1

Optimal Anti-Jamming Strategy in Sensor Networks

Presenter: Yu Wen ChiouAdvisor: Yeong Sung Lin

Communications (ICC), 2012 IEEE International Conference

2

Agenda

I. IntroductionII. Model and Problem FormulationIII. Optimal Anti-jamming StrategyIV. Performance EvaluationV. Conclusion

3

I. Introduction

4

Introduction

• Wireless Sensor Network 的特性• Low-cost• Easy to deploy• Unattended operation• Ability of withstanding harsh environmental conditions

• Wireless Sensor Network 的應用• Environment monitoring• Event detection

• Wireless Sensor Network 的關鍵問題• Sensor nodes are typically powered by batteries and hence limited in power supply.

5

Introduction(Cont’d)

• Radio Jamming Attacks• Sensor networks transmit wireless signals over the open shared media. This

leaves a sensor network vulnerable to radio jamming attacks.• Ex, Corrupt control packets, the jammer just keeps sending packets like RTS

to prevent transmission of legitimate packets.• Cause severe damage to the sensor network with only modest overhead.

• Anti-Jamming Methods• Energy-efficient• Ex, channel surfing, error correction codes, transmission power adjustment

6

Introduction(Cont’d)

• 研究動機• These existing countermeasures of jamming attacks are usually suitable for a

limited range of jamming conditions with varying operation cost. In the real world scenario, jamming attacks may be very different in nature and may change over time.• Different nodes suffer different degrees of radio jamming. Thus, it is

inefficient for a whole sensor network simply to apply a single anti-jamming technique. This may result in poor performance of anti-jamming and/or still suffer serious performance degradation of energy consumption.

7

Introduction(Cont’d)

• 研究目的• This paper focuses on proposing an adaptive approach to anti-jamming for

sensor networks, by combining the strength of different anti-jamming techniques.• We are formulating the anti-jamming problem of the sensor network as a

Markov decision process and propose an algorithm for computing the best anti-jamming strategy.

8

II. Model and Problem FormulationA. Problem DescriptionB. Anti-jamming Techniques

9

Model and Problem Formulation

A. Problem DescriptionN A set of nodes.

J A set of a few jammers in the environment.

K For each node, there are anti-jamming techniques available for different jamming conditions.Channel condition.

τ Node 定期評估 channel condition 的週期 (period)

( )n t

Based on , the node chooses the proper anti-jamming technique to deal with different jamming signal.

( )n t

10

Model and Problem Formulation(Cont’d)

• For the node, each anti-jamming technique has different cost, . We evaluate this cost as a function of the energy that the node will consume in the next period.

Ck

( ) ( )( ( )) (1)k t k tC f E

時間點 t 時，所選用的 anti-jamming 技術 k

Node 定期評估 channel condition 的週期能耗函數

11

Model and Problem Formulation(Cont’d)

• The performance reward of the node using specific anti-technique is the improvement of the communication between the nodes.

( )R ( ( )) ( ) ( ) (2)k t Q t Q t Q t

時間點 t 時，所選用的 anti-jamming 技術 k ，所帶來的 performance reward

時間點 t 時的 communication quality時間點 t 過了一個週期的時間點

12

Model and Problem Formulation(Cont’d)

• 本文的目的是要為每個 node 選用一個合適的 anti-jamming 技術，選用的準則必須同時考量上述第 (1), (2) 式的 cost 及 reward 。• We make the objective by minimizing the product of cost and

reward.• We focus on a relatively long time period, denoted as T, so the nodes

totally make [T/τ] decisions./

( ) ( )( ) 0

min ( ( )) (3)i i

i

T

k t k tk t i

R Q t C

The set of anti-jamming techniques

13

Model and Problem Formulation(Cont’d)

B. Anti-jamming Techniques• Without loss of generality, this paper assumes that each sensor node is

capable of applying three representative anti-jamming techniques.1. Transmission Power Adjustment: With this technique, a sender

node increases its transmission power, and thus increases the SNR at the receiver node. This technique is suitable under a slight jamming condition. This technique introduces modest energy cost.

2. Error-Correcting Code(ECC): An error-correcting code is used for correcting some error bits that occurred during transmission.

14

Model and Problem Formulation(Cont’d)

3. Channel Hopping(Channel Surfing): With this technique, a sensor node will change the working channel when it detects strong jamming signals in the current channel.

