Secrecy Capacity Scaling of Large-Scale Cognitive Networks Yitao Chen1, Jinbei Zhang1, Xinbing Wang1, Xiaohua Tian1, Weijie Wu1, Fan Fu2, Chee Wei Tan3 1 Dept. of Electronic Engineering, Shanghai Jiao Tong University 2 Dept. of Computer Science and Engineering, Shanghai Jiao Tong University 3 Dept. of Computer Science, City University of Hong Kong Outline Introduction Network Model and Definition Independent Eavesdroppers Colluding Eavesdroppers Conclusion 2 Motivations Security is a major concern in wireless networks Mobile Payment Privacy Virtual Property Military Communication 3 Motivations Cryptographic methods Key distribution Rapid growth of computation power Improvement on decoding technology Physical Layer Security Assume eavesdroppers have infinite computation power Require the intended receiver should have a stronger channel than eavesdroppers Provable security capacity C log(1 SNR ) log(1 SNRe ) 4 Related works Secrecy capacity in large-scale networks Guard zone [9] Artificial noise + Fading gain (CSI needed) [8] Using artificial noise generated by receivers to suppress eavesdroppers’ channel quality [11] Cited from [8] [9] O. Koyluoglu, E. Koksal, E. Gammel, “On Secrecy Capacity Scaling in Wireless Networks”, IEEE Trans. Inform. Theory, May 2012. [8] S. Vasudevan, D. Goeckel and D. Towsley, “Security-capacity Trade-off in Large Wireless Networks using Keyless Secrecy,” in Proc. ACM MobiHoc, Chicago, Illinois, USA, Sept. 2010. [11] J. Zhang, L. Fu, X. Wang, “Asymptotic analysis on secrecy capacity in large-scale wireless networks,” in IEEE/ACM Trans. Netw., Feb. 2014. 5 Motivations Limited spectrum resources and CR networks Key questions: What is the impact of security in cognitive networks? What is the performance we can achieve? 6 Outline Introduction Network Model and Definition Independent Eavesdroppers Colluding Eavesdroppers Conclusion 7 Network Model and Definition – I/III Network Area: a 𝑛 × 𝑛 square Legitimate Nodes 𝑛 primary users {𝑋𝑖 }𝑛 , 𝑚 secondary users {𝑌𝑖 }𝑚 I.I.D Self-interference cancelation [17] adopted CSI unknown Eavesdroppers 𝑛𝜙𝑒 (𝑛) eavesdroppers Location positions unknown CSI unknown Cited from [17] [17] J. I. Choiy, M. Jainy, K. Srinivasany, P. Levis and S. Katti, “Achieving Single Channel, Full Duplex Wireless Communication”, in ACM Mobicom’10, Chicago, USA, Sept. 2010. 8 Network Model and Definition – II/III Random permutation traffic, no cross network traffic Communication Model Physical Model: Primary user i transmits to primary user j Interference from other primary TXs Interference from secondary TXs Interference from other primary RXs Interference from secondary TXs Define the physical model for secondary users and eavesdroppers similarly. 