Transcript Document

Simulation of
Large-Scale Wireless
Ad Hoc Networks
Luiz Felipe Perrone
Cybersecurity Research Group
Institute for Security Technology Studies
7/20/2015
Cybersecurity Research Group -- ISTS
1
Acknowledments

Some of the slides you’ll see have been
recycled from other presentations.

A few of the slides you’ll see come from
presentations by Deborah Estrin (UCLA) and
David Culler (UC Berkeley).
7/20/2015
Cybersecurity Research Group -- ISTS
2
Why wireless ad hoc networks?
Customizable
Easy to deploy
No infrastructure
Good in changing environments.
Allows for node mobility.
Can be designed for self-configurability.
Can be designed for scalability.
7/20/2015
Cybersecurity Research Group -- ISTS
3
Sensor Networks: A Motivating Example
M
Intelligence,
Surveillance,
Emergency Response
M
7/20/2015
Cybersecurity Research Group -- ISTS
4
Technical Challenges
Energy constraints: no wires, no power source.
Level of dynamics: weather, terrain, RF
interference, network traffic.
Scaling: very large number of nodes complicates
protocol design.
7/20/2015
Cybersecurity Research Group -- ISTS
5
A Deeper Look into the Technical
Challenges
Protocol Stack
Application
Transport
(Yes, I did throw away a couple of
layers, but who doesn’t?)
Network
Link
Physical
7/20/2015
Cybersecurity Research Group -- ISTS
6
Physical Layer
Design communication for maximum
scalability and reliability:
Modulation (AM, FM, FSK, etc).
Use of the spectrum (FDMA, TDMA,
CDMA).
Noise, interference, multipath effects,
shadowing…
7/20/2015
Cybersecurity Research Group -- ISTS
7
Data Link Layer: Medium Access Control
(Coordinated access to a shared resource)
Power is a scarce resource (so is the RF
spectrum).
Collisions lead to wasted power (AND
wasted spectrum).
Need to impose some kind of access
discipline so as to avoid collisions.
7/20/2015
Cybersecurity Research Group -- ISTS
8
The Hidden Node Problem (MAC Layer)
Station B can hear stations A and C.
Stations A and C can’t hear each other.
How can we coordinate transmissions from A and C so as to avoid
collisions?
A
7/20/2015
B
C
Cybersecurity Research Group -- ISTS
9
The Hidden Node Problem (MAC Layer)
Station B can hear stations A and C.
Stations A and C can’t hear each other.
How can we coordinate transmissions from A and C so as to avoid
collisions?
A
B
C
Solution: RTS/CTS/DATA/ACK handshake – A sends RTS
to B, B sends CTS to A, C hears CTS and stays quiet, A sends DATA
to B, B replies to A with an ACK.
7/20/2015
Cybersecurity Research Group -- ISTS
10
The Exposed Node Problem (MAC Layer)
A
B
C
D
An exposed node is one that is in range of the transmitter, but
outside range of the receiver.
Problem: exposed nodes reduce bandwidth.
7/20/2015
Cybersecurity Research Group -- ISTS
11
The MAC Layer Challenge
Maximize throughput:


Minimize collisions.
Avoid exposed nodes.
An interesting option: schedule medium
access.
Related challenges:


Clock synchronization.
Distributed coordination for determining schedule.
7/20/2015
Cybersecurity Research Group -- ISTS
12
The Network Layer Challenge
How do we build routes dynamically?


Pro-active algorithms.
Reactive algorithms.
Will the routing protocol scale up to LARGE
networks?
Can routing adapt to changes in network traffic,
propagation conditions, etc.?
Packet forwarding costs power. Can we do routing
in a way that balances power consumption?
7/20/2015
Cybersecurity Research Group -- ISTS
13
Power Consumption Issues
Nodes may not be rechargeable.
Power conservation leads to maximum network lifetime.
Communication is orders of magnitude more power
hungry than computation (need for data compression,
data fusion?).
Actuation is orders of magnitude more power hungry
than communication.
7/20/2015
Cybersecurity Research Group -- ISTS
14
Security Issues
Desirable properties:





