The Role of Pervasive Sensing in Connected Mobile Environments

Rute Sofia
Rute.sofia@ulusofona.pt/rsofia@gmail.com
Technical University of Munich
2014.04.01
The Relevancy of Pervasive Sensing
In Connected Mobile Environments

2
Overview
Connected
Mobile
Environments
1. User-centric
networking
2. Mobility/
Roaming
Aspects
3. Social
Behavior in
Connected
Mobile
4. Pervasive
sensing
01-04-2014 2Rute Sofia, rute.sofia@ulusofona.pt,
http://copelabs.ulusofona.pt/~rsofia

Connected Mobile Environments
Some Food for Thought
• Connected Mobile Environments are about People
• Citizen integration into the system
• Services personalization on-the-go (learn with an
adaptive user profile – roaming/commuter behavior)
• Connected Mobile Environments require looking
into new paradigms
• User-centric networking
• Integration of social notions into the OSI stack
• Mobility Modeling is essential to ensure Internet
user/customer satisfaction
• Roaming prediction assists the most varied areas :
residential services; intelligent transportation systems;
smart environments
• Better QoE – more customers
• Connected Mobile Environments require knowledge
acquired on-the-fly
• Participatory sensing requires too much user
involvement
• Pervasive sensing is the key to generate new services
Connected Mobile
Environments
1. User-centric
networking
2. Mobility/
Roaming
Aspects
3. Social Behavior
in Connected
Mobile
4. Pervasive
sensing

4
Connected
Mobile
Environments
1. User-
centric
networking
2. Mobility/
Roaming
Aspects
3. Social
Behavior in
Connected
Mobile
4. Pervasive
sensing

 Shared access
 Why: broader roaming
 Neglected cooperation
potential
 Density implies bad
resource management
 Neglected available
spectrum
 Spectrum overlaps
 Architectures follow the
Internet end-to-end
principle
 Clear splitting between
network and end-user
devices
 Mobility is limited
 Roaming with specific
areas; static, centralized
models
Private
User-centric Networking
User as Prosumer, Networking Aspects
User-centric networking

End-user devices
Direct sharing
Shared connectivity
Micro-provider device
User-centric relaying
Virtual node aggregated
relaying
Multi-hop relaying Direct aggregated relaying
MiFi, Whisher
Internet access point
Devices sharing connectivity
People-centric Architectures in Connected Environments*
*R. Sofia, P. Mendes, User-provided networks, Consumer as Provider. IEEE Communications
Magazine, vol. 46, pp86-91, 2008.
FON, ZON

Relevant Aspects in Connected Mobile Environments
Trust Management
• Dynamically built
• Mimics human trust
Advanced Routing
• Based on user’s interests
• Opportunistic
Advanced
Resource Management
• Take advantage of
overlap and interference
Mobility Prediction
• Improve QoE
• Assist network operation
User-centricity
???
Destination: Mall
Getting off at next bus stop

Circles of Trust in Connected Mobile Environments
Trust in Networking
Node
Wireless device that belongs to 1 individual (owner)
 Only the owner is responsible, for the case of shared
devices
An individual may own different nodes
Trust association
Unidirectional association between two nodes
Related to nodes’ interests and social networking
perception
Has a specific trust level or trust weight
Trust weight is based on QoE, previous history, etc.
Two nodes may hold more than one trust association
(due to QoE or type of traffic)
Trust Propagation
Based on direct and indirect recommendations
Passively overhead (e.g. via wireless beacons)
Based on dispositional trust adjusting over time

Circles of Trust in Connected Mobile Environments, Example*
•Community 1 is already active
•ULOOP “Gateway” (equipment providing the shared Internet access) in red
•Maria in Blue, Tom in green
•In Community 1, Maria holds a trust level e.g. 4
Maria’s community 1 trust level: 4
Tom - new user in
Community 1 –
Maria provides
Internet access,
basic services
Tom gets the
quarantine level –
trust association
cost from Tom to
Maria = 2
*EU FP7 ULOOP – User-centric Wireless Local Loop
http://uloop.eu/

