Asymptotic Analysis on Secrecy Capacity in Large-Scale Wireless Networks Jinbei Zhang, Luoyifu, Xinbing Wang Department of Electronic Engineering Shanghai Jiao Tong University Aug. 13, 2013 Outline Introduction Motivations Related works Objectives Network Model and Definition Secrecy Capacity for Independent Eavesdroppers Secrecy Capacity for Colluding Eavesdroppers Discussion Conclusion and Future Work Asymptotic Analysis on Secrecy Capacity in Large-Scale Wireless Networks 2 Motivations Secrecy is a Major Concern in Wireless Networks. Mobile Phone Wallet Military networks … Asymptotic Analysis on Secrecy Capacity in Large-Scale Wireless Networks 3 Related works – I/II Properties of Secrecy Graph Cited From [5] Cited From [5] [4] M. Haenggi, “The Secrecy Graph and Some of Its Properties”, in Proc. IEEE ISIT, Toronto, Canada, July 2008. [5] P. C. Pinto, J. Barros, M. Z. Win, “Wireless Secrecy in Large-Scale Networks.” in Proc. IEEE ITA’11, California, USA, Feb. 2011. Asymptotic Analysis on Secrecy Capacity in Large-Scale Wireless Networks 4 Related works – II/II Secrecy Capacity in large-scale networks, Mobile Networks [16] Guard Zone [13] Artificial Noise+Fading Gain(CSI needed) [12] Cited from [12] [16] Y. Liang, H. V. Poor and L. Ying, “Secrecy Throughput of MANETs under Passive and Active Attacks”, in IEEE Trans. Inform. Theory, Vol. 57, No. 10, Oct. 2011. [13] O. Koyluoglu, E. Koksal, E. Gammel, “On Secrecy Capacity Scaling in Wireless Networks”, submitted to IEEE Trans. Inform. Theory, Apr. 2010. [12] 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. Asymptotic Analysis on Secrecy Capacity in Large-Scale Wireless Networks 5 Objectives Several questions arise: CSI information is difficult to obtain Artificial noises also degrade legitimate receivers’ channels Cost on capacity is quite large to utilize fading gain How to effectively improve the secrecy capacity? Asymptotic Analysis on Secrecy Capacity in Large-Scale Wireless Networks 6 Outline Introduction Network Model and Definition Secrecy Capacity for Independent Eavesdroppers Secrecy Capacity for Colluding Eavesdroppers Discussion Conclusion and Future Work Asymptotic Analysis on Secrecy Capacity in Large-Scale Wireless Networks 7 Network Model and Definition – I/II Legitimate Nodes: Self-interference cancelation[16] adopted 3 antennas per-node CSI information unknown Eavesdroppers: Location positions unknown CSI information 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. Asymptotic Analysis on Secrecy Capacity in Large-Scale Wireless Networks 8 Network Model and Definition – II/II Network Model: Extended networks: n nodes randomly distributed in a network with size n. Static Physical channel model SINRij SINRie Plt ( xi , x j ) N 0 kT \{i} Plt ( xk , x j ) kR \{i} Pr l ( xk , x j ) Plt ( xi , xe ) l ( xi , x j ) min(1, dij ) where N0 kT \{i} Plt ( xk , xe ) kR Pl ( x , x ) r k e Definition of secrecy capacity Rijs Rij Rie log2 (1 SINRij ) log 2 (1 SINRie ) where SINRie maxe SINRie Asymptotic Analysis on Secrecy Capacity in Large-Scale Wireless Networks 9 Outline Introduction Network Model and Definition Secrecy Capacity for Independent Eavesdroppers Lower Bound Upper Bound Secrecy Capacity for Colluding Eavesdroppers Discussion Conclusion and Future Work Asymptotic Analysis on Secrecy Capacity in Large-Scale Wireless Networks 10 Independent Eavesdroppers Capacity for Eavesdroppers Lemma 1: When a legitimate node t is transmitting to a legitimate receiver r, the maximum rate that an independent eavesdropper e can obtain is upper-bounded by Pd P Re min( t te , t (1 dtr ) ) N0 Pr Received Power where dtr is the Euclidean distance between legitimate node t and node r and dte is the distance between legitimate node t and eavesdropper e. Asymptotic Analysis on Secrecy Capacity in Large-Scale Wireless Networks 11 Independent Eavesdroppers Case 1: when dte and d re both greater 1, Plt ( xt , xe ) SINRe N 0 kT \{i} Plt ( xk , xe ) kR Pr l ( xk , xe ) Plt ( xt , xe ) Pd = t te Pl Prdre r ( xr, xe ) Pd Pt t te (1 d ) tr Pr( drt dte ) Pr Case 2-4: dte 1, dre 1; dte 1, dre 1; dte 1, dre 1. Asymptotic Analysis on Secrecy Capacity in Large-Scale Wireless Networks 12 