AIGSA-based multi-jammer localization in wireless networks

Xianglin Wei

doi:10.1016/j.asoc.2021.107131

2021-05-01

PRE

Xianglin Wei

Tongxiang Wang *

DOI：10.1016/j.asoc.2021.107131

原文链接

原文求助

摘要

En 中文

Jamming attacks, which exploit the shared and openness nature of the wireless mediums, are one of the major threats to wireless networks since they may significantly degrade the performance of or even corrupt wireless networks by continuously preventing wireless channels from serving intended users. To facilitate conducting anti-jamming countermeasures, a few single jammer localization algorithms have been introduced and evaluated. In contrast, only few efforts have been made for locating multiple jammers, for instance X-ray and Cluster algorithms. They are developed based on the assumption that a jammer's affected area is roughly a circle; however, their localization accuracy is unsatisfactory if multiple jammers' affected areas are overlapped. In this backdrop, a Multi-jammer Localization Algorithm based on Alternating Iteration and Gravitational Search Algorithm (MLA-AIGSA) is put forward in this paper. On the contrary to existing proposals, MLA-AIGSA neither relies on the mapping between the received signal power and two nodes'; mutual distance nor counts on the ideal assumption about the shape of a jammer's affected area. MLA-AIGSA contains two steps, i.e. estimating the number of jammers and Multi-jammer localization. At first, Region Growth Algorithm (RGA) is adopted to derive the number of jammers based on collected Received Jamming Signal Strength (RJSS) values. Then, Alternating Iteration (AIt) and GSA are combined to estimate and refine multiple jammers' positions in an iterative way. The overall time complexity of MLA-AIGSA is in quadratic time to the number of the particles and is in linear time to the number of AIt's and GSA's iterations. Extensive experiments are conducted to validate MLA-AIGSA's correctness and effectiveness, and the results show that MLA-AIGSA's average localization error (around 3 m) is nearly 70% lower than state-of-the-art proposals, including X-ray (17 m) and Cluster (10 m) algorithms. (C) 2021 Elsevier B.V. All rights reserved.