Search for a command to run...
This work aspires to provide a trustworthy solution for target localization in adverse environments, where malicious nodes, capable of manipulating distance measurements (i.e., performing spoofing attacks), are present, thus hindering accurate localization. Besides localization, its other goal is to identify (detect) which of the nodes participating in the process are malicious. This problem becomes extremely important with the forthcoming expansion of IoT and smart cities applications, that depend on accurate localization, and the presence of malicious attackers can represent serious security threats if not taken into consideration. This is the case with most existing localization systems which makes them highly vulnerable to spoofing attacks. In addition, existing methods that are intended for adversarial settings consider very specific settings or require additional knowledge about the system model, making them only partially secure. Therefore, this work proposes a novel voting scheme based on clustering and weighted central mass to securely solve the localization problem and detect attackers. The proposed solution has two main phases: 1) Choosing a cluster of suitable points of interest by taking advantage of the problem geometry to assigning votes in order to localize the target, and 2) Attacker detection by exploiting the location estimate and basic statistics. The proposed method is assessed in terms of localization accuracy, success in attacker detection, and computational complexity in different settings. Computer simulations and real-world experiments corroborate the effectiveness of the proposed scheme compared to state-of-the-art methods, showing that it can accomplish an error reduction of <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" id="m1"> <mml:mrow> <mml:mn>30</mml:mn> <mml:mtext> </mml:mtext> <mml:mi>%</mml:mi> </mml:mrow> </mml:math> and is capable of achieving almost perfect attacker detection rate when the ratio between attacker intensity and noise standard deviation is significant.