15

Model and Problem Formulation(Cont’d)

• 每一個節點都會有一組 anti-jamming 技術可供應用，表示成：0 1 2 3={ , , , }

0

1

2

3

: Null technique: Transmission power adjustment: Error-correcting code: Channel hopping(Channel surfing)

16

III. Optimal Anti-jamming StrategyA. System StateB. Transition ProbabilityC. Cost of Anti-jamming TechniquesD. Policy DeterminationE. Algorithm Framework

17

Optimal Anti-jamming Strategy

• 在此章節中，將仔細描述，如何為每個 node 選用最佳的 anti-jamming 技術。• We formulate the problem as a 4-tuple:

(S, , , )P C

( , ')

S

P s s

C

: the set of all the node states that describe the channel conditions (3-A)

: the set of anti-jamming techniques (2-B)

: the probability that the node state becomes s’ after technique δ is performed at state s. (3-B)

: the cost of the anti-jamming technique (3-C)

18

Optimal Anti-jamming Strategy(Cont’d)

A. System States (S)

• 考量到 sensor nodes 的有限運算能力，因此降低演算法的複雜度是非常重要的。故本文僅使用四種狀態，表示成 S = {0, 1, 2, 3} 。這四種狀態表示四種不同的 jamming conditions ，及其所對應的 anti-jamming strategies ( 前述之 Δ ={δ0, δ1, δ2, δ3}) 。• 根據過往的文獻，使用 PDR 及 RSSI 來表示 jamming signals ，本文亦使用 PDR 及 RSSI 作為判斷 S 的依據。• 當 S 值越大時，表示 jammer 出現； S 值越小時，表示 jammer 未出現。

19

Optimal Anti-jamming Strategy(Cont’d)

• PDR and RSSI  Value Descriptions Anti-jamming Technique State (0, 1, 2, 3)

PDR Low   Choose the worse case Bigger state

High   Prefer a light anti-jamming strategy

Smaller state

High Enough   The current anti-jamming strategy is effective.

Current state

RSSI Low(Below normal value)

The link is weak   Normal state

Certain level Normal case   Current state

High Jammer is present   Bigger state

20

Optimal Anti-jamming Strategy(Cont’d)

These two functions output the system state depending on the current countermeasure

(PDR < Φ or RSSI > K)

PDR 值大時，表示jammer 未出現， prefer a light anti-jamming strategyRSSI 值小時，表示jammer 未出現， normal state

Jammer 出現，使state 值為大

21

Optimal Anti-jamming Strategy(Cont’d)

B. Transition Probability (P)

• Since the state of nodes changes due to the varying jamming conditions, the transition probability of the states also describes the variation of jamming signals.• The transition probability is acquired by analyzing historical data.

22

Pr( (t ), S(t), )P (S( ), ( )) (4)Pr( ( ),

//

i j i i j

i i i

St S tS t

k k kk k k

Optimal Anti-jamming Strategy(Cont’d)

• When S( t - τ ) = i and S(t) = j. Then, the transition probability can be calculated as:

The total number of nodes that reach the state i

運用 δ 技術時， state 為 i 的node 中，有多少比例會由state i 變為 state j

i j

23

Optimal Anti-jamming Strategy(Cont’d)

C. Cost of Anti-jamming Techniques (C)

1. Adjusting Transmission power It is the raised power multiplied the packet transmission time.

1 1

/ (5)n n

I tx pkt tx pkt bitsi i

C P t P L R

The cost of the increasing power action.

The packet length.

The bits transmission rate.

The time of transmitting every packet.

傳送每個封包所消耗的能量

在週期 τ 內的封包數

24

Optimal Anti-jamming Strategy(Cont’d)

2. Error-Correcting Codes(ECC)• The energy consumed by this technique is the power that used for the

encoding and decoding process and transmitting the redundant bits. • For the error-correcting codes has to be undertaken by both of the

communicating nodes, the notification process also consume extra energy.

1

( / ) (6)n

EC tx EC bits dec notii

C P L R E E

The length of the encoded packets.

The energy that spend for decoding the packets or correcting the errors of the received signal.

The energy expended for the notification process

傳送每個封包所消耗的能量

25

Optimal Anti-jamming Strategy(Cont’d)

3. Channel Hopping(Channel Surfing)• As the error-correcting code strategy, the nodes have to notify the

neighbors that it will change to another channel, for the neighbors undertaking the same strategy to keep the connectivity of the network.• When the nodes take this strategy, the intermediate nodes also need to

send some packets when it switches to each channel.