Independent Eavesdroppers Capacity for Legitimate Nodes Lemma 2: When a legitimate node t is transmitting to a legitimate receiver which is located d cells apart, the minimum rate that the legitimate node can receive is lower-bounded by c2 Pd , where c 2 t is a constant. S (d ) ) N 0 I (d ) Pt (c(d 1)) log(1 ) N 0 c1 ( Pt Pr )(kc) c2' Pt (c(d 1)) c2 Pd t R(d ) log(1 1 when choosing k ( Pr ) and c 2 is a constant. Asymptotic Analysis on Secrecy Capacity in Large-Scale Wireless Networks 13 Independent Eavesdroppers Secrecy Capacity for Each Cell Theorem 1: For any legitimate transmitter-receiver pair which is spaced at a distance of d cells apart, there exists an Rs (d ) (d 4 ), so that the receiver can receive at a rate of Rs (d ) securely from the transmitter. 1 ( R (d ) Re ) 2 (k d ) Pt 1 (c2 Pd c3 d ) t 2 (k d ) Pr Rs (d ) Choose P 2 c3 d 2 r c2 1 k ( Pr ) (d 2 ) Rs (d ) (d 4 ) Asymptotic Analysis on Secrecy Capacity in Large-Scale Wireless Networks 14 Independent Eavesdroppers Highway System Draining Phase Highway Phase Delivery Phase Rs (d ) (d 4 ) Theorem 2: With n legitimate nodes poisson distributed, the achievable per-node secrecy throughput under the existence of 1 independent eavesdroppers is ( ) . n Asymptotic Analysis on Secrecy Capacity in Large-Scale Wireless Networks 15 Independent Eavesdroppers Optimality of Our Scheme Theorem 2: When n nodes is identically and randomly located in a wireless network and source-destination pairs are randomly chosen, the per-node throughput λ(n) is upper bounded by ( 1 ) . n [18] P. Gupta and P. Kumar, “The Capacity of Wireless Networks”, in IEEE Trans. Inform. Theory, Vol. 46, No. 2, pp. 388-404, Mar. 2000. Asymptotic Analysis on Secrecy Capacity in Large-Scale Wireless Networks 16 Outline Introduction Network Model and Definition Secrecy Capacity for Independent Eavesdroppers Secrecy Capacity for Colluding Eavesdroppers Lower Bound Upper Bound Discussion Conclusion and Future Work Asymptotic Analysis on Secrecy Capacity in Large-Scale Wireless Networks 17 Colluding Eavesdroppers Eavesdroppers Collude Assume that the eavesdropper can employ maximum ratio combining to maximize the SINR which means that the correlation across the antennas is ignored. Theorem 4: If eavesdroppers are equipped with A(n) antennas, the 2 s( n ) is( 1 A( n) ) . per-node secrecy capacity n Asymptotic Analysis on Secrecy Capacity in Large-Scale Wireless Networks 18 Colluding Eavesdroppers Eavesdroppers Collude Assume that each eavesdropper equipped with one antenna and different eavesdroppers can collude to decode the message. Pd P t te Re min( , t (1 dtr ) ) N0 Pr SI NRe SI NRei j e 1 SI NR1j i 2 j ei 2f ( n ) e( n )SI NRe1 SI NRi j 2f ( n) ( n)SI NR e i 2 P 2f ( n ) e( n ) t (1 d r t ) Pr ei Pr 2f ( n ) e( n ) t i 1 N0 i 2 Pt Pt 2 2 e( n ) ( r (1 d r t ) r1 i Pr N0 i 1 2 1 2 ) Asymptotic Analysis on Secrecy Capacity in Large-Scale Wireless Networks 1 Colluding Eavesdroppers Lower Bound Theorem 5: Consider the wireless network B where legitimate nodes and eavesdroppers are independent poisson distributed with parameter 1 eand ( n) respectively, the per-node secrecy capacity is ( s( n) ( 1 n 1 n e( n) l og 2 2 2 2 ) , e( n) ( l og 2 n) , e( n) O( l og Asymptotic Analysis on Secrecy Capacity in Large-Scale Wireless Networks n) n) 20 Colluding Eavesdroppers Lower Bound Lemma 5: When the intensity of the eavesdroppers is e( n )=O( n ) for any constant β>0, partitioning the network into disjoint regions with Nei the number of nodes inside constant size c and denoting by region i, we have where P( Nei v , i ) 1 1 v +1 Theorem 6: If eavesdroppers are poisson-distributed in the network with intensity e( n) O( n ) for any constant β>0, the per-node secrecy capacity is ( 1 ) . n Asymptotic Analysis on Secrecy Capacity in Large-Scale Wireless Networks 21 Colluding Eavesdroppers Upper Bound Nei 1 e( n )k 2 2 SI NRe N J 0 Sj Ij 3k ) 4 Nei N0 I j ( c11 e( n )k 2 k e( n ) 1 2 s( n) ( Asymptotic Analysis on Secrecy Capacity in Large-Scale Wireless Networks 1 k 2 ) n 22 Colluding Eavesdroppers Upper Bound k 2 e( n)k (1) k e( n ) s( n) ( 1 2 1 k 2 ) n Asymptotic Analysis on Secrecy Capacity in Large-Scale Wireless Networks 