Evaluation of Range-Free Localization Algorithms Against Node Compromise Attack

Çankaya University Journal of Science and Engineering Volume 14, No. 1 (2017) 099-124 Evaluation of Range-Free Localization Algorithms Against Node Compromise Attack Seyed Saber Banihashemian, Fazlollah Adibnia*, Mehdi Agha Sarram Department of Computer Engineering, Yazd University, Yazd, Iran, e-mail: , , * Corresponding author Abstract: Different range-free algorithms are proposed for location estimation in Wireless Sensor Networks. In these algorithms, the network is assumed to have no error and false data. This article attempts to evaluate and compare the effect of malicious data produced by node compromised attacks in some of the range-free algorithms: DV-hop, LSVM, and NN. The false data may be produced by the malicious anchor nodes or compromised sensor nodes. The resistance of these algorithms against node compromise attacks is compared. The results show that although DV-hop has less localization error compared to the two other algorithms in a normal condition, in the case of attacks LSVM has less localization error. Further, in this research work, a new criterion is proposed for studying and comparing the border problem issue in the localization algorithms. Using the simulation results from various algorithms, the outcomes have been used for comparison, where it can be considered that LSVM has better performance in the border problem compared with the other studied algorithms. Keywords: Localization, Wireless Sensor Networks, Range-free, Node compromise attack, border problem. 1. Introduction Wireless sensor networks are an important technology, and many research works have been done in this field. Recent advances in wireless communications and electronic industries have made it possible to develop multi-functional sensors with low cost and low energy consumption. These sensors are small and can communicate with each other over short distances. Inexpensive, intelligent and networked sensors have introduced new opportunities to control houses, cities and the environment around. Further, sensor networks have a wide ISSN 1309 – 6788 © 2017 Çankaya University 100 S.S. Banihashemian et al. range of functionalities in the military industry and have introduced new capabilities in identifying, investigating and other strategic applications [1]. Supplying all nodes in the network with GPS devices is infeasible due to limitations in the size of sensors, cost, and energy consumption. Sensor nodes with unknown locations can estimate their positions through localization algorithms and by means of available knowledge of distance measurements to some nodes with known locations. The nodes with the known locations are called Anchor or Beacon. These nodes can obtain their locations using the Global Positioning System or setting up with a fixed known coordinates. The nodes with unknown location are called sensor nodes, non-anchor or normal nodes and we use these terms interchangeability. The locations of normal nodes are estimated with the help of the localization algorithms. In this paper, we use normal nodes and sensor nodes interchangeability. Also, we use the term ‘nodes’ for all sensor nodes and beacons. To obtain the location coordinates of normal nodes, the localization procedure is divided into two parts. In the first part, data such as distance, connectivity, angles between nodes and also the location of anchor nodes are collected [2]. The distance between the neighbors can be measured by the Received Signal Strength [3], Time of Arrival [4] and the Time Difference of Arrival. The distance between multi-hop nodes can be measured by DV-hop [4] and DVdistance [4]. In the second part, the location of an unknown node is determined by means of the collected data. The localization systems can be categorized in different ways. They can be divided into nodecentric and infrastructure-centric [5, 6]. In the former, the sensor nodes compute their locations on their own. In the latter, infrastructure (base station) or trusted nodes identify the locations of sensor nodes. These systems can also be divided into one-hop and multi-hop. In the first approach, normal nodes are localized based on the one-hop neighbors of anchor nodes. In the second approach, distant anchor nodes are also used. Localization systems can also be categorized into range-based and range-free [7]. In the ranged-based systems, geographical distances or angles between nodes are measured in the data collection stage but in range-free systems, there is no need to obtain these measurements. For the resiliency of these algorithms against attacks, some algorithms are proposed for securing the localization algorithms [8, 9, 10, 11, 12, 15]. In the previous proposed multihop range-free algorithms such as DV-hop, LSVM and NN [4, 16, 17], the localization performance is studied and evaluated only in normal conditions. It is assumed that the error free data is used for localization and there is no attack on the network. Although many algorithms are proposed for securing the localization algorithms [8, 9, 10, 11, CUJSE 14, No. 1 (2017) 101 12, 15] and for reducing the effect of attacks on the localization process, no studies have been conducted regarding the effect of data errors caused by attacks on the range-free multihop localization algorithms mentioned above. Here, the term attack means the node compromise attack. For instance, in multihop range-free algorithms, if a malicious anchor node announces a wrong location or if a malicious sensor node increases or decreases the number of hops, it will result in false data for the use in the localization process. The initial objective of this article is to analyze the effect of false data on the localization process. This study can be used in later studies to overcome the weakness of the localization algorithms caused by the node compromise attacks. The next objective is to propose a criterion that can be used for studying the border problem issue in isotropic sensor networks. The border problem is an important issue in the localization process and is partially dealt with in [16]. The utilized method in [16] cannot be used for comparing the performance of localization algorithms in the border problem. The structure of this article has been organized as follows: in the next section, the previous related methods are reviewed. In Section 3, the problem statement is highlighted and essential criteria for the evaluation are then argued. In Section Four, results of the simulations are presented and at the end, the conclusion is discussed. We have used some notations through the paper. Table 1 explains notation used throughout this paper. TABLE 1. Used notation [𝐷𝑥 ∗ 𝐷𝑦 ] 𝐷𝑥 𝐷𝑦 𝐶𝑋𝑖 𝐶𝑌𝑖 𝑁𝑥 𝑁𝑦 (𝑥𝑎 , 𝑦𝑎 ) (𝑥′𝑎 , 𝑦′𝑎 ) ℎ(𝐴𝑖 , 𝐴𝑗 ) ℎ′(𝐴𝑖 , 𝐴𝑗 ) ℎ𝑜𝑝𝑠𝑖𝑧𝑒𝑖 𝑚 𝐾 𝐶𝑖𝑥 𝑦 𝐶𝑖 dimensions of the deployment area the length of the 𝑥 axis the length of the 𝑦 axis class 𝑖 on 𝑋 axis in LSVM and NN algorithms class 𝑖 on 𝑌 axis in LSVM and NN algorithms number of location classes on 𝑋 axis in L (...truncated)