1. Presented by: Dr. Attila Altay Yavuz
Mohamed Grissa, Attila Altay Yavuz and Bechir Hamdaoui
Presented by: Dr. Attila Altay Yavuz

2. Cognitive Radio
Cognitive Radio：Increase channel utilization via dynamic spectrum access
Traditional Spectrum Allocation
Primary User (PU)
PU uses the spectrum exclusively
Cognitive Radio allows dynamic access to the spectrum and introduces 2 types of users:
Primary users: they are licensed users that have the spectrum for their exclusive use. In the traditional allocation policy PUs do not use the spectrum all the time -> waste of spectrum resource.
Secondary users: these are introduced to exploit the idle spectrum whenever the PU is not using it. Once PU is back, SUs must vacate the spectrum.
Cognitive Radio
Secondary User (SU)
SUs can access the idle spectrum
November 13, 2018

3. Cognitive Radio: System Model
SUs can identify the idle spectrum based on:
Non-Cooperative Spectrum Sensing
Cooperative Spectrum Sensing
Distributed
Centralized
(-) Inaccurate decision
(+) simplicity
Simple spectrum sensing is not accurate due to physical phenomena like shadowing, fading, … but it is very simple
Cooperation is added to improve decision accuracy by overcoming the problems of shadowing …
(+) More accurate decision
(-) Complex
(+) More accurate decision
(+) Less complex
FC: Fusion Center, SU: Secondary User, PU: Primary User
November 13, 2018

4. Centralized Cooperative Spectrum Sensing
𝝉 : energy threshold
RSS: received signal strength
Centralized Cooperative Spectrum Sensing
Aggregation-based spectrum availability decision
𝟏 𝒏 𝑹𝑺𝑺𝒊 ? 𝝉
decision
decision
decision
decision
decision
RSS5
RSS4
RSS3
There are two types of approaches to make a decision about spectrum availability in CRNs:
The first type is soft decision rules where FC receives some measurements from Sus, e.g. RSS and computes some statistics over those measurements e.g. aggregation or average and then makes a decision.
RSS2
RSS1
RSS: Received Signal Strength
𝝉 : energy sensing threshold
November 13, 2018
FC: Fusion Center, SU: Secondary User

5. Centralized Cooperative Spectrum Sensing
𝝉 : energy threshold
RSS: received signal strength
Centralized Cooperative Spectrum Sensing
Voting-based spectrum availability decision
Votes count
decision
decision
decision
decision
decision 𝝉 𝝉 𝝉 𝝉 𝝉
𝑹𝑺𝑺𝟓 ? 𝝉 b5
𝑹𝑺𝑺4? 𝝉 b4
𝑹𝑺𝑺3? 𝝉 b3
𝑹𝑺𝑺2? 𝝉 b2
The second type is hard decision rules, e.g. voting, where users make local decisions and send their decisions to FC who combines these decision to make a final decision
𝑹𝑺𝑺1? 𝝉 b1
FC: Fusion Center, SU: Secondary User
November 13, 2018

6. Outline Research Challenge: Location Privacy
Limitations of the State-of-the-Art
Proposed Scheme: Main Idea
Proposed Scheme: Details
Security Analysis
Analysis and Comparison
November 13, 2018

7. Location Privacy Issue
SUs’ sensing reports (e.g., RSS) are highly correlated to their locations
RSS
RSS
Location information can easily be obtained
RSS
Challenge in CRNs: Sensing reports being sent to FC lead to leakage of location information
Without protecting sensing reports, locations of SUs are exposed to FC
November 13, 2018

8. Location Privacy Issue
SUs the CRN
Location information can be used to determine a lot of information about an individual’s beliefs, preferences, and behavior
Despite its importance, little work was drawn to deal with it in the literature
November 13, 2018

9. Outline Research Challenge: Location Privacy
Limitations of the State-of-the-Art
Proposed Scheme: Main Idea
Proposed Scheme: Details
Security Analysis
Analysis and Comparison
November 13, 2018