Confidentiality
Authenticity
Integrity
Freshness
Scalability





Availability
Accessibility
Self-organization
Non-repudiation
Flexibility
As of today, the network can be vulnerable at multiple levels:
 PHY: radio jamming.
 MAC: DoS via fake requests or schedules.
 NET: fake route advertisements (black hole attack).
 A funny but scary notion: “caveman” attacks.
7/20/2015
Cybersecurity Research Group -- ISTS
15
The ISO/OSI RF Model and Wireless
We need a good architecture for wireless ad
hoc nets, but we don’t have it yet.
Current designs do not lend themselves to
interoperability (try to plug out a layer and
plug in a new one!).
Power conservation spans multiple protocol
layers.
7/20/2015
Cybersecurity Research Group -- ISTS
16
The Need for Simulation
Protocol design has always been a tough problem.
Protocol validation and verification have always been
even tougher.
We have a complex system that defies mathematical
analysis.
This system has several components tightly interconnected: interactions complicate behavior.
Experiments will call for repeatability and
controllability.
7/20/2015
Cybersecurity Research Group -- ISTS
17
Rapid Simulation: A Tough Goal
Radio propagation: a continuous process in
continuous time.
Teletraffic: a discrete process in continuous time.
The simulation must cope with time scales of very
different resolution. Mixing them and achieving high
performance could be a tough goal.
7/20/2015
Cybersecurity Research Group -- ISTS
18
Wish List for a WAN Simulator
Detail
Completeness
Performance
Scalability
7/20/2015
Cybersecurity Research Group -- ISTS
19
Related Work
CMU: Monarch Project
http://www.monarch.cs.cmu.edu/cmu-ns.html
detailed radio propagation models, complete implementations of MAC
and routing algorithms, scenario generation, visualization tools,
network emulation, etc.
UCLA: SensorSim (pre-release stage)
http://nesl.ee.ucla.edu/projects/sensorsim/
sensing channel and sensor models, battery models, lightweight
protocol stacks for wireless microsensors, scenario generation, and
hybrid simulation.
Common major drawback: they are based on ns-2.
7/20/2015
Cybersecurity Research Group -- ISTS
20
The Architecture of SWAN
Physical Process
read terrain
features
Power Consumption
Model
Terrain
Model
read terrain
features
memory
Mobility
Model
Protocol
Graph
time
run
thread
OS
Model
(DaSSF
Runtime
Kernel)
Host Model
read terrain
features
7/20/2015
RF Channel Model
Cybersecurity Research Group -- ISTS
21
Where Things Get Complicated
Physical Processes: We need to simulate different physical
phenomena accurately and rapidly.
RF Channel Model: Propagation models are mathematically very
complex. We need to abstract and take only the most relevant
details, without this scalability is impaired.
Scale: Large number of nodes consume large amounts of memory.
Large number of nodes mean large number of computing threads
adding a big burden to scheduling.
Direct execution: Different code, potentially different behavior. We
want to allow the simulator to run the same code that runs in the real
system.
7/20/2015
Cybersecurity Research Group -- ISTS
22
The SWAN Protocol Stack Today
Application
AODV
Socket
UDP
TCP
IP
NIC
7/20/2015
ARP
NIC
MAC
MAC
PHY
PHY
Cybersecurity Research Group -- ISTS
23
What We’re Doing with SWAN

Evaluate routing protocols’ robustness to
dynamic changes in propagation conditions.

Evaluate routing protocols’ robustness to
“caveman attacks”.

Evaluate the impact of ARP in the simulation
of wireless ad hoc networks.
7/20/2015
Cybersecurity Research Group -- ISTS
24
A Duckling Called TOSSF (TinyOS+SSF)
TinyOS: An event-driven component based programming model that
powers SmartDust platforms.
Goal: Use existing SWAN framework to allow simulation by direct
execution of TinyOS applications.
TinyOS Component
Internal Tasks
Commands
7/20/2015
Internal State
Events
Cybersecurity Research Group -- ISTS
25
A TinyOS Application: a Protocol Graph
Application = graph of components + scheduler
sensing application
application
Routing Layer
routing
Messaging Layer
messaging
packet
byte
bit
7/20/2015
Radio Packet
UART Packet
Radio byte
UART byte
RFM
photo
clocks
Cybersecurity Research Group -- ISTS
ADC
Temp
SW
i2c
HW
26
Last Words

Simulation is a key technology to the development
of network protocols.

However… the behavior observed in simulations
can be trusted only as long as the models used
have been validated and verified.

Do you like this kind of stuff? Want to work with us?
Visit http://www.ists.dartmouth.edu
7/20/2015
Cybersecurity Research Group -- ISTS
27