Computing Trust in Connected Environments, the ULOOP suite*
•https://play.google.com/store/apps/details?id=eu.uloop.t31.android
•EU FP7 ULOOP, http://uloop.eu
1. Id creation – one time, one user, multiple
devices
• Crypto-id generation (SHA-256(Public Key) of a
KeyPair)
• Validation – any Id validator
• 1 crypto-id, multiple node-ids: Crypto-id || encrypted
MAC
2. Trust Setup (one time, one user)
• Dispositional trust - the way I trust thirds/the
world
• Affects trust computation
• Trust table <crypto-id, trust level, ageing>
• Wallet – initial set of credits
Bank
(Monetization)
Id Validator
(One time)
3. Trust negotiation and management
• During the MAC authentication phase
• Negotiates trust – uses tokens and credits
• Example: Token tk(i,j) = tl(i,j)*sqrt(c), tl in
[0,1]; c in [0,inf]
4. Rewards and Cooperation
• Cooperation manager mediates negotiation
• Triggers request for rewards
• Reward manager handles payments (monetization)
• Payments are secured with RSA

Trust as a metric to provide QoS/QoE*
Hostapd
STA
Auth Response
CAC
Request
Queue
Request with the highest
priority.
Resource
Manager
0: OK
32: Not OK
Request priority p(i,j)=Tokens(i,j)*Tl(j,i)
j: gateway
i: node requesting service
Tl: trust level
*R. Sofia, L. Lopes, Trust as a Fariness Parameter for
Quality of Experience, Book Chapter contribution,
Wireless Networks, Springer LCNS User-centric
Networking - Future Perspectives, 2014.

MAC 1 MAC 2
Internet
mac80211- TX
ST
A1
ST
A2
24
Mbps
18
Mbps
ACK
AC
K
Throughput of high rate
station is impaired by the
low rate station.
Making the MAC Layer Fairer, Dynamic Frequency Sharing*

MAC 1 MAC 2
Internet
ST
A1
ST
A2
Throughput of
the high rate station is not
impaired by the low rate station.
The low rate station maintains
the same throughput with only
the addition of a small processing
time for the superframe.
Making the MAC Layer Fairer, Dynamic Frequency Sharing*
*L. Lopes, R. Sofia, H. Haci, H. Zhu,
A Proposal for Dynamic Spectrum Sharing in Wireless Networks, under submission, 2014.
H. Osman, H. Zhu, H. Haci, L. Lopes, R. Sofia. “Method and Apparatus for communication
in a wireless network” (EP 13191667.8), August 2013

01-04-2014 Rute Sofia, rute.sofia@ulusofona.pt,
14
Avg Throughput Difference between the two stations (%)
Our proposal IEEE802.11g
Scenario I 3.59 82.32
Scenario II, LoS, UDP 27.04 27.04 68.11
Scenario II, LoS, TCP 16.65 16.65 44.44
Scenario II, nLoS, UDP 81.02 48.57
Scenario II, nLoS, TCP 41.97 82.95
Trust as a metric to provide QoS/QoE*
*R. Sofia, L. Lopes, Trust as a Fariness Parameter for Quality of Experience
in Wireless Networks, Book Chapter contribution References
pringer LCNS User-centric Networking - Future Perspectives, 2014.

15
Connected
Mobile
Environments
1. User-
centric
networking
2. Mobility/
Roaming
Aspects
3. Social
Behavior in
Connected
Mobile
4. Pervasive
sensing
User IS part of the
network
Trust as a new QoE
metric – better QoE,
fairer QoS
Wrapping-up
Connected
Mobile
Environments
1. User-
centric
networking
2. Mobility/
Roaming
Aspects
3. Social
Behavior in
Connected
Mobile
4. Pervasive
sensing

Internet
Home
Coffee
MN
Mobility anchorpoint
Identifier @Home
Identifier @Visited Network
Remarks
End-user device has an identifier for each place
Can be the same, e.g., 3GPP
For the network, identifier is the same
Always the one used when starting the communication
Endpoint identifier: never changes
Visited network identifier is a “location marker”
Mobility anchorpoint
Tracks mapping between identifiers, as well as active identifiers
Transport
(OSI L4)
Network
(OSI L3)
Data/Link
(OSI L2,L1)
Application
(OSI L5-L7)
SIP/IMS
M-SCTP
HIP
Mobile IP
Layer 2 Mobility
(micro-mobility)
3GPP/4G/5G
Wi-FiWiMAX
802.21
Mobility Aspects
Today’s Mobility Management Solutions

