SDR4IoT an Experiment Part of Fed4FIRE+ in w-iLab.t Testbeds - - PowerPoint PPT Presentation

▶
sdr4iot
SMART_READER_LITE
LIVE PREVIEW

SDR4IoT an Experiment Part of Fed4FIRE+ in w-iLab.t Testbeds - - PowerPoint PPT Presentation

Mar 31, 2023 136 likes •541 views

SDR4IoT an Experiment Part of Fed4FIRE+ in w-iLab.t Testbeds @alexis0duque #FOSDEM #FSR #SDR #IoT #ML @alexis0duque ` Who am I? Alexis DUQUE Director of Research & Development @alexis0duque alexisduque alexisd@rtone.fr

slide-1
SLIDE 1

@alexis0duque #FOSDEM #FSR #SDR #IoT #ML

SDR4IoT

… an Experiment Part of Fed4FIRE+ in w-iLab.t Testbeds

@alexis0duque

`

slide-2
SLIDE 2

@alexis0duque #FOSDEM #FSR #SDR #IoT #ML

Alexis DUQUE Director of Research & Development

  • @alexis0duque
  • alexisduque
  • alexisd@rtone.fr
  • alexisduque.me
  • https://goo.gl/oNUWu6

Who am I?

slide-3
SLIDE 3

@alexis0duque #FOSDEM #FSR #SDR #IoT #ML

Outline

  • Motivation
  • Fed4FIRE+ H2020 project
  • Wireless testbeds w-iLab.t at iMEC
  • Accessing the testbed
  • Our experiment
  • Further Work
slide-4
SLIDE 4

@alexis0duque #FOSDEM #FSR #SDR #IoT #ML

Motivation

SDR hardware is popularizing Software library are maturing Lot of interest and work in academia So, lets embed & use SDR4IoT! 😅 Passive Authentication Localization Fingerprinting

slide-5
SLIDE 5

@alexis0duque #FOSDEM #FSR #SDR #IoT #ML

Motivation

slide-6
SLIDE 6

@alexis0duque #FOSDEM #FSR #SDR #IoT #ML

State of the Art

  • R. C. Bunescu, “Deep Learning Convolutional Neural

Networks for Radio Identification.”, 2018.

  • S. Rajendran, W. Meert, D. Giustiniano, V. Lenders, and S. Pollin,

“Deep Learning Models for Wireless Signal Classification with Distributed Low-Cost Spectrum Sensors.”, 2018.

  • C. Morin, L. S. Cardoso, J. Hoydis, J.-M. Gorce, and T. Vial,

“Transmitter Classification with Supervised Deep Learning,” in CROWNCOM’19, 2019.

slide-7
SLIDE 7

@alexis0duque #FOSDEM #FSR #SDR #IoT #ML

SoA Limitation

Not real-world IoT device Not conventional communication protocols Not reproducible Too small dataset: few devices, few RF traces, ... Nodes position do not change, ML doesn’t learn localization

slide-8
SLIDE 8

@alexis0duque #FOSDEM #FSR #SDR #IoT #ML

Our Idea

Use off the shelf emitter form true IoT nodes Widely used RF protocols in 2.4 GHz ISM band SDR-based receiver Collect and share a large dataset and reproducible RF fingerprints Further rely on Machine Learning for authentication and localization But we need extensive experiments -> large testbed

slide-9
SLIDE 9

@alexis0duque #FOSDEM #FSR #SDR #IoT #ML

Fed4FIRE+

Fed4FIRE+ is a project under the EU pogramme H2020,

  • ffering the largest federation worldwide of Next Generation

Internet (NGI) testbeds, which provide open, accessible and reliable facilities supporting a wide variety of different research and innovation communities and initiatives in Europe, including the 5G PPP projects and initiatives.

slide-10
SLIDE 10

@alexis0duque #FOSDEM #FSR #SDR #IoT #ML

Fed4FIRE+ SME Open Call

slide-11
SLIDE 11

@alexis0duque #FOSDEM #FSR #SDR #IoT #ML

slide-12
SLIDE 12

@alexis0duque #FOSDEM #FSR #SDR #IoT #ML

IMEC w-iLab.t testbeds

slide-13
SLIDE 13

@alexis0duque #FOSDEM #FSR #SDR #IoT #ML

2 testbed locations http://doc.ilabt.imec.be

  • w-iLab.1: datacenter with 44 embedded PCs (WiFi and

sensor nodes)

  • w-iLab.2: industrial room with 100 fixed +15 mobile nodes

(WiFi, sensor, LTE, SDR) Use cases: wireless, sensor, mobile, networking, SDR, 3rd party hardware

IMEC w-iLab.t testbeds

slide-14
SLIDE 14

@alexis0duque #FOSDEM #FSR #SDR #IoT #ML

IMEC w-iLab.t testbeds - IoT Devices

Zolertia Re-Mote

  • 2.4GHz / 868MHz
  • UWB-shield

Zolertia Z1 RM090 Nordic Semiconductor nRF52 DK

slide-15
SLIDE 15

@alexis0duque #FOSDEM #FSR #SDR #IoT #ML

IMEC w-iLab.t testbeds - SDR

USRP B210 (x4) & B200 (x4) USRP N210 (x6) USRP x310 (x2) Xilinx ZedBoard Zync-7000 SoC (x1) ZC706 with Zync-7000 SoC (x3)

slide-16
SLIDE 16

@alexis0duque #FOSDEM #FSR #SDR #IoT #ML

IMEC w-iLab.2 testbed

slide-17
SLIDE 17

@alexis0duque #FOSDEM #FSR #SDR #IoT #ML

IMEC w-iLab.2 testbed

slide-18
SLIDE 18

@alexis0duque #FOSDEM #FSR #SDR #IoT #ML

Our Experiment - Access Testbed

First Time : Request an account at https://authority.ilabt.iminds.be/signup.php

slide-19
SLIDE 19

@alexis0duque #FOSDEM #FSR #SDR #IoT #ML

slide-20
SLIDE 20

@alexis0duque #FOSDEM #FSR #SDR #IoT #ML

Our Experiment - Access Testbed

First Time : Request an account at https://authority.ilabt.iminds.be/signup.php

  • 1. Browse and reserve resources at

http://inventory.wilab2.ilabt.iminds.be/.

