Elsevier

Computer Communications

Volume 163, 1 November 2020, Pages 162-175
Computer Communications

Prevention of Hello Flood Attack in IoT using combination of Deep Learning with Improved Rider Optimization Algorithm

https://doi.org/10.1016/j.comcom.2020.03.031Get rights and content

Abstract

IoT are prone to vulnerabilities as a result of a lack of centralized management, dynamic topologies, and predefined boundary. There are diverse attacks that affect the performance of IoT network. Flooding attack is a DoS attack that brings down the network by flooding with a huge count of HELLO packets, routed to a destination that does not exist. The main intent of this paper is to develop a novel robust model for detecting and preventing HELLO flooding attacks using optimized deep learning approach. In this proposed research model, the steps like Cluster head selection, k-paths generation, HELLO flooding attack detection and prevention, and optimal shortest path selection are employed. Once after the random cluster head selection and k-paths generation, few Route Discovery Frequency Vectors like Route Discover Time and Inter Route Discovery Time of each node is determined for detecting the HELLO flooding attack. Initially, a threshold function is used to match with the computed Received Signal Strength (RSS) of each node to detect the stranger node. Further, the HELLO flood attack is confirmed by the optimized Deep Belief Network (DBN), which is removed from the network subsequently. Once after securing the network, the shortest route path selection is done optimally by the improved meta-heuristic algorithm. Here, improved Rider Optimization Algorithm (ROA) termed as Bypass-Linked Attacker Update-based ROA (BAU-ROA) is used for performing the optimal DBN as well as optimal shortest path selection. The objective constraints like node trust, distance between the nodes, delay of transmission, and packet loss ratio are considered for performing the optimal shortest path selection. Finally, the experimental evaluation of various performance measures validates the fruitful performance of the proposed model. Based on the analysis, the const function of proposed BAU-ROA is 28.1% superior to D-DHOA, 34.1% superior to DHOA, and 39.2% superior to WOA at 5th iteration. When considering the 10th iteration, the developed BAU-ROA is 12.2% superior to DHOA, 19.3% superior to D-DHOA, 28.1% superior to ROA, and 39.2% superior to WOA.

Introduction

IoT devices are used for collecting the information and identify hazards successive to tragedies and also for localizing the affected people [1]. However, IoT is not holding up the existence of tragedy, which could be a very important thing for transferring catastrophe readiness with each other by active information, for example, the prediction of disaster and also caution to the systems [2], etc.

Flooding attack is a specific kind of DoS attack existing in MANETs, in which the malicious nodes imitate the legitimate nodes in all situations except the paths discovered by them. The discovered paths are usually intending to use the processing resources of the remaining nodes. This kind of attack is very easy to execute using on-demand routing protocols such as AODV [3], [4]. Some of the flooding attacks are HELLO, DATA, and RREQ. The most dangerous flooding attack is RREQ since it is very simple to generate a disaster of RREQ packets and lead huge damage. The HELLO flooding attack is related to the network layer attack [5], which targets the routing protocols. The protocols require the nodes used for communicating the HELLO packets for reporting to the neighbours they are existing. The node is considered as the local node when a node receives the data packet from another node within its range.

Many research studies have been introduced in the field of RREQ flooding attacks, which significantly concentrates on detection models that depend on the frequency of RREQ packets for sending the information [6]. In order to detect an attack, each node makes use of constant or dynamic threshold value. This threshold value is computed on the basis of the count of RREQs derived by the node for each unit time. If the node receives more RREQ, then that neighbour node is denoted a as malicious node. The above-mentioned models have some defects in dealing with the dynamics of networks. If the flooding attack is well known in wired networks [7], [8], [9], the solutions are not directly implemented for securing mobile DTN, which is having transient connectivity and not accessible at any time. Several solutions are developed for preventing the flooding attack existing in DTN. In [10], a lightweight validation scheme was proposed. In [11], the maintenance of dropping strategy for node’s message queue. In [12], [13], the internal flooding attackers were not detected. Other contributions like [14], [15] limit the nodes’ message creation rates and the detection of internal adversaries, which is beyond the rate limits. However, the rate limit method is not reliable for networks because it is not predictable broadcast demands.

The major contributions of the paper are portrayed as follows.

  • To develop a secure IoT-WSN with several processing steps like Cluster head selection, k-paths generation, HELLO flooding attack detection and prevention, and optimal shortest path selection.

  • To detect and prevent the HELLO flood attack from the IoT-WSN using the BAU-ROA-DBN with the input feature vectors like Route Discover Time and Inter Route Discovery Time.