9 Network Model and Definition – III/III Definition of Per Hop Secrecy Throughput: Independent eavesdropper Colluding eavesdroppers Definition of Asymptotic Capacity 𝜆𝑝 𝑛 = Θ(𝑔(𝑛)), if Similarly, we can define the asymptotic per-node capacity for the secondary network 10 Outline Introduction Network Model and Definition Independent Eavesdroppers Colluding Eavesdroppers Conclusion 11 Independent Eavesdroppers Physical Feasibility of Security Primary Networks 𝑝 𝑆𝐼𝑁𝑅𝑖𝑗 ≥ 𝛾𝑝 and Successful transmission 𝑝 𝑆𝐼𝑁𝑅𝑖𝑒 ≤ 𝛾𝑒 No eavesdropper can decode the message Secondary Networks 𝑠 𝑆𝐼𝑁𝑅𝑖𝑗 ≥ 𝛾𝑠 and 𝑠 𝑆𝐼𝑁𝑅𝑖𝑒 ≤ 𝛾𝑒 𝛾𝑒 < min{𝛾𝑝 , 𝛾𝑠 } Operation Rules: • Primary users disregard secondary users; • Secondary users should affect primary users little. 12 Independent Eavesdroppers Intuitive Primary Networks 𝑝 𝑆𝐼𝑁𝑅𝑖𝑒 𝑝𝑟𝑖𝑚𝑎𝑟𝑦 𝑃𝑟𝑒𝑐𝑒𝑖𝑣𝑒𝑟 Concurrent Transmission Range Secrecy Capacity Secondary Networks 𝑠𝑒𝑐𝑜𝑛𝑑𝑎𝑟𝑦 𝑃𝑟𝑒𝑐𝑒𝑖𝑣𝑒𝑟 Unknown ? Good or bad for primary nodes ? Good or bad for eavesdroppers Depend on SUs’ locations 13 Independent Eavesdroppers Primary T-R pair (node i to node j) • For other primary transmitter k and receiver l 𝑝 𝑆𝐼𝑁𝑅𝑖𝑗 ≥ 𝛾𝑝 𝑝 𝑆𝐼𝑁𝑅𝑖𝑒 ≤ 𝛾𝑒 • For other secondary transmitter k and receiver l 14 Independent Eavesdroppers Scheduling scheme Cell Partition Round-Robin Scheduling: • Tessellate the network into cells. • Different cells take turn to transmit. • Secondary users can transmit in non-occupied cells with the guarantee of affecting primary transmissions little. Figure: Simple 9-TDMA 15 Independent Eavesdroppers Routing scheme Highway System – Draining Phase – Highway Phase – Delivery Phase Bottleneck: Highway Phase (nodes need to relay packets for others) Distance of primary T-R pairs is 1. Distance of primary concurrent transmission range is Θ(1). Secrecy Capacity is Θ( 1/𝑛) for primary network. Secrecy Capacity is Θ( 1/m) for secondary network. No order cost comparing to the scenario without security concern! 16 Outline Introduction Network Model and Definition Independent Eavesdroppers Colluding Eavesdroppers Difference with previous case Conclusion 17 Colluding Eavesdroppers SINR of Colluding Eavesdroppers – maximum ratio combining of SINR Bound the SINR of eavesdroppers: Disjoint rings with same size. Eavesdroppers in the same ring has a similar SINR. Artificial noise + Path loss gain + Cooperation 18 Colluding Eavesdroppers Choice of Concurrent Transmission Range k k , artificial noise , throughput k , SINR of eavesdroppers , security 𝑆 𝑑 𝐺 𝑑 = log 1 + 𝑁0 + 𝐼 𝑑 𝑝 ≥ log(1 + ≥ ≥ 𝑃𝑡 1+ 2𝑐 𝑑+1 ) 𝑁0 +𝑏7′ 𝑃𝑡𝑠 +𝑃𝑟𝑠 𝜙𝑠 𝑛 𝑘𝑐 −𝛼 𝑝 𝑏7′′ 𝑃𝑡 1 + 𝑝 𝑏7 𝑃𝑡 𝑑 −𝛼 when choosing 𝑘 = Θ −𝛼 (𝑃𝑟𝑠 𝜙𝑠 𝑛 2𝑐 𝑑 + 1 1 𝛼 −𝛼 ) and 𝑏7 is a constant. 19 Colluding Eavesdroppers Result comparison Cooperation in cognitive networks helps to increase secrecy capacity, compared to stand-alone networks [11]. [11] J. Zhang, L. Fu, X. Wang, “Asymptotic analysis on secrecy capacity in large-scale wireless networks,” to appear in IEEE/ACM Trans. Netw., 2013. 20 Outline Introduction Network Model and Definition Independent Eavesdroppers’ Case Colluding Eavesdroppers’ Case Conclusion 21 Conclusion In this paper, we study physical layer security in cognitive networks. Our scheme adopting self-interference cancellation is very efficient. Cooperation between secondary network and primary network in CR networks can help to strengthen physical layer security. 22 Thank you !