22. Integration of False Data Detection With Data Aggregation and Confidential Transmission in...

7/31/2019 22. Integration of False Data Detection With Data Aggregation and Confidential Transmission in Wireless Sensor Networks

1/47

Introduction Assumptions and Limitations DAA Performance Analysis Conclusion

Integration of False Data Detection

with Data Aggregation

and Confidential Transmission

in Wireless Sensor Networks

S. Ozdemir H. Cam

IEEE

IEEE/ACM Transactions on Networking, 2009

Presented by Gowun Jeong

1 / 4 7
http://find/http://goback/


2/47


Outline

Introduction

Assumptions and Limitations

Data Aggregation and Authentication Protocol (DAA)Step 1: Monitoring Node Selection for an Aggregator

Step 2: Sensor Node Pairing

Step 3: Integration of Secure Data Aggregation and False

Data Detection

Performance Analysis

Conclusion

2 / 4 7
http://find/


3/47


Outline

Introduction





Data Detection


Conclusion

3 / 4 7
http://find/


4/47


Security Vulnerability of Wireless Sensor Networks

Security attacks False Data Injection (FDI)

Compromised nodes (CNs) decrease data integrity.

Data Forgery Eavesdropping

Where FDI by CNs possibly occurs?

Data Aggregation (DA) Data Forwarding (DF)

False data transmission depletes the constrained battery power; and the bandwidth utilisation.

4 / 4 7
http://find/


5/47









5 / 4 7


6/47









6 / 4 7

C
http://find/


7/47









7 / 4 7

I t d ti A ti d Li it ti DAA P f A l i C l i


8/47


False Data Detection (FDD)

Conventional work

Most discussed FDD during DF.

Challenge! Any data change between twocommunicating endpoints is considered as

FDI. Ozdemir and Cams approach

attempts to correctly determine whether any data alterationis due to DA or FDI.

A Data Aggregation and Authentication protocol

against up to T CNs over the encrypted data for FDD both by a data aggregator and by a non-aggregating

node

8 / 4 7

http://find/


9/47



Conventional work







node

9 / 4 7



10/47



Conventional work







node

10/47

http://find/


11/47



Conventional work







node

11/47



12/47


Outline

Introduction





Data Detection


Conclusion

12/47

http://find/


13/47


Basic Assumptions

Network

A densely deployed sensor network of certain large size

Sensor

Overlapping sensing ranges

Role change Sensor nodes rotatively assumes the role of data aggregator.

Limited computation and communication capabilities

Message

Time-stamped Nonce used to prevent reply attacks

Intrusion ways to compromise nodes

Physical capturing Radio communication channel attack

13/47

http://find/


14/47

p y

Network Topology

Data aggregators are chosen in such a way that1. there are at least T nodes, called forwarding nodes, on

the path between any two consecutive data aggregators;and

2. each data aggregator has at least T neighbouring nodes.

14/47

http://find/


15/47

p y

Generation of MACs

Only data aggregators encrypt and decrypt the aggregated

data.

The forwarding nodes first verify data integrity using MACs

and then relay the data if it is not false. Two Full-size MACs (FMACs), each of which consisting of

T + 1 subMACs, for a pair of plain and encrypted data

One computed by a data aggregator T subMACs generated by its T monitoring nodes

The same Pseudo-Random Number Generator (PRNG),termed f

Random numbers between 1 and 32

15/47



16/47

Generation of MACs

subMAC generation of data D by neighbouring node Ni ofdata aggregator Au for its pairmate Fj

1. Establish the shared key Ki,j between Ni and Fj.2. Compute MAC(D) using Ki,j.3. Assuming that S denotes the size of MAC(D) in bits, selects

S/(T + 1) bits to form subMAC(D) using its PRNG and Ki,jas the seed.

subMAC verification of D by Fj for its pairmate Ni1. Compute the MAC(D).2. Run its PRNG S/(T + 1) times to generate subMAC(D)

with Ki,j as the seed.3. Compare two subMAC(D)s.

PRNG synchronisation achieved by packet sequence

numbers

16/47

http://find/


17/47

Key Establishment

Pairwise key establishment Sybil attacks

A compromised node fakes multiple identities to establish

pair relations with more than one monitoring nodes. To prevent from Sybil attacks, a monitoring node can share

a pairwise key with another node in multiple hops.

Group key establishment

Group key Kugroup for data aggregator Au and its

neighbouring nodes is used to select the monitoring nodesand to protect data confidentiality while data transmitting.

17/47



18/47

Limitations

The value of T depends strictly on several factors, such as

geographical area conditions, modes of deployment, and

so on.

The pairwise key establishment between non-neighbouring

nodes takes more time than that between direct

neighbouring nodes.

Compromising only one legitimate group member

discloses not only some or all of the past group keys butalso the current group key.

18/47



19/47

Outline

Introduction





Data Detection


Conclusion

19/47



20/47

Notations used in DAA

20/47

http://find/


21/47

Outline

Introduction


Data Aggregation and Authentication Protocol (DAA)

Step 1: Monitoring Node Selection for an Aggregator



Data Detection


Conclusion

21/47

http://find/


22/47

Algorithm MNS (Monitoring Node Selection)

Table: Choose T monitoring nodes from n neighbouring nodes of Au

1. Au all nodes request two random numbers with node ID

2. Ni Au Ra and Rb generated by f(Ku,i)MACKu,i(Ra | Rb)3. Au all nodes {N1, . . . , Nn} in the receiving order

{R1, . . . , R2n} labeled in an ascending orderMACKugroup(R1 | | R2n)

4-1. Ni Au (verified)EKu,

i

(MACKugroup

(R1 | | R2n))

4-2. Ni Au, Njs (unverified)restart from 1.5. Ni for 1 k T, compute

Ik = [(n1+k

j=k Rj + Ku

group)mod(n)] + 1

to determine T monitoring node IDs of Au

22/47



23/47

Outline

Introduction






Data Detection


Conclusion

23/47

http://find/


24/47

Three Types of Node Pairs

2T + 1 node pairs are formed.

