ntu im oplab providing survivability against jamming attack for multi-radio multi-channel wireless...

NTU IMOPLAB

Providing survivability against jamming attack for multi-radio multi-channel wireless

mesh networksJournal of Network and Computer Applications

Author: Shanshan Jiang, Yuan Xue

Advisor: Frank,Yeong-Sung Lin

Presented by Jia-Ling Pan

12010/06/08

COMPANY LOGO

Outline

Introduction System model Routing and channel assignment

without jamming attacks Optimal restoration strategies under

jamming attacks

2

2010/06/08

COMPANY LOGO

Outline

Dynamic channel assignment under jamming attacks

Static channel assignment under jamming attacks

Performance degradation model Performance evaluation Concluding remarks

2010/06/08

3

COMPANY LOGO

Introduction

Wireless mesh network are capable of communicating with each other and cooperating to relay traffic throughout the network via multiple hops.

Built upon open wireless medium, wireless mesh network is particularly vulnerable to jamming attacks.

The ability to deal with jamming attacks and maintain an acceptable level of service degradation in presence of jamming attack is thus a crucial issue.

4

2010/06/08

COMPANY LOGO

Introduction

When jamming occurs, the traffic going through that jamming area is disrupted.

The network either switches to different channels other than those of the jammers, and/or its traffic needs to be rerouted around the jamming area.

This paper investigates the jamming defense strategies via the joint design of traffic rerouting, channel re-assignment, and scheduling in a multi-radio multi-channel wireless mesh network.

5

2010/06/08

COMPANY LOGO

Introduction

First formulate the optimal network restoration problem as linear programming problem, which gives an upper bound on the achievable network throughput.

Consider two strategies, namely global restoration and local restoration.

Based on the LP solutions, providing a greedy scheduling algorithm using dynamic channel assignment, which schedules both the network traffic and the jamming traffic.

6

2010/06/08

COMPANY LOGO

Introduction

Further provide a greedy static edge channel assignment algorithm, where a channel is assigned to an edge at the beginning and will remain fixed over all time slots.

Define two performance degradation indices, transient disruption index (TDI) and throughput degradation index (THI).

7

2010/06/08

COMPANY LOGO

Outline



jamming attacks

8

2010/06/08

COMPANY LOGO

System model

Normal node model Consider a multi-radio multi-channel wireless mesh

network and model it as a directed graph G=(V,E,C).

Each node v is equipped with k(v) radios. E=ET υ EI

All nodes have a uniform transmission range (RT) and a uniform interference range (RI).

A transmission edge e=(v,v’) Є ET is formed if r(v,v’) ≤ RT.

An interference edge e=(v,v’) Є EI is formed if RT ≤ r(v,v’) ≤ RI.

9

2010/06/08

COMPANY LOGO

System model

Wireless channel capacity is the same for all edge using channel c (Φc).

Packet transmission from node v to v’ is successful if and only if ：

There is a transmission edge e=(v,v’) Є ET.

Node v and v’ have radios that support a common channel c .

Any other node v” Є V within the interference range of the sending node v or the receiving node v’, i.e., e=(v,v”) Є ETυEI

or e=(v’,v”) Є ETυEI, is not transmitting on channel c .

10

2010/06/08

COMPANY LOGO

System model

I(e) ： Interference set which contains the transmission edges that interfere with transmission edge e, e Є ET.

The traffic between any pair of nodes as a flow and denote it as f Є F.

sf ： the sending node of flow f

rf ： the receiving node of flow f

df ： the demand of flow f The traffic of flow f will be routed over multiple

paths and multiple channels ： xf(e,c) ： The amount of flow f’ s traffic being

routed on edge e over channel c.11

2010/06/08

COMPANY LOGO

System model

∑ f Є F xf(e,c) ： The amount of all flows’ traffic on edge e over channel c

Jamming node model jc Є Jc ： A wireless jammer node at channel c.

Jc ： The set of all the jammers detected at channel c.

J ： The set of all the jammers over all the channels.

Gjc ： Constant traffic generating rate 0 ≤ Gjc ≤ Φc.

All the jamming nodes have a uniform jamming range RJ.

12

2010/06/08

COMPANY LOGO

System model

Assume that they are smart jammers that can totally occupy the channels when sending jamming traffic.

Jc(e),e Є ETυEI ： the set of jammers who have one or both of the two end nodes of the edge e within its jamming range.

ET(jc) ： the set of transmission edges whose sending or receiving node is within the jamming range of jc.

Assume that two jammers are not within the jamming range of each other.

13

2010/06/08

COMPANY LOGO

Outline



jamming attacks

14

2010/06/08

COMPANY LOGO

Routing and channel assignment without jamming attacks

First study the routing and channel assignment problem in a multi-radio multi-channel wireless mesh network when there is no jamming node.

The goal is achieving the maximum throughput.

Find a best strategy that can minimize the performance degradation to defend against jamming attacks.

