Jammer localization in wireless networks: An experimentation-driven approach

Abstract

Jamming attacks have become prevalent during the last few years facilitated by the open access to the shared wireless medium as well as the increased motivation and easiness to create damage as a result of sophistication of wireless devices, both legitimate and jamming ones. Among the challenges that a wireless network faces while trying to confront the jammer, jammer localization is of utmost importance. This entails estimating the physical location of the jammer. Successful jammer localization can trigger a series of corrective measures to ensure sustainable network operation. However, locating the jammer is a difficult problem. Our primary goal in this paper is to design a simple, lightweight and generic approach for localizing a jamming device through a set of measurable parameters. The key observation guiding our design, is that the Packet Delivery Ratio (PDR) that can be readily measured locally by a device decreases as a receiver moves closer to the jammer. Further, we draw on the gradient-descent principle from optimization theory, and we adapt it to operate on the discrete plane of the network topology so that the jamming device location can be estimated. The very nature of the gradient-descent algorithm allows the distributed execution of our localization scheme. In this paper, we compute and experimentally validate the impact of jammer on the PDR of a link and we show that this impact decreases as the link moves away from the jammer. We further design a distributed, lightweight jammer localization system, which does not require any modifications to the driver/firmware of commercial NICs, while we implement a prototype system to evaluate our scheme on our 802.11 indoor testbed. Finally, we evaluate the performance of our system via extensive simulations in larger scale settings. Its performance in terms of average location estimation error in combination with its simplicity and distributed operations hold great promise.