AA-type pair One pair between Au and AfMF-type pair T pairs between M

kof Au

and Fj towards AfMN-type pair T pairs between Mk of Au

and Ni of Af

T Mks selected in Step 1 distinctly choose

their own pairmates to form MF-type andMN-type pairs.

24/47



25/47

Pairmate Selection

1. Af Fj Au pairmate discovery messageNis of AfMACKf,u(Nis)

Fjs IDs for 1 j h2. Au T Mks MACKugroup(F1 | | Fh) for new, random

forwarding node labelingMACKugroup(Nis)s

3. Mk Au one forwarding node

one neighbouring node4. Au T Mks two pairmate lists of size T5. Mk pairmate verification

25/47

http://find/


26/47

Outline

Introduction






Data Detection


Conclusion

26/47



27/47

Data Confidentiality

One pairmate computes a subMAC, and the otherpairmate verifies the subMAC.

subMACs for plain data are used for FDD during DA.

subMACs for encrypted data are used for FDD during DF.

Each data aggregator forms two FMACs as the followingfigure.

Au determines the order of subMACs and informs eachforwarding node about its subMAC location individually.

probability of FDI at a forwarding node = (1/2)32

27/47

http://find/


28/47

Algorithm SDFC

28/47



29/47

Algorithm SDFC

29/47



30/47

Algorithm SDFC

30/47



31/47

Algorithm SDFC

31/47



32/47

Algorithm SDFC

32/47

http://find/


33/47

Algorithm SDFC

33/47

http://find/


34/47

Algorithm SDFC

34/47

http://find/


35/47

Algorithm SDFC

35/47

http://find/


36/47

Outline

Introduction






Data Detection


Conclusion

36/47

http://find/


37/47

Security Analysis of Algorithm SDFC

Lemma 1Assuming that Au is compromised and there are additional at

most T 1 collaborating compromised nodes among theneighbouring nodes of Au and Af, any false data injected by Auare detected by the Afs neighbouring nodes only in SDFC.

Data verification by the monitoring nodes of Au and the

neighbouring nodes of Af

Lemma 2

Assuming that Au and Af are not compromised, any false datainjected by any subset of Aus forwarding nodes are detected by

Af in SDFC.

Data verification by Af

37/47



38/47

Security Analysis of FMAC and subMAC

Changing the size of MAC

Security Level vs. Communication Overhead Probability of FDI at a node = (1/2)32 for 4-byte FMACs

Probability of FDI into a subMAC = (1/2)32/(T+1)

The size of FMAC = T + 1

38/47

http://find/


39/47

Security Analysis of FMAC and subMAC

Changing the size of MAC

Security Level vs. Communication Overhead Probability of FDI at a node = (1/2)32 for 4-byte FMACs

Probability of FDI into a subMAC = (1/2)32/(T+1)

The size of FMAC = T + 1

39/47



40/47

Computational Cost of Algorithm SDFC

Computation Traditional Work SDFCMAC 1 4(T + 1)

= (T + 1) subMACs 2 FMACs a pair

Aggregation 1 T + 1

= 1 by aggregator+ T by monitors

Encryption/ 1 T + 2Decryption = 1 encryption by Au

+ T decryptions by monitors

+1 decryption by A

f

Only the first MAC computation consumes much resource.

Data transmission requires much more energy than data

computing in wireless sensor networks.

40/47

http://find/


41/47

Communication Cost of Algorithm SDFC

DADD the amount (in bytes) of data transmission using ADD of two FMACsDtradAuth the amount (in bytes) of data transmission using the traditional scheme of a MAC

Ltos the length (in bytes) of an authenticated and encrypted data packet

the number of data packets generated by legitimate nodes

the number of false data packets injected by up to T compromised nodes

Hd the average number of hops between two consecutive data aggregators

H the average number of hops that a data packet travels in the network

DADD = (Ltos + 4)(H+ Hd) + T(Ltos + 4)( + ) +4T

T + 1( + )

DtradAuth = LtosH( + )

data transmission by a data aggregator

data transmission by T monitors

subMACs transmission by T monitors

41/47

http://goforward/http://find/http://goback/


42/47









T + 1( + )





42/47



43/47









T + 1( + )





43/47

http://find/


44/47









T + 1( + )





44/47

http://find/


45/47

Cost Comparison

Ltos = 41, H = 50, Hd 12 and / 0.2

Comparing (a) and (b), DADD more mildly increases than

DtradAuth.

(c) shows that the value of T trades off between security

and computation overhead in the network.

(c) also illustrates the impact of data aggregation.

45/47

http://find/


46/47

Outline

Introduction



Step 1: Monitoring Node Selection for an AggregatorStep 2: Sensor Node Pairing


Data Detection


Conclusion

46/47



47/47

Contributions and Future Work

Contributions

False data detection during data aggregation Integration of data confidentiality and false data detection

Less communication overhead (by fixing the size of eachFMAC)

Future work

Security and efficiency improvement in networks whereevery sensor enables data forwarding and aggregation at

the same time

47/47

22. Integration of False Data Detection With Data Aggregation and Confidential Transmission in...

Documents

Transcript of 22. Integration of False Data Detection With Data Aggregation and Confidential Transmission in...