Friends & Foes: Preventing Selfishness in Mobile Ad Hoc - - PowerPoint PPT Presentation

friends foes preventing selfishness in mobile ad hoc
SMART_READER_LITE
LIVE PREVIEW

Friends & Foes: Preventing Selfishness in Mobile Ad Hoc - - PowerPoint PPT Presentation

Friends & Foes: Preventing Selfishness in Mobile Ad Hoc Networks Hugo Miranda and Lu s Rodrigues hmiranda@di.fc.ul.pt Universidade de Lisboa DIAL-NP - LaSIGE MDC03 p.1 Outline Motivation Load balancing Selfishness


slide-1
SLIDE 1

Friends & Foes: Preventing Selfishness in Mobile Ad Hoc Networks

Hugo Miranda and Lu´ ıs Rodrigues

hmiranda@di.fc.ul.pt

Universidade de Lisboa DIAL-NP - LaSIGE

MDC’03 – p.1

slide-2
SLIDE 2

Outline

  • Motivation
  • Load balancing
  • Selfishness prevention
  • Protocol overview
  • Load balancing evaluation
  • Conclusions

MDC’03 – p.2

slide-3
SLIDE 3

Motivation

  • Mobile Ad Hoc Networks (MANETs) depend of

the individual behavior of the nodes

  • Open MANETs users may be “resource selfish”
  • Routing protocols may exacerbate the problem if

routes remain static

MDC’03 – p.3

slide-4
SLIDE 4

Dynamic Source Routing (DSR)

A B C D E F

  • 3 routing message types: requests, replies and

errors

  • Data messages include the route to be followed
  • To minimize routing messages:
  • learn routes in data messages
  • if a route is known, reply to route requests

MDC’03 – p.4

slide-5
SLIDE 5

Load-balancing in DSR

Two scenarios with:

  • 36 nodes, on a 6 × 6 matrix
  • nodes reach each of their neighbors
  • no node movement
  • no transmission errors
  • each node (i, j) sends 20 UDP datagrams to:

Scenario #1 Node (j, i) Scenario #2 Node (2, 3)

MDC’03 – p.5

slide-6
SLIDE 6

Scenario #1

30 32 51 30 71 26 49 133 176 162 89 49 73 102 225 337 156 48 81 94 211 175 182 32 31 178 116 120 95 55 32 55 34 53 33 52

  • 3468(2997 + 471)

MAC frames sent

  • 9 (25%) of the

nodes sent 1802 (52%) of the frames

  • 3 (8%) of the nodes

sent 773 (22%) of the frames

  • 50 routes used (12 of them once)
  • Node (2, 3) sent 337 frames in 24 routes

MDC’03 – p.6

slide-7
SLIDE 7

Scenario #2

24 25 57 22 21 21 21 83 95 61 21 22 41 62 85 21 21 42 141 61 21 21 81 42 21 21 21 21 21 21 21 42 41 81 132 6

  • 1561(1496 + 65)

MAC frames sent

  • 42 routes used (2 of

them once)

  • The 8 nodes surrounding (2, 3) have sent 90

frames on average. Standard deviation: 31.5

MDC’03 – p.7

slide-8
SLIDE 8
  • Therefore. . .
  • MANET routing protocols privilege efficiency
  • ver load balancing
  • Some nodes may unfairly exhaust their resources

serving others

  • The lack of load balancing mechanisms and the

absence of consequences from selfish behavior, motivates users for being “resource selfish”

MDC’03 – p.8

slide-9
SLIDE 9

Related Work

  • Path rater
  • Notifies routing protocols to avoid selfish

nodes

  • Selfish nodes may freely use the network
  • Confidant
  • Nodes refuse to forward messages from those

with a bad reputation

  • Nodes are always forbidden to be selfish (no

fairness)

MDC’03 – p.9

slide-10
SLIDE 10

Related Work - cont.

