Improving the model of decision making about abnormal network state using a positioning system

Abstract

We have proposed to supplement the model of decision making about abnormal states of a wireless network under conditions of uncertainty by another attribute ‒ the location of wireless mobile and stationary devices in a controlled network.The method of trilateration, based on the measurement of signal strength at three points, is considered. This method has a high accuracy of determining the location of a wireless device, provided that the most accurate model of radio waves propagation is constructed. However, given the specificity of radio waves propagation inside the premise, it is rather difficult to build such a model for them. Therefore, it is proposed to use the method of radio fingerprints. This method is based on the construction of radio maps for each of the three access points, which indicates the signal level from a typical wireless device located at a certain number of reference points. We have also considered the possibility of the combined application of two methods, which will make it possible to determine the location of a wireless device even when it is outside the radio map.Experimental studies were carried out, including the creation of radio maps for a room of area 70 m² with 26 reference points. We employed three identical routers and a smartphone. During the experiment, it turned out that, depending on the orientation of the mobile device (in fact, its antenna), the measured power changes, so the radio maps were constructed based on average power for six different positions of the mobile device. It is shown that the level of the signal is almost independent of the door and window position in the room.This analysis of the principles of organizing various types of attacks on wireless networks has revealed that accounting for the position makes it possible to detect attacks of the types man in the middle and false access point that were not identified by the base model. In addition, the improved model allows determining the source of interference at the muting attack