The Research On Key Technologies Of Jamming Attacks In Wireless Sensor Networks

Holding the features of easy-to-deploy, low-cost, high-resilience andgood-scalability, Wireless sensor networks (WSNs) can been widely used in manyapplication areas, such as military affairs, environmental protection, health cares, humanlife, disaster salvage. Therefore, WSNs have attracted people’s attention from bothindustry and academia. In WSNs, the data transmitted by the sensing nodes are providedfor the demanding user, regardless of time, space and environmental changes. And thishas become one of most important ways to render the information for the human people.At the same time, however, owning to the characteristics of openness, broadcasting anddynamic vulnerability, WSNs have to be forced to face the problem of reliability andsecurity. Specifically, the interference has been always a big problem in the field ofwireless communication. Besides the unintended interference induced by thecommunication among the normal sensor nodes, the intended interference, known as thejamming attack which is launched by the attacker through emitting the radio orcomprising the normal sensing nodes, is a great threat to the security of the wirelesscommunication.In our thesis, we show the importance of studying the issues of jamming attack andthe challenges we have to face. And we survey the state-of-the-art about the jammingattacks. The issues of jamming attack can be classified into two categories: attacks andcountermeasures. On behalf of attacking, it mainly concerns about the modeling ofjamming attack and its metrics of efficiences. In the existing work, the jamming modelswere proposed from the time domain, and there is no consideration from the spacialperspective. On behalf of countermeasures, the detection and defense mechanisumshave been proposed. In the existing work, the detection methods have been studied well.However, few works have been done on the issue of active defense, especially thejammer localization. In order to resolve the shortcomings of the existing work, we studythe issues of jammer localization, jamming attack models and game-theory-based attack&countermeasures, respectively.Classical jamming attack models in WSNs in the time domain have been proposed,such as constant jammer, deceptive jammer, random jammer, and reactive jammer. Inthe application-oriented sensor networks, the location information of sensor nodes arequite important. We consider a new problem: given k jammers, how does the attackerminimize the pair-wise connectivity among the nodes in a Wireless Sensor Network(WSN)? We call this problem k-Jammer Deployment Problem (k-JDP). To the best ofour knowledge, this is the first attempt at considering the position-critical jammingattack against wireless sensor network. We mainly make three contributions. First, we prove that the decision version of k-JDP is NP-complete even in the ideal situationwhere the attacker has full knowledge of the topology information of sensor network.Second, we propose a mathematical formulation based on Integer Programming (IP)model which yields an optimal solution. Third, we present a heuristic algorithm HAJDP,and compare it with the IP model. Numerical results show that our heuristic algorithm iscomputationally efficient.In the existing work on jammer localization, the accuracy was not guaranteed, andthe computation overhead was too high. Aiming to solve such kind of shortcomings, wepropose Catch the Jammer (CJ), an efficient jammer localization scheme where sensornodes collaborate with each other to compute the coordinates of the jamming attacker.As opposed to existing localization techniques which locate jamming attacker based onthe characteristics of the received signal, CJ only requires victim nodes’ locationinformation. Nodes located at the border of the jammed area will be able to exchangetheir location information and determine the jammer’s coordinates thanks to the use ofdifferent computational geometry algorithms. Nodes first compute a convex hull for theset of victim nodes based on their coordinates. They further extract the correspondingsmallest circle that covers all nodes in the convex hull in order to achieve a goodaccuracy on the coordinates of the adversary. The correctness of our proposedlocalization scheme is verified. The simulation results show that CJ outperforms mostof the existing localization algorithms depending on the variation of the jammer’stransmission range, the sensitivity to node density, and/or the position of the jammer.Thanks to the simplicity and low-cost, the Received-Signal-Strength (RSS)-basedlocalization algorithms have been widely adopted in wireless sensor networks. However,existing RSS-based localization scheme can not be directly used in the jamming attackcircumstances. The reason is that, in the radio propagation model, the distance betweensensor nodes is determined by the value of RSS, the transmitting power and the pathloss exponent parameters, etc. Opposed to normal sensor localization, the attacker or thejammer will not collaborate with other legitimate sensor nodes such that the parametersof propagation model can not be rendered. In order to solve the problem aforementioned,we propose a range-based jammer localization scheme, in which the parameters of radiopropagation model can be cancelled. We first linearly approximate the exponentialrelationship between RSS and distance, and this makes the jammer localization beindependent on the parameters of radio propagation model. Experimental resultsindicate the feasibility and efficiency of our proposed scheme if four or more RSSmeasurements can be provided by different sensing nodes.The existing research results on issues of jamming attacks in wireless sensornetworks are mainly concerned with two difference independent parts: attack andcountermeasures. In the jamming circumstances, how do the attacker and the network (sensor nodes) interact with each other? What’s the theoritical base of the changing ofstrategies for both the attacker and the network? Is there a stable state in such a jammingattack situation? It is of great importance to resolve the questions aforementioned,because it will help us to understand the essence of network attack and correspondingcountermeasures. Based on the analysis, we formalize one attacker and the network of nsensor nodes as a (n+1) player finite non-cooperative game. The goal of sensor nodes isto maximize the utility function that captures both their mutual interference and thesecurity of the chosen path for data forwarding. At the same time, the adversary or theattacker tries to reduce the overall networks’ capacity by choosing whether to jammingor eavesdropping, or use both of these two strategies. Based on knowledge of gametheory, we conclude that there is mixed strategy Nash equilibrium (MSNE) for thisgame. In order to find such MSNE, the fictitious play learning model is introduced, andthrough the simulation, we verified its efficiency.