  • To find the shortest route paths in IoT-WSN enabled by the proposed BAU-ROA approach by considering the objective constraints like node trust, distance between the nodes, delay of transmission, and packet loss ratio.

  • To validate the performance of the proposed and conventional models in terms of convergence analysis, and analysis on latency, normalized energy, and shortest path length.

The entire paper is modelled as shown below: Literature review and the features and challenges of existing flooding attack detection and prevention models are discussed in Section 2. Section 3 specifies the detection and prevention of hello flood attack in medical Internet of Things. Detection and prevention of hello flood attacked in medical IoT are mentioned in Section 4. Moreover, the development of improved rider optimization algorithm is shown in Section 5. Section 6 illustrates the selection of shortest paths by proposed rider optimization algorithms with the computation of energy dissipation. Results and discussions are mentioned in Section 7. The final conclusions of the paper are shown in Section 8.

Section snippets

Related works

In 2019, Luong et al. [16] have suggested a new FADA for MANETs on the basis of the machine learning algorithm. The method was dependent on the data of each node’s route discovery history for capturing the equivalent behaviours and features of nodes that belonged to a similar class for deciding whether the node was malicious. Moreover, FAPRP was suggested by expanding the actual AODV protocol and combining FADA algorithm. The performance analysis of the suggested model was done based on packet

System model

Consider the medical IoT-WSN system with huge numbers of sensor nodes denoted as je. The data from each and every sensor node is transmitted to a single sink node or base station represented as BS. A direct data transmission is carried out between the nodes with wireless connection, in which the communication performs within the radio range. Every node in the network takes of the role of data packet transmission, which will be under the network with dimension X and Y in metres. Clusters of

Creation of k-paths

From the IoT-WSN, the source node ns and destination node nd has to be chosen after selecting the cluster head, which is done by a graph structure. By this way, the user can choose the source node and destination node along with ID. Moreover, the shortest path of the network is selected from the generated k-paths. This k-path routing procedure is based on the correlation of finding the minimum count of possible bandwidth or shortest path between two cluster heads in the transmission fields that

Conventional ROA

The inspiration of ROA [22], [23] is based on the cluster of riders that is going in the direction of destination. Consider some of the rider’s group, which are going towards the same target for winning the race. In the conventional ROA, there are four groups of riders such as “bypass riders, followers, overtakers, and attackers”. This algorithm is following some steps that are described below and should be followed by every rider in the group.

Parameter Initialization of rider and group:

Solution encoding

The proposed BAU-ROA is employed for defining the shortest path that chooses the best among the cluster heads from IoT-WSN model after preventing the HELLO flood attack. The solution pattern before encoding is depicted in Fig. 4. Here, the cluster head is indicated by CHz, and the count of clusters is denoted as CH. The cluster heads that are in progress is considered for choosing the shortest route path after detecting and preventing HELLO flooding attack. Let, CHz1 is the source node, and CHz

Experimental setup

The developed shortest route path selection with detection and prevention of HELLO flood attack was implemented in MATLAB 2018a, and the analysis was carried out. For performing the experimental analysis, the number of rounds was considered as 800, the number of nodes allotted as cluster heads was 35, and the initial energy of each node was kept constant as 0.02 Joules. The performance analysis of the suggested approach was compared with state-of-the-art WOA [25], DA [26], DHOA [27], ROA [22],

Conclusion

This paper has introduced a novel algorithm for detecting and preventing HELLO flooding attack by optimized deep learning mechanism. This model included cluster head selection; k-paths generation, detection, and prevention of HELLO flooding attack, and optimal shortest path selection phases. After selecting the cluster head, and k-paths, some Route Discovery Frequency Vectors such as Inter Route Discovery Time and Route Discovery Time of every node was defined in order to detect and prevent

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

References (28)

  • Madhura Mahajan, Dr. K.T.V. Reddy, Manita Rajput, Design and simulation of a blacklisting technique for detection of...
  • ZargarS.T. et al.

    A survey of defense mechanisms against distributed denial of service (DDoS) flooding attacks

    IEEE Commun. Surv. Tutor.

    (2013)
  • BhuyanM.H. et al.

    Detecting distributed denial of service attacks: Methods, tools and future directions

    Comput. J.

    (2014)
  • KumarSunil et al.

    Direct trust-based security scheme for RREQ flooding attack in mobile ad hoc networks

    Int. J. Electron.

    (2017)
  • Cited by (0)

    View full text