Mobility Aspects
Distributed Mobility Management, an Example
Distributed Mobility
Management
IETF DMM Working Group
Envisions distribution via
replication, or splitting
Our view: splitting and
placing some functionality
closer to the user
Today’s centralized
approaches
Control Plane
Mobile node registration in a central
mobility anchor point (MAP)
Secure exchange for binding
Data Plane
VO
User B
Sessions always on
Access Provider 1
Access Node
Access Provider 2
Access Node
Residential Gateway
User A
Mobility ControlPlane
Residential Gateway
Control /
Data Plane
AAA
Mobility
Anchorpoint

Mobility Aspects
The True Challenges in Connected Mobile
Efficient selection of anchor points
Based on user satisfaction degree Based on network policies
Flexible mobility architectures
Distributed mobility management
Better for flater networks
Better for variable topologies
A new meaning to “anywhere, anytime”
Ways to estimate WHEN is session continuity
required (mobility estimation) and WHICH services
are required
Use any opportunity to transmit on user’s interests

Mobility Aspects
The Rise of Mobility Modeling
Modeling and realistic operation – aren’t these two distinct fields
Optimal routing,
incorporate sensitivity
to movement
Simplified network
operation, predict future
movement aspects
Optimized mobility
management–
introduce estimation
Better connectivity
models – place better
mobility functionality
Why not just use ANY available model ?
Modeling impacts severely network operation
Social mobility modeling captures some realistic
human movement features
Why do we need mobility models anyway?
In connected mobile, nodes move freely – based
on human behavior and routines
Node movement patterns and relative location
(node to group) is relevant to ensure a good
communication

Mobility Aspects
The Rise of SOCIAL Mobility Modeling
• Intel Lablet, Cambridge UK
• Distribution of bluetooth devices to people in different
settings (Cambridge, Infocom 2005, Tokyo)
• Wireless Topology Discovery Project (UCSD)
• Darthmouth CRAWDAD
Traces
• Up to now, most popular, but started around 2000
• Hard to assess applicability to reality
• E.g., end-user portable devices move with humans, not
randomly
• Traces corroborate gaps in applicability to reality
Synthetic
• Attempts to model human roaming behavior/movement
• Often based on the notion of social attractiveness
• Still present several gaps: obstacle collision; pause time
modeling; community detection
Social Mobility Modeling

Mobility Aspects
Social Behavior Modeling

Mobility Aspects
An example for roaming estimation*
• Passively captures networking context data that
characterizes visits to networks over time
Tracking
• Based on learnt behavior, infers time left to
handover and potential handover targets
Predicts
* R. Sofia (COPELABS), EP 13186562.9, 08.2013 – patent pending.
• Roaming patterns share affinity with human
routines
• Tracking such routine allows prediction
Motivation
• Improve handovers – reduce signaling
• Quality of Service/Better resource
management
• Better selection of mobility anchor points
Benefits

Mobility Aspects
An example for roaming estimation*
Visited network 1, rank 0.8
Visited network 2,
rank 0.3
Visited network 3,
rank 0.9
Today
MTracker
 User equipment application that passively
ranks visited Wi-Fi networks over time, by
collecting a few parameter
 Number of visits
 Average visit time in seconds
 Information concerning the visited network
e.g. BSSID and SSID
 Attractiveness of the network to the user:
trust level the user has on that ULOOP
gateway
 Rank is periodically updated based on
the collected parameters
 Mtracker estimates a potential move
 Selects best network based on rank
 Consideres average visit time as the indicator
that a move can occur
*http://www.youtube.com/watch?v=yzGJyDlamRU