10. Aggregation-based Approaches
decryption
𝒊=𝟏 𝒏 𝑬(𝑹𝑺𝑺𝒊)
𝑹𝑺𝑺𝒊
𝑬(𝑹𝑺𝑺𝟏)
𝑬(𝑹𝑺𝑺𝟐)
𝑬(𝑹𝑺𝑺𝒏)
𝑬 is an additive homomorphic encryption scheme
(e.g., use DLP [3], Paillier [4], or Elliptic Curve [5])
[3] Shuai Li, Haojin Zhu, Zhaoyu Gao, Xinping Guan, Kai Xing, and Xuemin Shen. Location privacy preservation in collaborative spectrum sensing. In INFOCOM, 2012 Proceedings IEEE, pages729–737. IEEE, 2012.
[4] L. Chen, R. Lu, and Z. Cao, “PDAFT: A privacy-preserving data aggregation scheme with fault tolerance for smart grid communications,” Peer-to-Peer Networking and Applications, pp. 1–11, 2014.
[5] Koblitz, N. (1987). Elliptic curve cryptosystems. Mathematics of computation,48(177),
November 13, 2018

11. Limitations of Aggregation-based Methods
Prone to differential privacy attack
SUs may frequently join or leave
𝑹 = 𝐷𝑒𝑐( E 𝑹𝑺𝑺𝟏 + E 𝑹𝑺𝑺𝟐 + E 𝑹𝑺𝑺𝟑 )
𝑹′ = 𝐷𝑒𝑐( E 𝑹𝑺𝑺𝟏 + E 𝑹𝑺𝑺𝟐 + E 𝑹𝑺𝑺𝟑 +E 𝑹𝑺𝑺𝟒 )
𝑹′−𝑹=𝑹𝑺𝑺𝟒 location disclosed
No fault tolerance (e.g. [3])  All reports are needed
Substantial overhead: needing costly crypto operations or solving DLP problem [3,4]
Some cannot handle network dynamism [3] (i.e., multiple users leaving/joining)
[3] Shuai Li, Haojin Zhu, Zhaoyu Gao, Xinping Guan, Kai Xing, and Xuemin Shen. Location privacy preservation in collaborative spectrum sensing. In INFOCOM, 2012 Proceedings IEEE, pages729–737. IEEE, 2012.
[4] L. Chen, R. Lu, and Z. Cao, “PDAFT: A privacy-preserving data aggregation scheme with fault tolerance for smart grid communications,” Peer-to-Peer Networking and Applications, pp. 1–11, 2014.
November 13, 2018

12. Outline Research Challenge: Location Privacy
Limitations of the State-of-the-Art
Proposed Scheme: Main Idea
Proposed Scheme: Details
Security Analysis
Analysis and Comparison
November 13, 2018

13. Contribution: Attributes of Our Scheme (LPOS)
Low computational overhead
Logarithmic expensive operations versus linear
Low communication overhead
Smaller tags w.r.t. aggregation-based methods
High location privacy
No differential privacy attack, RSS values are protected
Fault tolerance and robustness against network dynamism
Scalability
November 13, 2018

14. The Main Idea: Objectives
To recap: Existing approaches rely on comparison operations: FC
1) Aggregation: FC compares ( 𝟏 𝒏 𝑹𝑺𝑺𝒊 ,𝝉) and sends decision to SUs
𝑹𝑺𝑺𝒊 𝝉
SUi
2) Voting: SUi compares (𝑹𝑺𝑺𝒊 ,𝝉) and sends vote back to FC to decide
In 1): 𝑹𝑺𝑺𝒊 is disclosed to FC via DLP  location disclosed
In 2): 𝝉 is disclosed to SUs  prone to malicious users
Ideal solution: Enable comparisons of 𝝉 and RSSs
Without disclosing 𝝉 and RSS and with
Minimal overhead and fault tolerance
Our goal
November 13, 2018

15. The Main Idea: Basic SchemeFC
SUi FC 𝝉 𝝉 𝝉 𝝉
𝑹𝑺𝑺𝒊 𝒃𝟏 𝒃𝒏 𝒃𝟐
𝒏 times YM YM YM
𝑹𝑺𝑺𝟏
𝑹𝑺𝑺𝟐
𝑹𝑺𝑺𝒏
SU1
SU2
SUn
𝒃←𝝉 ≤?𝑹𝑺𝑺𝒊
Linear # of PKC operations not scalable
is a secure comparison protocol e.g. Yao Millionaire(YM) Only b but nothing else learned
We need to make it sub-linear for better scalability
November 13, 2018

