Connected device region identification
Various aspects of methods, systems, and use cases include region identification of a malicious device based on crowdsourced locations. A method may include generating a grid of bins of a local radio coverage region, receiving location data from a plurality of devices in the local radio coverage region, and classifying locations of the plurality of devices with respect to the bins. The method may include associating the classified locations of the plurality of devices to the received location data for corresponding devices of the plurality of devices, and generating a model, from the associated classified locations and the received location data.
1 . An edge device for region identification of a malicious device based on crowdsourced locations, the edge device comprising:
processing circuitry; and
memory including instructions, which when executed by the processing circuitry, cause the processing circuitry to perform operations including:
receiving location data from a plurality of devices in a local radio coverage region;
classifying locations of each of the plurality of devices;
associating the classified locations of each of the plurality of devices to the received location data for corresponding devices of the plurality of devices;
generating a model, from the associated classified locations and the received location data;
receiving, from a device, a received signal including a reported location of the device;
determining an actual location of the device using the model, wherein the model is a neural network;
updating the neural network based on neural network information received from a second edge device corresponding to a second device location generated at the second edge device; and
outputting, from the model, an indication that the device is malicious based on a mismatch between the reported location and the actual location.
2 . The edge device of claim 1 , wherein the local radio coverage region includes a grid of bins and each bin in the grid of bins is one square meter in size.
3 . The edge device of claim 1 , wherein the local radio coverage region includes a grid of bins and each bin in the grid of bins has a respective unique bin identifier, and wherein the model is to output a unique bin identifier corresponding to the received signal.
4 . The edge device of claim 1 , wherein the local radio coverage region includes a two-dimensional grid of bins.
5 . The edge device of claim 1 , wherein the plurality of devices are devices that are trusted or known to the edge device.
6 . The edge device of claim 1 , wherein receiving the location data includes passively and periodically receiving the location data.
7 . The edge device of claim 1 , wherein receiving the location data includes actively receiving the location data on-demand based on a request.
8 . The edge device of claim 1 , wherein the location data includes reported cartesian coordinates of a respective device of the plurality of devices.
9 . The edge device of claim 1 , wherein the location data includes at least one of a timestamp of sent data, a received signal strength indicator (RSSI), a Reference Signal Received Power (RSRP), a Reference Signal Received Quality (RSRQ), a Power Headroom Report (PHR), or an interference level of a respective device of the plurality of devices.
10 . The edge device of claim 9 , wherein associating the classified locations includes associating the classified locations with a RSSI, RSRQ, or RSRP received from each device.
11 . The edge device of claim 1 , wherein classifying locations of each of the plurality of devices includes using hypothesis testing including testing what bin of a grid of bins a location is in using a M-ary test where a value of M in the M-ary test corresponds to a number of bins in a column of the grid of bins.
12 . The edge device of claim 1 , wherein the local radio coverage region includes a grid of bins and wherein associating the classified locations includes using a trained neural network to map multipath channel gains with bins of the grid of bins.
13 . The edge device of claim 1 , wherein the model includes a stochastic mapping of channel statistics to bins.
14 . The edge device of claim 1 , further comprising operations including using the model to determine bin location, in a grid of bins, of the received signal, wherein the received signal is a potentially malicious signal.
15 . An apparatus for region identification of a malicious device based on crowdsourced locations, the apparatus comprising:
means for receiving location data from a plurality of devices in a local radio coverage region;
means for classifying locations of each of the plurality of devices;
means for associating the classified locations of each of the plurality of devices to the received location data for corresponding devices of the plurality of devices;
means for generating a model, from the associated classified locations and the received location data; and
means for receiving, from a device, a received signal including a reported location of the device;
means for determining an actual location of the device using the model, wherein the model is a neural network;
means for updating the neural network based on neural network information received from a second edge device corresponding to a second device location generated at the second edge device; and
means for outputting, from the model, an indication that the device is malicious based on a mismatch between the reported location and the actual location.
16 . The apparatus of claim 15 , wherein the plurality of devices are devices that are trusted or known to the apparatus.
17 . The apparatus of claim 15 , wherein the location data includes at least one of a timestamp of sent data, a received signal strength indicator (RSSI), a Reference Signal Received Power (RSRP), a Reference Signal Received Quality (RSRQ), a Power Headroom Report (PHR), or an interference level of a respective device of the plurality of devices.
18 . The apparatus of claim 17 , wherein the means for associating the classified locations include means for associating the classified locations with a RSSI, RSRQ, or RSRP received from each device.
19 . A method for region identification of a malicious device based on crowdsourced locations, the method comprising:
receiving, at an edge device, location data from a plurality of devices in a local radio coverage region;
classifying locations of each of the plurality of devices;
associating the classified locations of each of the plurality of devices to the received location data for corresponding devices of the plurality of devices;
generating a model for the edge device from the associated classified locations and the received location data; and
receiving, from a device, a received signal including a reported location of the device;
determining an actual location of the device using the model, wherein the model is a neural network;
updating the neural network based on neural network information received from a second edge device corresponding to a second device location generated at the second edge device; and
outputting, from the model, an indication that the device is malicious based on a mismatch between the reported location and the actual location.
20 . At least one non-transitory machine-readable medium including instructions for region identification of a malicious device based on crowdsourced locations, which when executed by processing circuitry, cause the processing circuitry to perform operations comprising:
receiving location data from a plurality of devices in a local radio coverage region;
classifying locations of each of the plurality of devices;
associating the classified locations of each of the plurality of devices to the received location data for corresponding devices of the plurality of devices;
generating a model, from the associated classified locations and the received location data; and
receiving, from a device, a received signal including a reported location of the device;
determining an actual location of the device using the model, wherein the model is a neural network;
updating the neural network based on neural network information received from a second edge device corresponding to a second device location generated at the second edge device; and
outputting, from the model, an indication that the device is malicious based on a mismatch between the reported location and the actual location.