  • Terminodes
  • Virtual currency possibly mapped in real

money

  • Each hop of a message would charge some

nuglets (beans)

  • Tamper-proof device prevents frauds
  • Requires PKI
  • Power-aware routing
  • Different metrics
  • Rely on the information provided by each

node

MDC’03 – p.10

slide-11
SLIDE 11

Desirable Properties

  • Fair selfishness, if some node becomes highly

loaded

  • Nodes should remember the past behavior of
  • ther nodes
  • The protocol should be optional
  • Low overhead

MDC’03 – p.11

slide-12
SLIDE 12

Selfishness Prevention Protocol

  • Fair selfishness: improves load balancing
  • Excludes selfish nodes
  • Node reintegration
  • Charges per message
  • Requirements
  • A route composed of only non-selfish nodes

must exist between any two selfish nodes

  • Selfish nodes do not cooperate between them

MDC’03 – p.12

slide-13
SLIDE 13

Algorithm overview

Periodically, each node p broadcasts: friendsp The set of nodes to whom he is willing to provide services foesp The set of nodes to whom he refuses to provide services selfishp The set of nodes that lied to him, by declaring him as friend Nodes rate their neighbors by crossing the information received

MDC’03 – p.13

slide-14
SLIDE 14

Internal data structures

Each node p keeps a record for each of his neighbors q with: creditsq

p Messages forwarded on behalf of q

maxCreditsq

p Acceptable number of credits

friendsq

p Nodes to whom q provides services

foesq

p Nodes to whom q refuses to provide services

deadbeatq

p Evaluates if q is still in the neighborhood

MDC’03 – p.14

slide-15
SLIDE 15

Who’s selfish?

  • Decision is taken locally. Node q with the ratio:

♯friendsq

p

♯friendsq

p + ♯foesq p

bellow an acceptable threshold will be considered selfish by p.

  • Selfish nodes will only be able to send messages

until their credits reach 0

MDC’03 – p.15

slide-16
SLIDE 16

Other considerations

  • Re-integration;
  • Protocol transparency;
  • Subverting the protocol;
  • Integration with routing protocols
  • Do not forward route discovery messages

issued by foe nodes

  • Send route errors for messages to be

forwarded to foe neighbors

MDC’03 – p.16

slide-17
SLIDE 17

Evaluation # 1

30 38 53 30 90 26 51 173 187 155 70 48 106 74 185 285 141 51 93 157 189 213 174 52 50 158 125 121 105 50 57 33 50 29 78 32

  • 3559(3060 + 499)

MAC frames sent [3468]

  • 9 (25%) of the

nodes sent 1721 (48%) of the frames [1802]

  • 3 (8%) of the nodes

sent 687 (19.3%) of the frames [773]

  • 53 [50] routes used (12 of them once)
  • Node (2, 3): 285 [337] frames in 23 [24] routes

MDC’03 – p.17

slide-18
SLIDE 18

Evaluation # 2

24 57 22 21 21 25 69 85 61 20 31 44 72 84 21 30 43 105 97 21 29 48 78 31 25 27 22 23 23 39 41 91 125 6 29 42

  • 1632(1553 + 79)

MAC frames sent [1561]

  • 84 [42] routes used

(20 of them once)

  • The 8 nodes surrounding (2, 3) are closer to
  • average. Standard deviation: 19.7 [31.5]

MDC’03 – p.18

slide-19
SLIDE 19

Summary of Evaluation

  • Slightly increase on:
  • Number of frames
  • Number of routes
  • New routes alleviate overloaded nodes

MDC’03 – p.19

slide-20
SLIDE 20

Future Work

  • How to make the protocol more robust;
  • Validation of the protocol;
  • Investigate the use of other metrics;

MDC’03 – p.20

slide-21
SLIDE 21

Conclusions

  • When nodes remain in the same position, routing

protocols may present unfair load distribution

  • Selfishness prevention in MANETs is a relatively

new subject

  • A new algorithm that enhances load balancing

while banning selfish users from the MANET was presented

MDC’03 – p.21