16. The Main Idea: O(n)  𝑙𝑜𝑔 2 (n)
What if FC knows the relative order of RSSs but nothing else ?
𝐄_K(𝑹𝑺𝑺𝟏)≤𝐄_K(𝑹𝑺𝑺𝟐) ≤ ⋯≤ 𝐄_K(𝑹𝑺𝑺𝒏)
Reduces # of secure comparisons drastically
We can use a binary search-like secure comparison strategy 𝛰 𝑛 →𝑙𝑜𝑔2(𝑛)
November 13, 2018

17. The Main Idea: O(n)  𝑙𝑜𝑔 2 (n)
Benefit of knowing relative order 𝝉 𝒊
𝐄_K(𝑹𝑺𝑺𝟏)≤𝐄_K(𝑹𝑺𝑺𝟐) ≤ ⋯≤𝐄_K 𝑹𝑺𝑺𝒏/2 ≤⋯≤𝐄_K 𝑹𝑺𝑺𝒊 ≤ 𝐄_K(𝑹𝑺 𝑺 𝒊+𝟏 )≤⋯≤ 𝐄_K(𝑹𝑺𝑺𝒏)
YM(𝑹𝑺𝑺𝟏,𝝉)
YM( 𝑹𝑺𝑺 𝒏/𝟐 ,𝝉)
YM(𝑹𝑺𝑺𝒊,𝝉)
YM( 𝑹𝑺𝑺 𝒊+𝟏 ,𝝉)
YM( 𝑹𝑺𝑺 𝒏 ,𝝉)
Compare with min and max values
Otherwise, use a binary search-like strategy to find the position of 𝝉
FC learns that 𝒊 users have RSS smaller/greater than 𝝉 : Worst-case 𝑙𝑜𝑔 2 (𝑛) comparisons
FC compares 𝒊 to a threshold 𝝀 (we used optimal expression with half-voting rule)
November 13, 2018

18. Main Idea: How to Obtain Relative Order?
Order Preserving Encryption (OPE): Deterministic symmetric encryption (e.g., [8]) whose encryption function preserves numerical ordering of the plaintexts.
OPE uses the same key to preserve the order
We use TGECDH [7] protocol to establish a group key among SUs
Scales logarithmically (i.e., 𝑙𝑜𝑔 2 𝑛 ), rarely executed (per-sensing period, join/leave)
𝑷𝟏 ≤ 𝑷𝟐 ≤ ⋯ ≤ 𝑷𝒏
OPEK
OPEK
OPEK
𝑪𝟏 ≤ 𝑪 𝟐 ≤ ⋯ ≤ 𝑪𝒏
[7] Wang, Y., Ramamurthy, B., & Zou, X. (2006, June). The performance of elliptic curve based group Diffie-Hellman protocols for secure group communication over ad hoc networks. In Communications, ICC'06. IEEE International Conference on (Vol. 5, pp ). IEEE.
[8] Alexandra Boldyreva, Nathan Chenette, Younho Lee, and Adam O’neill. Order-preserving symmetric encryption. In Advances in Cryptology-EUROCRYPT 2009, pages 224–241. Springer, 2009
November 13, 2018

19. Outline Research Challenge: Location Privacy
Limitations of the State-of-the-Art
Proposed Scheme: Main Idea
Proposed Scheme: Details
Security Analysis
Analysis and Comparison
November 13, 2018

20. Location Privacy for Optimal Sensing LPOS𝝉
𝑟 𝑚𝑖𝑛
𝑟 5
𝑟 3
𝑟 7
𝑟 4
𝑟 8
𝑟 2
𝑟 𝑚𝑎𝑥 1
𝑟 𝐼−1
November 13, 2018

21. Outline Research Challenge: Location Privacy
Limitations of the State-of-the-Art
Proposed Scheme: Main Idea
Proposed Scheme: Details
Security Analysis
Analysis and Comparison
November 13, 2018

