 and P2) are assigned appropriate weights that are application-dependent. In applications where more energy saving is required, weight (w1) assigned to energy parameter, (P1) is given higher value, e.g., in military applications. In applications supporting high mobility node e.g., vehicular applications, the weight (w2) assigned to the mobility parameter (P2) is assigned higher values. In totality, the summation of w1 and w2 must be 1.

The total weight of any path is given by the sum of all nodes’ weights lying on that path. As mentioned below, the weight W of path p_i, denoted by W(p_i) is defined by Eq. (12) below:

All the paths having weight, W(p_i) above the defined threshold are accepted, and the paths not satisfying a particular predefined threshold for the path weights are rejected. Hence, the selected paths are those paths which satisfy the following weight condition.

Let the weight of the path with maximum weight value be denoted as Max(W(p_i).The path p_i with weight w(p_i) is selected only if it satisfies the Eq. (13):

All the other received paths are rejected. The various steps for the selection of optimal paths are shown in a flow diagram in Fig. 5.

The primary aim of the proposed method is to increase thesecurity of MANETs by implementing signature generation and cryptographic mechanisms. The working process of the proposed Secure Cryptography-based Clustering Mechanism (SCCM) can be described briefly in the steps below:

Once the cluster members register themselves with the CH, the key generation and sharing processes are performed in each cluster. When the data transmission request is initiated by the source towards the destination, a secure route is established between them. It uses the proposed weighted multipath routing protocol as the base routing technique.

Before transmission, the original packet is encrypted using the Elliptic Curve Cryptography (ECC) technique. ECC belongs to the category of asymmetric public key encryption that generates smaller and more efficient keys very fast compared to other techniques. It is based on the elliptic curve theory that uses the properties of elliptic curve equation. ECC creates keys by simply generating a random integer in a certain range though securely. These can be used as private keys. The ECC points i.e., pairs of integers coordinates lying on the curve are used as public keys.

Next, the signature is produced for the encrypted data using Schnorr’s signature generation technique. Schnorr’s digital signature is known for its simplicity. It is based on Schnorr’s algorithm and its security is built on the property of un-traceability of discrete logarithmic problem. The ECC encrypted and Schnorr’s signed data packet is sent to the destination through the respective Cluster Heads. The destination node on accepting the packet, regenerates the signature for decrypting the original data. During this process, the received packet is verified for correctness. The destination node discards the received packet, and transmits the error report to its CH. Thus, the proposed scheme ensures message confidentiality, integrity and authentication.

Figure 6 below illustrates the flow of various steps through which the proposed protocol works:

The WMECS protocol is used to perform the multipath routing among the source node and destination node, where multiple paths are selected during data transmission. This provides an alternate path if there is any fault or failure on the selected path. It also reduces the data loss and delay in the network by using multiple paths.

After establishing the secured path between the source and destination, the Elliptic Curve Cryptography technique is utilized to encrypt the data to be sent. ECC offers fast computation and reduced resource consumption for cryptographic operations at minimum cost. The power of the ECC algorithm is fully based on the key and the alphabetical table. It also gives a possible outcome for the data by enabling the secure transmission of keys between the communication entities. Different characteristics are symbolized in this technique as the coordinates of the curves. Because of its capability to generate the complex encrypted data that is difficult to decipher by unauthorized user, ECC has been selected for the proposed technique.

After data encryption, the Schnorr’s signature generation algorithm is used for the generation of the signature for the encrypted data. It is a kind of key generation mechanism that integrates both digital signature and public key encryption schemes. It analyzes the discrete logarithmic problem for generating the digital signature, which increases the security of the network. The signature generation process has the following steps:

In this technique, the source verifies the public key of the packet using the digital certificate. The keys for generating the cipher text are generated using a random integer. The one-way key hash function is deployed in the mechanism to create an encrypted text which is transmitted to the destination with the generated signature. The one-way keyed hash function alters input messages of several lengths into output series of fixed length, called a hash value, which is usually shorter than the input length. Hash values are often used to spot input sequences, i.e., to allot to them some distinctive values that illustrate them.

The network performance is analyzed for SCCM and compared with other existing non-ECC secure protocols, namely PRIME, CORMAN and F3TM in the attackers’ presence in the network.

This signifies the approach’s efficiency in the presence of malicious nodes. It is measured in frames/packets. It is calculated based on link maintenance, discovery of nodes and latency. Typically, the generated control packets and received data packets are helpful in finding the overhead ratio of the network. Based on this, the presence of theattacker in the network is identified and blocked as the attacker doesn’t have the destination signature.

Figure 7 illustrates the control packet overhead of existing PRIME, CORMAN, F3TM and proposed SCCM protocols with respect to the amount of attackers. When likened to existing techniques, the proposed SCCM efficiently decreased the control packet overhead by transmitting the message through CH and use of an efficient security mechanism.

Figures 8 and 9 illustrate the PDR of the existing and proposed protocols with respect to the amount of nodes and attackers.

From the above, we noted that if the amount of attackers in the network increase, the PDR of the network decreases. When likened to existing techniques, the proposed SCCM shows better PDR by efficiently blocking the malicious nodes in the network.

It can be seen from the above that with increase in the amount of nodes, PDR stays almost constant for all the protocols. It is maximum for SCCM as it uses multiple paths for forwarding the packets. So, in case there is any failure, it uses an alternate path for further communication, which improves the PDR.

Figures 10 and 11 illustrate the average end-to-end delay of the proposed and existing protocols in terms of the amount of nodes and attackers. In this evaluation, it is observed that the proposed SCCM provides the lowestend-to-end delay, when likened to existing techniques. This is due to the use of efficient CH, multiple paths and lightweight security for transmission in the proposed scheme.

Figure 12 illustrates the performance of the proposed SSCM technique. It has an improved throughput in the attackers’ presence, when compared to the other techniques. This is due to the combined use of clustering, multipath routing, and efficient security that makes the performance better.

Protocols are analyzed for performance with respect to throughput and delay with increase in number of attacker nodes.

Figure 13 describes the average end-to-end delay of both the existing and proposed methods with respect to the amount of attacker nodes. With increase in the amount of malicious nodes, the packet starts getting delays. The result reveals that the proposed SCCM provides less end-to-end delay compared to the other methods due to the use of efficient security and routing.

In this analysis, we proved that the suggested SSCM technique has a better throughput when likened to the other techniques. This is due to the combined use of clustering, multipath routing and efficient security.

Figure 14 describes the impact of increase in the amount of attackers on throughput. With increase in theamount of malicious nodes, the throughput starts decreasing for all as packets are dropped. The decrease is lowest in the proposed SCCM technique when compared to the other techniques due to the use of a pool of CHs and multipath routing with an efficient security mechanism.