This document provides an overview of building a distributed sensor network called the Harbinger Sensor Network Project (HSNP) for collecting wireless traffic data and performing analytics. The HSNP has three stages: 1) creating sensor nodes using inexpensive hardware like Raspberry Pis to passively collect wireless data, 2) centralized data management to aggregate the sensor data, and 3) analytics and visualizations of the collected data. The document outlines the objectives, requirements, hardware suggestions, data schema, sensor code, and considerations for designing and deploying the sensor network grid.

1. Sensory Perception:
A DIY Approach to Building a Sensor Network

2. rbx@rbx:~# whoami
 Tim Fowler
 Prinicpal IT Security Consultant
 Mountainsec, LLC
 @roobixx

3. rbx@rbx:~# info
 Christian
 Member of WNCInfoSec & Proverbs Hackers
 OSWP
 Open Source Advocate
 Lover/Hater of all things Wifi
 Triathlete*
* A Work in Progress

4. Where it all started...

5. Background
● CreepyDOL ● Snoopy & Snoopy-
NG

6. From that came this...

8. ...and now comes this...

9. Harbinger Sensor Network
har·bin·ger
/ härb nj r/ˈ ə ə
noun
a person or thing that announces or
signals the approach of another.

10. HSNP: Objectives
 Create a distributed sensor network for data
collection and traffic flow / migratory pathway
analysis
 Create analytics engine for data processing and
visualization rendering

11. Why?

12. Because I can!

13. Why reinvent the wheel?

14. Harbinger Sensor Network Project
 Harbinger is a wireless sensor network project
for historical trend and predictive analysis
 Unlike other sensor projects, Harbinger is
completely passive on the airwaves, making it
virtually undetectable.
 Harbinger is an open source project that is
licensed under the GPL v3

15. HSNP: Requirements
 Must be open source
 Must be as hardware agnostic as possible to
keep cost down and allow for the use of
whatever platform is available.
 Keep dependencies & requirements to a
minimum
 Must scale

16. HSNP: Project Breakdown
The HSNP has three primary components which
breakdown into stages of development
 Stage 1: Sensor Node(s)
 Stage 2: Centralized Data Management
 Stage 3: Analytics and Data Visualizations

17. Stage 1: Sensor Node(s)
This is where we are at today!

18. Stage 1: Node Requirements
 Must be able to run python (2.7)
 SQLite3
 A wireless interface that supports Monitor mode
 That's it!

19. Stage 1: Hardware Suggestions
Hardware Platforms:
 Raspberry Pi A/B/B+/2
 Beaglebone Black
 TP-Link MR3040*
Wireless Adapters:
 Alfa Network AWUS036h

TP-Link tl-wn722n*

20. HSNP: Data Requirements
● Clients
– Probe Requests
– Date which client appeared in the network
– Last seen in the network
● Access Points
– Probe Responses
– Time of Responses

21. HSNP: Data Schema

22. Where to get
started?

23. “Git” the code!

24. https://github.com/roobixx/harbinger

25. HSNP: Sensor Code
# git clone https://github.com/roobixx/harbinger.git
# cd harbinger/
# ./harbinger.sh

26. HSNP: sensor.py

27. HSNP: db.py

28. HSNP: node.db

29. You have node(s)
now what?

30. HSNP: Designing the “Grid”
Do you need 100%
coverage?

31. HSNP: Designing the “Grid”
Questions to consider when deploying sensors
● Where are the choke points?
● Can a device in the grid be seen by multiple
nodes?
● What happens if a node(s) goes down??

32. HSNP: Deploying the Nodes
Depending on how expansive your coverage
area is will dictate how many nodes are
required for your grid.

33. HSNP: Deploying the Nodes
● Start my surveying the landscape!
● Use Wigle or Kismet to map all Access Points
within your desired coverage area
● You will use this data to fine tune your node
placements

34. HSNP: Node Placement
● Determine the exact area you desire to cover
● Estimate where the choke points in your grid
are at that require coverage.
● Test a sensor at each point looking for the
Access Points that are within rang of the sensor
● Cross-reference the Wigle scan to determine
your sensors effective range at current
placement.

35. HSNP: Effective Range of Detection
● ERD will vary from sensor to sensor even if
they are all the same hardware configuration

36. HSNP: Deploying the Nodes

37. ERD is highly variable and will change
frequently.
Constant calibration is require to maintain
accurate data within the grid

38. Stage 2: Centralized Data Management
● Not a lot to say on this at this point but
obviously being able to gather all of the sensor
data into a single location for processing is
necessary.

39. Stage 3: Analytics & Data Visualization
Coming soon...?

40. Comments or Questions?