24
Connected
Mobile
Environments
1. User-
centric
networking
2. Mobility/
Roaming
Aspects
3. Social
Behavior in
Networking
4. Pervasive
sensing
•User-centricity: new
paradigms in mobile
•Mobility management
requires new approaches,
closer to human behavior
•Mobility estimation – a
relevant trend
Mobility Aspects
Wrapping-up
Connected
Mobile
Environments
1. User-
centric
networking
2. Mobility/
Roaming
Aspects
3. Social
Behavior in
Connected
Mobile
4. Pervasive
sensing

Future Connected Environments, Social behavior integration
Movement patterns
QoE based on social principles (e.g. trust,
influence)
Internet value-chain (wholesale models)
Flow from providers to citizens still Value-chain is not assymetric
Citizen as prosumer
Citizen shares the network Citizen provides content
Citizen personalizes services in
real-time
The end-to-end principle still rules
The network is dumb The edges have the intelligence
Edges as of today: user-centric
networks
The Internet as Social Media
Citizens control/carry networking nodes Citizens exchange data, shared interests
Social Behavior in Connected Mobile
Integrating Social Metrics – closer to Human Behavior

Property/Parameter Social Capital Perspective Networking Perspective
Centrality: determine the
relative importance of a vertex
within the graph
The influence of a person on the social
structure
The impact a node has on the graph.
Importance here relates to information
dissemination.
Degree centrality: Nodes that
have more ties to other nodes
have a higher degree centrality.
Considers that such nodes are better
positioned (influence, information
dissemination). Alone, says little about
node influence. Together with the degree
centrality of neighbors, provides a better
measure
Nodes that have more ties to other nodes have
a higher degree centrality. These are not
necessarily better positioned.
Betweeness centrality: nodes
that have a high probability to
occur on a randomly chosen
shortest path between two
randomly chosen nodes have a
high betweenness.
Assists in finding “bridgers”: these are
nodes that limit clusters (interconnect
different clusters).
Links that are more central assist nodes in
better dissemination information, assuming a
plain connectivity model.
Closeness centrality: Sum of
its (shortest-path) distances to
any other node y normalized by
the maximum shortest-path
length.
High closeness centrality implies better
information propagation.
A node that has a higher number of shortest-
paths to all other nodes has a higher closeness
centrality. It also has a higher probability
of becoming a bottleneck
Link Strength: The strength of
a tie depends on the amount
of time spent on it and the
emotional intensity and intimacy
of the relation
If there is a strong tie between A
and B as well as between B and C, A and C
are
likely to develop a strong tie as well. This
ten-
dency cannot be observed for weak ties.
If there is a strong tie between A and B and
another between B and C, this says nothing
about A and C.
Disconnect Between Social and Network Perspectives

Why Should we Integrate Social Metrics ?
• Analyse social behavior (e.g. skype calls, specific area)
• Considers traffic locality
• Place networking nodes accordingly to usefulness
• improve adoption modellng
Connectivity Modeling
• Improve dissemination of information
• Take advantage of any possible contact
• Social interaction as the vehicle for dissemination
Routing
• Improve resource allocation
• Improve user satisfaction
Resource Management
• Estimate movement
• Anticipate movement
• Improve mobility management
• Develop new services
Mobility management and
modeling

Property/Para
meter
Social networking Analysis definition Networking Measurement
Metrics
Reach the degree of effective dissemination of certain
content or potential spread that a single
profilehas in the network
Rate of nodes reached;
proximity; propagation speed
Engagement the degree of participation and involvement of a
specific profile. A profile in networking can be
seen as e.g. a preferred location; an interest
towards a node/cluster/location.
growth of the followers of
profiles,
time spent in profiles,
number of visits
reciprocity
Link association strength
Influence the degree of attention and mobilization that a
certain profile can generate in other users
•FRINGE algorithm estimates the
impact of node based on direct
and indirect neighborhood
•Klout score
Integrating Social Metrics, Example*
* R. Sofia, P. Mendes, M. J. Damásio, S. Henriques, F. Giglietto, E. Giambitto and A. Bogliolo,
Moving Towards a Socially-Driven Internet Architectural Design (2012), in: ACM SIGCOMM CCR Newsletter, 42:3. 2012.