15

2010/06/08

COMPANY LOGO


The necessary conditions of channel assignment and scheduling for a multi-radio multi-channel wireless network are summarized as follows:

Node radio constraint ：

Channel congestion constraint ：

17

2010/06/08

COMPANY LOGO


Minimum flow throughput scaling factor(λ) ：Characterizes the throughput of a given routing with respect to a certain traffic demand.

18

2010/06/08

COMPANY LOGO


flow flow conservation conservation constraintconstraint

node radio node radio constraintconstraint

channel channel congestion congestion constraintconstraint

19

2010/06/08

COMPANY LOGO

Outline


without jamming attacks Optimal restoration strategies

under jamming attacks

20

2010/06/08

COMPANY LOGO

Optimal restoration strategies under jamming attacks

The disrupted network traffic can be rerouted to use other intermittent nodes away from the jamming area, or switched to another channel instead of using the jammed channel.

Include the jamming traffic into the channel congestion constraint:

21

2010/06/08

COMPANY LOGO


The network restoration via joint traffic rerouting and channel re-assignment under globalglobal and locallocal restoration strategies.

22

2010/06/08

COMPANY LOGO


Global restoration

flow flow conservation conservation constraintconstraint



23

2010/06/08

COMPANY LOGO


Local restoration Need to find the bypass flows that need to be

partially routed away from the jamming area. For these flows, their immediate upstream and

downstream nodes surrounding the jamming area should remain unchanged.

24

2010/06/08

COMPANY LOGO


Bypass flows For a jamming node jc and a flow f ： inf(jc) ： The set of nodes that are within the

jamming area of jc. pref(jc) ： The set of nodes sending data of f

directly to one or more nodes in inf(jc).

postf(jc) ： The set of nodes receiving data of f directly from one or more nodes in inf(jc).

25

2010/06/08

COMPANY LOGO


26

2010/06/08

COMPANY LOGO


Optimal bypass restoration Xbf (v,v’,jc)(e,c) ： The traffic demand of bf(v,v’,jc)

that is routed over edge e and channel c. The bypass flows need to share the wireless

channel capacity with the original flows, both of them need to be scaled again.

λb ： Scaling factor with bypass restoration.

27

2010/06/08

COMPANY LOGO




28

2010/06/08

COMPANY LOGO


bypassbypassflow flow conservation conservation constraintconstraint

29

2010/06/08

COMPANY LOGO

Outline




30

2010/06/08

COMPANY LOGO


The global restoration and the local restoration give an upper bound on the achievable network throughput.

Dynamic channel assignment problem ： A radio may need to switch to a different

channel at different time slots. Provides the maximum flexibility in channel

assignment and scheduling. Greedy approach

31

2010/06/08

COMPANY LOGO


Schedule both the network traffic on the edges and the jamming traffic.

I(e*) ： The set of transmission edges that interfere with edge e*.

E(Jc*) ： The set of transmission edges that

are within the jamming range of jammer jc*.

32

2010/06/08

COMPANY LOGO


33

2010/06/08

COMPANY LOGO


34

2010/06/08

COMPANY LOGO


round 0 1 2 3 4 5 6

e* e1 e2 e2 e1 e2 e1

x(e1,c) 3 2 2 2 1 1 0

x(e2,c) 3 3 2 1 1 0 0

35

V

V’

V’’

(e1,c)

(e2,c)

node set

v 1,1,2,2,3,3 1,1,2,2,3,3,…..

v’ 1,2,3 1,1,2,2,3,3,……

v’’ 1,2,3 1,1,2,2,3,3,……

edge-channelset

e1,c 1,2,3 1,2,3,4,…….

e2,c 1,2,3 1,2,3,4,…….

k(v)=k(v’)=k(v’’)=2color poorcolor set

2010/06/08

COMPANY LOGO


N ： The maximum number of time slots taken by all the edge-channel pairs.

The new scaling factor λSJ after scheduling

is calculated as ：

36

2010/06/08

COMPANY LOGO

Outline




37

2010/06/08

COMPANY LOGO


Dynamic channel assignment provides the maximum flexibility, but results in channel switching overhead.

Static edge channel assignment problem A channel is assigned to an edge at the

beginning and will remain fixed over all time slots.

Greedy approach

38

2010/06/08

COMPANY LOGO


Constraint set The node radio constraint and the channel

congestion constraint have a common structure.

Left sides of (1)&(10),have LL sets, each of which is composed of (edge, channel)pairs.

Right sides of (1)&(10), have LL fixed values, where L is the number of all the expanded inequalities.

39

2010/06/08

COMPANY LOGO


S1, S2,…., SL ： the sets of (edge, channel) pairs.

βS1-GS1

, βS2-GS2

,…., βSL-GSL

： their

corresponding values. If Si comes from the node radio constraint ：

βSi = k(v)Φc , GSi

= 0 .

If Si comes from the channel congestion constraint ：

βSi = Φc , GSi

= ∑jc Є Jc(e) Gjc

.

General form of Inequality (1) & (10) using constraint sets is defined as follows:

40

2010/06/08

COMPANY LOGO


Static channel assignment x(e) ： The amount of all flows’ traffic over all

the channels on edge e. For simplicity, assume that only one channel

can be assigned to a given edge. Therefore, x(e) is assigned to one particular one particular

channel assigned to edge e. The basic idea of static channel assignment

algorithm is to distribute the load on the constraint sets as much as possible among the given channels.

