secrecy in multiuser wireless communications: flirting with
TRANSCRIPT
Outline Introduction System Model Performance Upper Bounds Imperfect CSIT Ongoing Research
Secrecy in Multiuser Wireless Communications:Flirting with Many Girls or Boys at the Same
Time
Giovanni Geraci and Jinhong Yuan
Feb 18, 2013
Jinhong Yuan:
Secrecy in Multiuser MISO Wireless Communications
Outline Introduction System Model Performance Upper Bounds Imperfect CSIT Ongoing Research
Outline
Introduction: Physical Layer Security
System Model and Proposed Scheme
Performance
Upper Bounds
Imperfect Channel State Information
Ongoing Research
Jinhong Yuan:
Secrecy in Multiuser MISO Wireless Communications
Outline Introduction System Model Performance Upper Bounds Imperfect CSIT Ongoing Research
Outline
Introduction: Physical Layer Security
System Model and Proposed Scheme
Performance
Upper Bounds
Imperfect Channel State Information
Ongoing Research
Jinhong Yuan:
Secrecy in Multiuser MISO Wireless Communications
Outline Introduction System Model Performance Upper Bounds Imperfect CSIT Ongoing Research
Security in Wireless Networks
I Security is crucial for wireless communications:users rely on wireless devices to transmit sensitive data.
I Due to the broadcast nature of the physical medium,every node in the network is a potential eavesdropper.
Jinhong Yuan:
Secrecy in Multiuser MISO Wireless Communications
Outline Introduction System Model Performance Upper Bounds Imperfect CSIT Ongoing Research
Classical Cryptography
I Symmetric key: a secret key must be distributed
Jinhong Yuan:
Secrecy in Multiuser MISO Wireless Communications
Outline Introduction System Model Performance Upper Bounds Imperfect CSIT Ongoing Research
Classical Cryptography
I Asymmetric key:the eavesdropper must have limited computational power
Jinhong Yuan:
Secrecy in Multiuser MISO Wireless Communications
Outline Introduction System Model Performance Upper Bounds Imperfect CSIT Ongoing Research
Physical Layer Security
I Physical layer security exploits the randomness of the channel
I No secret key is needed
I Even an eavesdropper with unlimited computational powercannot extract any information from the received signal
I How? Example:
I Eavesdropper’s channel is noisier than the intended receiver’s
I Message is encoded in such a way that the intended receivercan recover it, but the eavesdropper cannot
I Some of the bits are used to “confuse” the eavesdropper
Jinhong Yuan:
Secrecy in Multiuser MISO Wireless Communications
Outline Introduction System Model Performance Upper Bounds Imperfect CSIT Ongoing Research
Physical Layer Security
I Physical layer security exploits the randomness of the channel
I No secret key is needed
I Even an eavesdropper with unlimited computational powercannot extract any information from the received signal
I How? Example:
I Eavesdropper’s channel is noisier than the intended receiver’s
I Message is encoded in such a way that the intended receivercan recover it, but the eavesdropper cannot
I Some of the bits are used to “confuse” the eavesdropper
Jinhong Yuan:
Secrecy in Multiuser MISO Wireless Communications
Outline Introduction System Model Performance Upper Bounds Imperfect CSIT Ongoing Research
Secrecy Rates
I The secrecy rate is the rate at which:
I The message is transmitted reliably to the intended user
I No information is leaked at the eavesdropper
I The secrecy rates achievable in multi-user networks withpractical transmission schemes remain an open problem.
Jinhong Yuan:
Secrecy in Multiuser MISO Wireless Communications
Outline Introduction System Model Performance Upper Bounds Imperfect CSIT Ongoing Research
Code construction: random binning
I Generate a message m out of 2nR possible messages.
I Associate m to a bin of size 2nR′, and select a message m′
inside the bin. Transmit the codeword (m,m′).
I If Cm are Ce are capacities to intended user and eavesdropper,then choose R + R ′ < Cm and R ′ > Ce ⇒ R < Cm − Ce .
Jinhong Yuan:
Secrecy in Multiuser MISO Wireless Communications
Outline Introduction System Model Performance Upper Bounds Imperfect CSIT Ongoing Research
Code construction: random binning
I Generate a message m out of 2nR possible messages.
I Associate m to a bin of size 2nR′, and select a message m′
inside the bin. Transmit the codeword (m,m′).
