A Service for Controlling Information Dissemination in Wireless Networks Xinfeng Li Jin Teng Boying Zhang Adam Champion and Dong Xuan IEEE Infocom2012 Information Dissemination 2 ID: 427760
Download Presentation The PPT/PDF document "TurfCast" is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Slide1
TurfCast: A Service for Controlling Information Dissemination in Wireless Networks
Xinfeng
Li, Jin
Teng
,
Boying
Zhang, Adam Champion and Dong
Xuan
IEEE Infocom2012Slide2
Information Dissemination
2Slide3
Information Dissemination with TurfCast
3
0/1 information dissemination over both temporal AND spatial domainSlide4
Outline
TurfCast
Applications
TurfCast
Key Components
TurfCode
TurfBurst
Implementation and Evaluations
Conclusion
4Slide5
TurfCast Applications
Electronic Proximity Advertising
Reward Loyal Customers with Special Promotions
5Slide6
TurfCast Applications (Cont’d)
Mobile Social Networking
Publishes personal information in a temporal/spatial bounded location
Location Verification
Generates physical presence proofs for users in proximity in an automatic manner
6Slide7
Possible Approaches For TurfCast
Power Control
Only spatial support
Hard to control
Cryptography
Computationally expensive
Difficult for fluid networks
7Slide8
Our Approach
Pure communication-based
Leverage fountain code
Include two key components
TurfCode
: for temporal information control
TurfBurst
: for spatial information control
8Slide9
TurfCode
A new type of nested 0-1 fountain code.
To achieve
temporal turfs
Fountain code:
Streams of packets coded from raw information
Linear combination of message chunks
Whenever enough packets are received, decoding is possible
Like collecting water from a fountain to fill a cup
Statistically more water drops, more information
9Slide10
Premature Decoding of Fountain Codes
A fraction of information can be decoded even if the number of received packets does not reach the correct decoding threshold
10
K: number of packets for correct decoding
Fountain Codes cannot be directly used for
TurfCastSlide11
0-1 Fountain Codes
We add a new scrambling
precoding
layer
To ensure that any linear combination of the received packets cannot reveal any block before intended decoding
11
m’=
Sm
c=
Bm
’Slide12
Scrambling Matrix Design
m’=
Sm
Not conventional bit-wise XOR operation. We have proved that scrambling with bit-wise XOR operation cannot be safe.
We use real numerical additions
The matrix S is designed to be a
trilinear
matrix
Easy to guarantee that a
trilinear
matrix is reversible
For decoding, we need the matrix to be reversible
We can randomly pick the diagonal elements
A small change in the diagonal matrix can result in a total different m’
These diagonal elements are the key to de-scramble the message
A hacker must guess all diagonal elements correctly to get any information
12Slide13
Nested 0-1 Fountain Codes
For multiple level of
TurfCast
, we need nested 0-1 fountain codes.
13Slide14
TurfBurst
Use the Shannon bounds to achieve spatial turfs.
Users at different locations have different capability of receiving information
Users far away gets lower Signal to Noise Ratio (SNR), and hence lower information amount according to the Shannon bound
With lower information amount, the user is able to decode less information
14Slide15
TurfBurst Design
TurfBurst
needs special attention when designed
Boundary not so sharp as the noise is probabilistic.
Need larger difference between layers to accommodate the noise.
However, in many scenarios, we have barriers that drastically reduce the SNR, which is to our advantage.
E.g., Room Walls
15Slide16
TurfCast Put Together
Messages of several different priorities to be sent.
Each priority level has distinctive requirement, i.e., when the message should be received and how far
Messages with lower priority cannot be received before those with higher priority.
Need to design meta-blocks based on
TurfCode
and
TurfBurst
to achieve this
16
TurfBurst
TurfCodeSlide17
A Quick Recap of Our Solutions
17
TurfCast
:
Multi-level Priority Information Dissemination
TurfCode
:
Temporal Control
TurfBurst
:
Spatial ControlSlide18
Implementation
On laptops and Android
Smartphones
Receiver side: Laptop using
aircrack-ng
for MAC level packet sniffing in
WiFi
.
Sender side:
Laptop
:
Use
aircrack-ng
for MAC level packet injection.
Nexus S: Socket programming on the TCP/IP level
18Slide19
Evaluations (Temporal)
19
User can get Layer 1 message in around 5-10 seconds
But to get Layer 2 messages, they need to linger around 10 seconds longer
To get Layer 3 messages, they need to stay still longer.Slide20
Evaluations (Spatial)
20
Users within the inner circle can get Layer 3 messages
Users between circle 1 and 2 gets significantly less Layer 3 messages
The farthest users receives little Layer 3 information.Slide21
Final Remarks
TurfCast
concept
0-1 Communications
Enable
new applications
Temporal and Spatial information control based on 0-1 fountain codes
TurfCode
for temporal control
TurfBurst
for spatial control
21Slide22
22
Thank you!