22. Security Analysis IND-CPA IND-OCPA
Theorem: LPOS leaks no information on ( 𝑹𝑺𝑺 𝒊 𝒋 𝒊=𝟏,𝒋=𝟏 𝒏,𝒍 ,𝝉) beyond IND-CPA secure 𝑽 𝒋 𝒋=𝟏 𝒍 . IND-OCPA secure order of tuple ( 𝒁 𝒋 =𝑶𝑷𝑬. 𝑬 𝑲 𝒋 𝑹𝑺𝑺 𝟏 𝒋 ,⋯,𝑶𝑷𝑬. 𝑬 𝑲 𝒋 𝑹𝑺𝑺 𝒏 𝒋 𝒊=𝟏,𝒋=𝟏 𝒏,𝒍 ) and 𝒃 𝒊 𝒋 𝒊=𝟏,𝒋=𝟏 𝒏,𝒍 to FC.
Proof
History lists 𝑳𝟏= 𝑽 𝒋 𝑳𝟐=( 𝒃 𝒊 𝒋 𝒊=𝟏 𝒏 , 𝑽 𝒋 , 𝒁 𝒋 ) where 𝑽 𝒋 = 𝒄𝒉𝒏 𝒊 𝒋 𝒊=𝟏 𝒏
Information history observed by all entities in the system for 𝒋=𝟏,…𝒏: 𝑽 𝒋 and 𝒁 𝒋
IND-CPA
IND-OCPA
Any membership status update: TGECDH guarantees key independence property: 𝑲 𝒋
November 13, 2018

23. Outline Research Challenge: Location Privacy
Limitations of the State-of-the-Art
Proposed Scheme: Main Idea
Proposed Scheme: Details
Security Analysis
Analysis and Comparison
November 13, 2018

24. Analysis and Comparison
LPOS is robust against differential privacy, unlike [3,4]
LPOS is fault tolerant and supports dynamism for multiple users, unlike [3]
Has better sensing performance thanks to the half-voting rule, unlike [3,4]
November 13, 2018
[3] Shuai Li, Haojin Zhu, Zhaoyu Gao, Xinping Guan, Kai Xing, and Xuemin Shen. Location privacy preservation in collaborative spectrum sensing. In INFOCOM, 2012 Proceedings IEEE, pages729–737. IEEE, 2012.
[4] L. Chen, R. Lu, and Z. Cao, “PDAFT: A privacy-preserving data aggregation scheme with fault tolerance for smart grid communications,” Peer-to-Peer Networking and Applications, pp. 1–11, 2014.

25. Thank you!
November 13, 2018

26. Appendix
November 13, 2018

27. Half Voting Rule [3]
It has a better sensing performance than aggregation-based rules
It does not expose users to the privacy issues, we mentioned earlier.
optimal spectrum sensing in voting-based CRNs that uses a voting threshold 𝝀.
≤|≥𝝀 ? 𝝉 1
𝑟 𝑚𝑖𝑛
𝑟 5
𝑟 3
𝑟 7
𝑟 𝐼−1
𝑟 8
𝑟 2
𝑟 𝑚𝑎𝑥
[3] Wei Zhang, Ranjan K Mallik, and Khaled Letaief. Cooperative spectrum sensing optimization in cognitive radio networks. In Communications, ICC’08. IEEE International Conference on, pages 3411–3415. IEEE, 2008.
November 13, 2018

28. Yao’s Millionaires’ Protocol
Initially proposed to solve the millionaires problem
Alice has value x
Bob has value y
Alice initiates the protocol to know whether x < y or x ≥ y
Only Alice knows the outcome 𝒃∈{𝟎,𝟏}
Alice doesn’t know y
Bob doesn’t know x
We use the protocol proposed by [9] which uses multiplicative homomorphic encryption (El Gamal).
Alice 𝒚 𝒙
Bob
𝒃=𝒙 ≤𝒚 YM
[9] Hsiao-Ying Lin and Wen-Guey Tzeng. An efficient solution to the millionaires’ problem based on homomorphic encryption. In Applied Cryptography and Network Security, pages 456–466. Springer, 2005.
November 13, 2018

29. [6] http://www.keylength.com/
November 13, 2018

30. Security Assumptions and Objectives
Semi-honest (i.e., honest-but-curious) model
FC may show interest in learning RSS values
SU may show interest in learning \tau
No collusion between entities
Objectives:
Hide RSS values from FC
Hide \tau from SUs
November 13, 2018