I If Cm are Ce are capacities to intended user and eavesdropper,then choose R + R ′ < Cm and R ′ > Ce ⇒ R < Cm − Ce .
Jinhong Yuan:
Secrecy in Multiuser MISO Wireless Communications
Outline Introduction System Model Performance Upper Bounds Imperfect CSIT Ongoing Research
Code construction: random binning
I Generate a message m out of 2nR possible messages.
I Associate m to a bin of size 2nR′, and select a message m′
inside the bin. Transmit the codeword (m,m′).
I If Cm are Ce are capacities to intended user and eavesdropper,then choose R + R ′ < Cm and R ′ > Ce ⇒ R < Cm − Ce .
Jinhong Yuan:
Secrecy in Multiuser MISO Wireless Communications
Outline Introduction System Model Performance Upper Bounds Imperfect CSIT Ongoing Research
Code construction: random binning
I The intended receiver can correctly decode
I The eavesdropper can decode within each bin,but he cannot distinguish between different bins!
Jinhong Yuan:
Secrecy in Multiuser MISO Wireless Communications
Outline Introduction System Model Performance Upper Bounds Imperfect CSIT Ongoing Research
Outline
Introduction: Physical Layer Security
System Model and Proposed Scheme
Performance
Upper Bounds
Imperfect Channel State Information
Ongoing Research
Jinhong Yuan:
Secrecy in Multiuser MISO Wireless Communications
Outline Introduction System Model Performance Upper Bounds Imperfect CSIT Ongoing Research
Multi-User MIMO with Malicious Users
How do we define the secrecy rates for this scenario?
Jinhong Yuan:
Secrecy in Multiuser MISO Wireless Communications
Outline Introduction System Model Performance Upper Bounds Imperfect CSIT Ongoing Research
Multi-User MIMO with Malicious Users
How do we define the secrecy rates for this scenario?
Jinhong Yuan:
Secrecy in Multiuser MISO Wireless Communications
Outline Introduction System Model Performance Upper Bounds Imperfect CSIT Ongoing Research
Worst-case Scenario
I The BS cannot predict the behavior of the users. We assumethat users can cooperate to jointly eavesdrop on other users.
I Worst-case for message uk : the K − 1 unintended users forman alliance k and jointly eavesdrop on user k.
Jinhong Yuan:
Secrecy in Multiuser MISO Wireless Communications
Outline Introduction System Model Performance Upper Bounds Imperfect CSIT Ongoing Research
Worst-case Scenario
I Secrecy rate Rs,k : reliable transmission to user k without
allowing k to obtain any information about uk .
I The secrecy sum-rate is given by Rs =∑K
k=1 Rs,k .
Jinhong Yuan:
Secrecy in Multiuser MISO Wireless Communications
Outline Introduction System Model Performance Upper Bounds Imperfect CSIT Ongoing Research
System Model
One Base Station (BS) with M transmit antennas serves Ksingle-antenna users. The vector of received signals y is given by
y = Hx + n
where
I x = 1√γWu is the vector of transmitted signals
I W = [w1, . . . ,wK ] is the linear precoding matrix
I H = [h1, . . . ,hK ]† is the Rayleigh fading channel matrix
I n is complex Gaussian noise with E[nn†]
= σ2I
I ρ = 1σ2 is the SNR.
Jinhong Yuan:
Secrecy in Multiuser MISO Wireless Communications
Outline Introduction System Model Performance Upper Bounds Imperfect CSIT Ongoing Research
Equivalent MISOME wiretap channel
I Each user k, with its own eavesdropper k and the transmitter,forms an equivalent MISOME wiretap channel[1]
I The eavesdropper k can cancel interference from uj , j 6= k
[1] A. Khisti and G. Wornell, ”Secure transmission with multiple antennas I: The MISOME wiretap channel,” IEEETrans. Inf. Theory, vol. 56, no. 7, pp. 3088−3104, Jul. 2010.
Jinhong Yuan:
Secrecy in Multiuser MISO Wireless Communications
Outline Introduction System Model Performance Upper Bounds Imperfect CSIT Ongoing Research
Equivalent MISOME wiretap channel
I For the k-th MISOME with linear precoding we obtain[2]
Rs,k = log2
(1 + SINRk
)− log2
(1 + SINR
k
)[2] I. Csiszar and J. Korner, ”Broadcast channels with confidential messages,” IEEE Trans. Inf. Theory, vol. 24,no. 3, pp. 339−348, May 1978.