41

2010/06/08

COMPANY LOGO


42

2010/06/08

COMPANY LOGO

Outline




43

2010/06/08

COMPANY LOGO

Performance degradation model

Challenge for choosing the restoration strategy ： Understand the tradeoff between the overhead

involved in repairing the failed traffic path(s) and the traffic throughput and network congestion after restoration.

Transient Disruption Index (TDI) Based on the repair overhead for the failed

traffic path(s) during restoration.

Throughput Degradation Index (THI) Characterizes throughput degradation of the

new network after restoration.44

2010/06/08

COMPANY LOGO


Transient disruption index (TDI) Use the number of modified routing table

entries as an estimate of the repair overhead for the failed path(s).

For local repair, only the boundary nodes outside the jamming area will try to find the alternative paths.

For global repair, the source node initiates a new route discovery and involves more routing table entry modifications.

45

2010/06/08

COMPANY LOGO


rv(c,v’) ： A routing table entry of node v’s routing table.

At a given channel c, it is calculated as the ratio of the total traffic of all the flows sending from node v to its next-hop node v’ to the total traffic of all the flows receiving at node v.

rv*(c,v’) ： Its corresponding routing table entry

for the new network under jamming. r(G) ： All the routing table entries of the nodes

in the network G.

46

2010/06/08

COMPANY LOGO


Throughput degradation index (THI) Use the changes of the minimum flow

throughput scaling factor λ as an estimate of the throughput degradation of the new network.

For local repair, it achieves partially optimal utilization of the network.

For global repair, all flows in the network will be considered in order to get an optimal utilization of the network.

47

2010/06/08

COMPANY LOGO

Outline




48

2010/06/08

COMPANY LOGO

Performance evaluation

Simulation setupSimulation setup Simulation result overview Comparison of TDI and THI under various

scenarios Comparison of λ under various scenarios

49

2010/06/08

COMPANY LOGO


54 wireless nodes are randomly deployed over a 1800*1080 m2 region.

RT = 250 m, RI = 250 m.

Channel capacity Φc(c Є C) is set as 1 Mbps. Three randomly distributed jamming nodes. RJ = 100 or 200 m. Traffic generating rates of the jammers are

from 0.2 to 0.8 Mbps.

50

2010/06/08

COMPANY LOGO


All the flows‘ traffic demand of 1 Mbps.

51

2010/06/08

COMPANY LOGO


Evaluate the performance of the global restoration and local restoration under two scenarios: Single channel

All the network nodes and jamming nodes use the same channel.

Multiple channels All the network nodes and jamming nodes

use multiple channels and │ C│ =5. Each network node is equipped with multiple radios. Jammers are able to send jamming traffic over all the channels.

52

2010/06/08

COMPANY LOGO


Simulation setup Simulation result overviewSimulation result overview Comparison of TDI and THI under various

scenarios Comparison of λ under various scenarios

53

2010/06/08

COMPANY LOGO


54

2010/06/08

COMPANY LOGO


55

2010/06/08

COMPANY LOGO


56

2010/06/08

COMPANY LOGO


57

2010/06/08

COMPANY LOGO


58

2010/06/08

COMPANY LOGO


Simulation setup Simulation result overview Comparison of TDI and THI under Comparison of TDI and THI under

various scenariosvarious scenarios Comparison of λ under various scenarios

59

2010/06/08

COMPANY LOGO


60

2010/06/08

COMPANY LOGO


61

2010/06/08

COMPANY LOGO


Simulation setup Simulation result overview Comparison of TDI and THI under various

scenarios ComparisonComparison of of λλ under various under various

scenariosscenarios

62

2010/06/08

COMPANY LOGO


63

2010/06/08

COMPANY LOGO


64

2010/06/08

COMPANY LOGO

Outline




65

2010/06/08

COMPANY LOGO

Concluding remarks

This paper investigates the network restoration problem in multi-radio multi-channel wireless mesh networks under jamming attacks.

The defense strategy dynamically adjusts the channel assignment and traffic routes to bypass the jamming area.

Two restoration strategies Global restoration Local restoration

66

2010/06/08

COMPANY LOGO

Concluding remarks

The goal is to minimize the performance degradation caused by the jamming attack.

Formulates as linear programming problems.

Network performance are evaluated via comprehensive simulation study under different jamming attack scenarios.

67

2010/06/08

COMPANY LOGO

Concluding remarks

Static channel assignment Inefficient use of radio resource also hinders

the network performance improvement when the number of radios is increased.

The post-restoration network performance (i.e., THI) may improve, when the network nodes are equipped with more radios, the disruption during the restoration (i.e., TDI) may get worse, as more radios need to switch channels.

68

2010/06/08

NTU IMOPLAB

69

ntu im oplab providing survivability against jamming attack for multi-radio multi-channel wireless...

Documents

node v v

jamming traffic

jamming area

network traffic

common channel

channel reassignment

jamming defense strategies

transmission edge e