CS1 1

C = / / (7)m n

tx noti bits tx pkt bits notii i

P L R P L R E

The energy expended for the notification process

The node has to inform the neighbors when it goes to a new channel. This is the energy expended to send those packets

The total channel switches

26

Optimal Anti-jamming Strategy(Cont’d)

D. Policy Determination• In the following, we use PDR to describe performance reward,

• And we define:

( ) ( ( )) ( ) ( ) (8)tR Q t PDR t PDR t

( )1 ( ( )), (9)(10)

t

i i

R S tC

the cost of technique δ

at state i.

同前述第 (2) 式，performance reward

γ 是 cost 的係數。這個係數設計的目的是為了調整使用這個科技的成本，調整時會參考使用這個科技對於質量的好壞影響。

Ex. More effective technique → R 大 → γ 小 → λ 小

27

Optimal Anti-jamming Strategy(Cont’d)

• Based on the definitions above, we devise the policy improvement algorithm to solve the MDP problem.• 我們令 為 node 從 state i 開始，演化 n 週期後的預期總成本。

Then we have:

1

1

(D) ( , ) ( ), 1, 2..., . (11)M

n ni i j

j

P i j D for i M

policyperiod

state

ni

The cost of the next n period

The cost introduced in the first period. (10)

28

Optimal Anti-jamming Strategy(Cont’d)

• The long run expected average cost per unit time could be expressed as:

1

(D)= (12)M

i ii

The steady distribution of the states

There are M states in total

29

Optimal Anti-jamming Strategy(Cont’d)

• We can have an approximate relationship when n is large:

( ) ( ) ( ) (13)ni iD n D D

The effect on the total expected cost due to beginning in state i

The long run expected average cost per unit time (12)

node 從 state i 開始，演化 n 週期後的預期總成本，同第 (11) 式

Optimal Anti-jamming Strategy(Cont’d)

30

• After substituting (13) into (11), we get:

1

1

(D) ( , ) ( ), 1, 2..., . (11)M

n ni i j

j

P i j D for i M

( ) ( ) ( ) (13)ni iD n D D

代入

1

(D)= + ( , ) ( ) (D), 1, 2, ..., . (14)M

j ij

P i j D i M

The long run expected average cost per unit time

Optimal Anti-jamming Strategy(Cont’d)

• The policy improvement algorithm starts by choosing an arbitrary policy and set . Then, it solves (14) to

We use to find another policy such that for each state i:

Where . When and are identical, this iteration process will stop. Otherwise, it sets n = n + 1 and this process continues.

nD ( ) 0M nD 1( ), ( )n nD D

2 ( ),..., ( ).n M nD D ( )i nD 1nD

1

min ( ) ( ) ( ) (15)M

i ij j ij

P D D

1( )i nd D 1nD nD

Optimal Anti-jamming Strategy(Cont’d)E. Algorithm Framework• The framework of the optimal anti-jamming algorithm is shown in Algorithm 2.

It updates the transition probabilities, which is for later policy determination.

The nodes choose a proper anti-jamming strategy based on the current policy

Algorithm 1.

the nodes begin to communicate with each other

It will send a notification to make sure the nodes are using the same anti-jamming strategy

33

IV. Performance EvaluationA. Metrics and SettingB. Performance Results

34

Performance Evaluation

• The nodes are uniformly distributed in an square area of 40m by 40m. • The distance between nodes are 5 meters.• The jammer placed in the middle of the network so that it jams as many

nodes as possible.• The packet length is 25 Bytes, and is 50 Bytes.• These values comply with the IEEE 802.15.4 Standard.

A. Metrics and Setting

ACKL dataL

35

Performance Evaluation(Cont’d)

• Figure 2 shows the effectiveness of the computed anti-jamming strategies.• Channel Surfing 及 Optimal

Strategy 在不同的 jamming conditions 下，都有很好的表現。

B. Performance Results Packet Delivery Ratio

Distance from the jammer

36

Performance Evaluation(Cont’d)

• Figure 3 shows the energy consumption of the nodes.• Optimal Strategy 的能量消耗較

Channel Surfing 為低。• With Figure 2 and Figure 3, we

conclude that our design is effective and efficient at the same time.

37

V. Conclusion

38

Conclusion

• We propose an approach for combining the strength of several jamming countermeasures and allow a sensor node to adopt the best anti-jamming technique.• Sensor nodes in the sensor network can adaptively change their anti-

jamming methods as the jamming condition changes over time.• The comprehensive simulation experiments have demonstrated that our

algorithm achieves good performance in terms of successful delivery rate and at the meanwhile consumes slightly more energy.