slide-21
SLIDE 21

@alexis0duque #FOSDEM #FSR #SDR #IoT #ML

slide-22
SLIDE 22

@alexis0duque #FOSDEM #FSR #SDR #IoT #ML

Our Experiment - Access Testbed

First Time : Request an account at https://authority.ilabt.iminds.be/signup.php

  • 1. Browse and reserve resources at

http://inventory.wilab2.ilabt.iminds.be/.

  • 2. Install jFed, start up jFed and login through the iLab.t
slide-23
SLIDE 23

@alexis0duque #FOSDEM #FSR #SDR #IoT #ML

Our Experiment - jFed

jFed Experimenter GUI and CLI allow end-users to provision and manage experiments

https://jfed.ilabt.imec.be/downloads/

slide-24
SLIDE 24

@alexis0duque #FOSDEM #FSR #SDR #IoT #ML

Our Experiment - Access Testbed

First Time : Request an account at https://authority.ilabt.iminds.be/signup.php

  • 1. Browse and reserve resources at

http://inventory.wilab2.ilabt.iminds.be/.

  • 2. Install jFed, start up jFed and login through the iLab.t
  • 3. Load/copy your experiments XML (.rspec)
  • 4. Click Run and give your experiment a name.
  • 5. Double click the nodes and you should have SSH access on the nodes.
slide-25
SLIDE 25

@alexis0duque #FOSDEM #FSR #SDR #IoT #ML

slide-26
SLIDE 26

@alexis0duque #FOSDEM #FSR #SDR #IoT #ML

Our Experiment - Emission

Huawai Nexus 6P

  • Custom Bluetooth Low Energy App
  • Advertising as a Bluetooth HRM Peripheral
  • 2.4 GHZ GFSK
slide-27
SLIDE 27

@alexis0duque #FOSDEM #FSR #SDR #IoT #ML

Our Experiment - Reception

USRP N210

  • GNU Radio Compagnon
  • Output Raw I/Q data (PHY)
  • Decode Bluetooth Low Energy Packets

rely on ble_dump1 by drtyhlpr

[1] https://github.com/drtyhlpr/ble_dump

slide-28
SLIDE 28

@alexis0duque #FOSDEM #FSR #SDR #IoT #ML

slide-29
SLIDE 29

@alexis0duque #FOSDEM #FSR #SDR #IoT #ML

Our Experiment - Scenario

Receiver (USRP) don’t move, emitters (smartphone) can move One emitters sequentially

  • same position
  • same receiver
  • different receivers
  • different positions

N emitters at the same time

slide-30
SLIDE 30

@alexis0duque #FOSDEM #FSR #SDR #IoT #ML

slide-31
SLIDE 31

@alexis0duque #FOSDEM #FSR #SDR #IoT #ML

slide-32
SLIDE 32

@alexis0duque #FOSDEM #FSR #SDR #IoT #ML

Our Experiment - Next steps

Write documentation and open-source codes, scripts, and datasets Zenodo

+ Other RF and node types + Extensive use of robot for mobility

slide-33
SLIDE 33

@alexis0duque #FOSDEM #FSR #SDR #IoT #ML

slide-34
SLIDE 34

@alexis0duque #FOSDEM #FSR #SDR #IoT #ML

Our Experiment - Next steps

Write documentation and open-source codes, scripts, and datasets Zenodo VirtualWall for data science and machine learning

  • Fingerprinting with deep learning
  • Try different deep learning algorithm
slide-35
SLIDE 35

@alexis0duque #FOSDEM #FSR #SDR #IoT #ML

Fed4FIRE+ SME Open Call

slide-36
SLIDE 36

@alexis0duque #FOSDEM #FSR #SDR #IoT #ML

Thanks!

w-iLab.t Team Brecht Vermeulen, Pieter Becue

slide-37
SLIDE 37

@alexis0duque #FOSDEM #FSR #SDR #IoT #ML

MACHINE LEARNING? TENSORFLOW LITE? INDOOR VERTICAL FARM ?

slide-38
SLIDE 38

@alexis0duque #FOSDEM #FSR #SDR #IoT #ML

References

https://www.fed4fire.eu/ https://www.fed4fire.eu/event/continuous-call-sme-cascaded-experiments/ https://jfed.ilabt.imec.be/ https://doc.ilabt.imec.be/ilabt/wilab/ https://doc.ilabt.imec.be/ilabt/wilab/tutorials/lte.html#control-an-android-sma rphone-via-adb https://doc.ilabt.imec.be/ilabt/jupyter/index.html https://zenodo.org/ https://gnuradio-eu-20.sciencesconf.org/ https://www.tensorflow.org https://rtone.eu

slide-39
SLIDE 39

@alexis0duque #FOSDEM #FSR #SDR #IoT #ML

Our Experiment - Outputs

2 Distinct Datasets Further serialized with Pickle <Y,X,NodeTx,NodeRx,Timestamp, I,Q> <Y,X,NodeTx,NodeRx,Timestamp, packet>