Crowd-Sourcing Wi-Fi Coverage Data to build Self- Mapping Radio Maps TNC2013 Gareth Ayres (Speaker) Jason Jones 2013
About the Authors Gareth Ayres (me) Wireless Network Officer, Swansea University PhD Student (part-time) Author of SU1X Jason Jones Lecturer Swansea University My PhD Tutor
Introduction – Key Terms I know I don’t need this slide here, but just in case… eduroam – (education roaming) is the secure, world-wide roaming access service developed for the international research and education community. Wi-Fi – WLAN based on any standard LBS – Location Based Service – A service that makes use of the context of a devices location
Introduction – Key Terms Supplicant – Software used to authenticate a device to a 802.1x network Location Fingerprint – a record of information (signal strength in this case) derived by a device about a given location at a specific time. (Coverage Data) Radio Map – A map of information (location fingerprints) about RF properties such as signal strength or signal-noise-ratio and associations between nodes. Abstract, not geographic.
The Problem/Solution Problem: Poor wireless coverage maps LBS’s poor as no location data Need rough AP locations at least Solution: Build self-mapping radio maps using crowd-sourced data Not brilliant, but a first step and better than present
Graph (or radio map) Graph theory is the study of graphs, which are mathematical structures used to model pairwise relations between objects. - wikipediagraphs Vertices – nodes, in this case wireless access points Edges – An association between nodes (not that type!) Weights – Used to describe proximity between associated nodes/vertices in this case
Simple Graph
Introduction – Setting the Scene Wi-Fi at Swansea University: CISCO WLAN controllers >1000 Wireless Access Points ~3000 staff, ~10000 students ~8000 users a day, ~6000 consecutively connected devices eduroam primary ssid Use SU1X for Windows configuration
Introduction – SU1X Supplicant Configuration Tool: SU1X – Open Source, Windows XPSP3->Win8 Configures a wireless profile Sets EAP credentials Like eduroamCAT…
Introduction - LocPriS LocPriS – A security and privacy preserving location based services development framework Paper: _60http://link.springer.com/chapter/ %2F _60 Part of my PhD project
Crowd-Sourcing Data The Idea: SU1X modified to capture location fingerprints Opt-out during installer Only runs on first-run and reauths. Location Fingerprints sent to LocPriS server HTTPS GET request PHP parses, cleans, validates data etc Fingerprint stored in MySQL DB
Typical Location Fingerprint Location Fingerprint: SSID, BSSID, ID, RSS, TYPE, TIME March 14 th 2012: rows from 172 devices eduroam AAAA g :38:52 eduroam FFFF g :38:52 eduroam BBBB g :38:52
Example Location Fingerprinting Two Devices: MN1 & MN2 Three Buildings, A, B & C 7 Access Points, A1..C3 (Potentially multiple BSSIDs but ignore this for now as we use eduroam)
Example Location Fingerprinting Two Location Fingerprints: MN1: [{A1, 80},{A2,46},{C1,72}] MN2: [{B2, 52},{C2,59},{C3,49}]
Example Location Fingerprinting E(Vx,Vy) = {MNi,Vx} + {MNi,Vy} Where: E(Vx,Vy)=Edge weight between Vx,Vy MNi = Mobile Node i Vx,Vy= Two Wi-Fi Access Points (Vertices)
Example Location Fingerprinting E(x,y) = {MN1,A1} + {MN1,A2} E(A1,A2)=80+46=126 E(B2,C3)=52+49=101 E(A2,C1)=72+46=118 E(B2,C2)=52+59=111 E(A1,C1)=80+72=152 E(C2,C3)=59+49=108
Force-Directed Layouts Force-Directed Graph Draws graphs according to an algorithm of physical system Forces set between edges (weights) Attract and Repel like springs Graph visualisation Using Gephi for visualisation in the following slides Moving to DS3.js for web based visualisation More control over force-directed layout algorithm
Data from March 11 th – 12 th Vertices Edges 167 Devices
Data from March 11 th - 13 th Vertices Edges 302 Devices
Data from March 11 th - 14 th Vertices Edges 461Devices
Data from March 11 th - 15 th Vertices Edges 633 Devices
Data from March 11 th - 16 th Vertices Edges 724 Devices
Data from March 11 th - 17 th Vertices Edges 856 Devices
Augmented Map and Anchors Radio map lacks orientation, position and scale in real world. Also suffers from flipping. Solutions is Anchor Nodes to produce an Augmented Radio Map (ARM): 1.Users manually locate themselves on map 2.GPS outside to give building periphery AP’s –Gathered from eduroam apps on smart phones?
Sharing and Distribution of Radio Maps Graph Exchange XML Format (GEXF) Use standard xml techniques Share and combine graphs/radio maps
Future Work Explorer/Improve new Force-Directed layout algorithms Different sources of fingerprints: RADIUS authentication logs Association logs D3.js visualisation of graphs
End Questions? Gareth Ayres