29
Connected
Mobile
Environments
1. User-
centric
networking
2. Mobility/
Roaming
Aspects
3. Social
Behavior in
Connected
Mobile
4. Pervasive
sensing
•Networking architectures must consider
social metrics – QoE and QoS
• Metrics must integrate social (knowledge
value) and technological perspectives
(technological adoption value)
•Where to start: trust, influence/reach
Social Behavior
Wrapping-up
Connected
Mobile
Environments
1. User-
centric
networking
2. Mobility/
Roaming
Aspects
3. Social
Behavior in
Connected
Mobile
4. Pervasive
sensing
Connected
Mobile
Environments
1. User-
centric
networking
2. Mobility/
Roaming
Aspects
3. Social
Behavior in
Connected
Mobile
4. Pervasive
sensing

Pervasive Sensing
Integrating Social Metrics: the How
• Always with the user
• New cars come equipped with GPS, navigation systems, and
lots of sensors
• Residential households with an average of 3 devices
Any portable device with
mobile/wireless capabilities
• Simplified and personalizeable systems
• Positioning
• Middleware that supports any type of sensor
Software defined solutions
• Increase the network on the go (user-centric networking)
• Seamless exchange stronger in trusted circles
Capability to share
connectivity
• Citizens that share routines – familiar strangers
• Exchange grows with trust
Relies on shared interests

Pervasive Sensing
Paradigms
• Devised for a specific purpose, e.g. health monitoring
• Focused on data capture and collection
• Example: fitness tracker
Personal sensing
• User is actively engaged – provides data or context
• Requires some infrastructure (software and hardware)
• Example: http://urban.cens.ucla.edu/
Participatory sensing
• Increase the network on the go (user-centric networking)
• Seamless exchange stronger in trusted circles
• Pervasive – carried by the citizen
• Depends on shared interests
Mobile-centric sensing

Pervasive Sensing
A New Platform for Connected Mobile Environments*
PerSense
Pervasive
Sensing
Framework
32
PerSense
Pervasive
Sensing
Framework
GPS
Accelerometer
WiFi
BluetoothCall log
Sensing and Data Capture (SEC) Behavior Inference (BE)
Context Modeling (COM)
Pervasive Data sharing (PERSH)
Opportunistic
Computing
Big Data
Analysis
Pervasive
Sensing Data
Centric
Opportunistic
Networking
Online Social Networks
Post
PerSense
Pervasive
Sensing
Framework
PerSense
Pervasive
Sensing
Framework
* P. Mendes. R. Sofia, PerSense, a Pervasive Sensing Framework. Under submission, 2014

PerSense
PerSense enabled Access Point
Standard Access Point
2- Bob goes Shopping as usual
PerSense updates roaming data
1- Bob meets Peter
PerSense social info updated
4- Bob’s tries to find Bob
•PerSense predicts that Bob’s at Market
•PerSense estimates arrival time to home
•PerSense is configure to call Authorities if
Bob does not arrive 1 hour after estimated
time
Home
3- Bob goes to the Market
PerSense info stored at Bob’s MOT.
Pervasive Sensing
Example, Intermittently Connected Mobile Environments

34
Summary
Connected
Mobile
Environments
1. User-
centric
networking
2. Mobility/
Roaming
Aspects
3. Social
Behavior in
Networking
4. Pervasive
sensing
•Networking architectures must
consider social metrics – QoE
and QoS
• Metrics must integrate social
(knowledge value) and
technological perspectives
(technological adoption value)
•Mobility management requires new
approaches, closer to human
behavior
•Mobility estimation as a
relevant trend to address
•User devices as part of the
network
•Trust as a new QoE metric
– better QoE, fairer QoS
•Essential to provide an
adequate contextual definition
of human roaming behavior
•Self-organizing, and non-
intrusive

The Role of Pervasive Sensing in Connected Mobile Environments

Recommended

Recommended

More Related Content

Similar to The Role of Pervasive Sensing in Connected Mobile Environments

Similar to The Role of Pervasive Sensing in Connected Mobile Environments (20)

More from Rute C. Sofia

More from Rute C. Sofia (20)

Recently uploaded

Recently uploaded (20)

The Role of Pervasive Sensing in Connected Mobile Environments