23 Colluding Eavesdroppers Upper Bound Theorem 7: Consider the wireless network B where legitimate nodes and eavesdroppers are independent e( n ) respectively, poisson distributed with parameter 1 and the per-node secrecy capacity is O( s( n ) O( 1 n 1 n e( n ) 2 2 ) , e( n ) (1) ) , e( n ) O(1) Asymptotic Analysis on Secrecy Capacity in Large-Scale Wireless Networks 24 Outline Introduction Network Model and Definition Secrecy Capacity for Independent Eavesdroppers Secrecy Capacity for Colluding Eavesdroppers Discussion Conclusion and Future Work Asymptotic Analysis on Secrecy Capacity in Large-Scale Wireless Networks 25 Discussions Secrecy Capacity in Random Networks Random networks: total node number is given Poisson networks: node numbers in different regions are independent When n goes to infinity, they are the same in the sense of probability Our results still hold in random networks [27] M. Penrose, “Random Geometric Graphs”, Oxford Univ. Press, Oxford, U.K., 2003. Asymptotic Analysis on Secrecy Capacity in Large-Scale Wireless Networks 26 Discussions Multicast Secrecy Capacity Corollary 1. Assume that legitimate nodes and eavesdroppers are independent poisson distributed with parameter 1 and e( n ) respectively. For each legitimate node, k − 1 nodes are randomly chosen as its destinations. For independent eavesdroppers case, n l og 4 n) k l og n n when k ( ) . n) l og n the aggregated multicast secrecy is ( k ( n ) l og n and is ( l og 4 when [24] X. Li, “Multicast Capacity of Wireless Ad Hoc Networks”, in IEEE/ACM Trans. Networking, Vol. 17, No. 3, pp. 950-961, 2009. Asymptotic Analysis on Secrecy Capacity in Large-Scale Wireless Networks 27 Discussions Secrecy Capacity in i.i.d Mobility Networks Corollary 2. Consider a cell-partitioned network under the two-hop relay algorithm proposed in [19], and assume that nodes change cells i.i.d. and uniformly over each cell every timeslot. For independent eavesdroppers case, the per-node secrecy capacity is (1) and the corresponding delay is ( n) For colluding case, the per-node secrecy capacity is ( f ( e( n ) ) ) and the corresponding delay is ( n ). f ( e( n) ) [19] M. J. Neely and E. Modiano, “Capacity and Delay Tradeoffs for Ad Hoc Mobile Networks”, in IEEE Trans. Inform. Theory, Vol. 51, No. 6, pp. 1917-1937, 2005.. Asymptotic Analysis on Secrecy Capacity in Large-Scale Wireless Networks 28 Discussions Secrecy Capacity under Random Walk Networks Corollary 3. Under random walk mobility model, nodes can only move to adjacent cells every timeslot. For independent eavesdroppers case, the per-node secrecy capacity is and the corresponding delay is (1) ( n l og n) . For colluding case, the per-node secrecy capacity is ( f ( e( n ) ) ) and the corresponding delay is ( n l og n ) . f ( e( n) ) [30] A. Gamal, J. Mammen, B. Prabhakar, and D. Shah, “Throughput-delay trade-off in wireless networks”, In Proceeding of IEEE INFOCOM, Hong Kong, China, Mar. 2004. Asymptotic Analysis on Secrecy Capacity in Large-Scale Wireless Networks 29 Outline Introduction Network Model and Definition Secrecy Capacity for Independent Eavesdroppers Secrecy Capacity for Colluding Eavesdroppers Discussion Conclusion and Future Work Asymptotic Analysis on Secrecy Capacity in Large-Scale Wireless Networks 30 Conclusions We derive the upper bound for secrecy capacity in largescale wireless networks by capturing the underling SINR relationship of eavesdroppers and legitimate nodes. The proposed scheme is order optimal for both the independent eavesdroppers and the colluding case. Our model relies weakly on the specific settings such as traffic pattern and mobility models of legitimate nodes and can be flexibly applied to more general cases and shed insights into the design and analysis of future wireless networks. Asymptotic Analysis on Secrecy Capacity in Large-Scale Wireless Networks 31 Future Work Secrecy capacity under active attacks The impact of heterogeneity networks The impact of dense networks and CR networks Asymptotic Analysis on Secrecy Capacity in Large-Scale Wireless Networks 32 Thank you ! Backup Revolve on its own Using 4 antennas Impact of Secrecy on Capacity in Large-Scale Wireless Networks 34