Jinhong Yuan:
Secrecy in Multiuser MISO Wireless Communications
Outline Introduction System Model Performance Upper Bounds Imperfect CSIT Ongoing Research
Secrecy sum-rates
I The secrecy sum-rate achievable by linear precoding is
Rs =K∑
k=1
[log2
(1+
∣∣∣h†kwk
∣∣∣2γσ2 +
∑j 6=k
∣∣∣h†kwj
∣∣∣2︸ ︷︷ ︸SINR term
)−log2
(1+
∑j 6=k |h
†jwk |2
γσ2
︸ ︷︷ ︸leakage term
)]
I We need to choose the precoding matrix W = [w1, . . . ,wK ]
Jinhong Yuan:
Secrecy in Multiuser MISO Wireless Communications
Outline Introduction System Model Performance Upper Bounds Imperfect CSIT Ongoing Research
Regularized Channel Inversion
I RCI precoding is better than plain CI, particularly at low SNR
W = H†(HH† + αIK
)−1
I We look for the αopt that maximizes
Rs =K∑
k=1
[log2
(1+
∣∣∣h†kwk
∣∣∣2γσ2 +
∑j 6=k
∣∣∣h†kwj
∣∣∣2)−log2
(1+
∑j 6=k |h
†jwk |2
γσ2
)]
I We study the RCI precoder in the large-system regime:M,K →∞, with their ratio β = K/M being held constant.
Jinhong Yuan:
Secrecy in Multiuser MISO Wireless Communications
Outline Introduction System Model Performance Upper Bounds Imperfect CSIT Ongoing Research
Regularized Channel Inversion
I RCI precoding is better than plain CI, particularly at low SNR
W = H†(HH† + αIK
)−1
I We look for the αopt that maximizes
Rs =K∑
k=1
[log2
(1+
∣∣∣h†kwk
∣∣∣2γσ2 +
∑j 6=k
∣∣∣h†kwj
∣∣∣2)−log2
(1+
∑j 6=k |h
†jwk |2
γσ2
)]
I We study the RCI precoder in the large-system regime:M,K →∞, with their ratio β = K/M being held constant.
Jinhong Yuan:
Secrecy in Multiuser MISO Wireless Communications
Outline Introduction System Model Performance Upper Bounds Imperfect CSIT Ongoing Research
Proposed SchemeTransmitted signal generated as x = Wu, where
W =
1√γH
H(HHH + MξIK
)−1for β ≤ 1
1√rγ
(HHH + MξIM
)−1HH for 1 < β < 2
0 for β ≥ 2
ξ =−2ρ2 (1− β)2 + 6ρβ + 2β2 − 2 [β (ρ+ 1)− ρ] ·
√· · ·
6ρ2 (β + 2) + 6ρβ
ρ is the transmit SNRβ = K/M is the number of users per antenna
r = max
(ρ
K log2β2
4(β−1)
, 1
)is a power reduction constant.
Jinhong Yuan:
Secrecy in Multiuser MISO Wireless Communications
Outline Introduction System Model Performance Upper Bounds Imperfect CSIT Ongoing Research
Proposed SchemeTransmitted signal generated as x = Wu, where
W =
1√γH
H(HHH + MξIK
)−1for β ≤ 1
1√rγ
(HHH + MξIM
)−1HH for 1 < β < 2
0 for β ≥ 2
ξ =−2ρ2 (1− β)2 + 6ρβ + 2β2 − 2 [β (ρ+ 1)− ρ] ·
√· · ·
6ρ2 (β + 2) + 6ρβ
ρ is the transmit SNRβ = K/M is the number of users per antenna
r = max
(ρ
K log2β2
4(β−1)
, 1
)is a power reduction constant.
Jinhong Yuan:
Secrecy in Multiuser MISO Wireless Communications
Outline Introduction System Model Performance Upper Bounds Imperfect CSIT Ongoing Research
Outline
Introduction: Physical Layer Security
System Model and Proposed Scheme
Performance
Upper Bounds
Imperfect Channel State Information
Ongoing Research
Jinhong Yuan:
Secrecy in Multiuser MISO Wireless Communications
Outline Introduction System Model Performance Upper Bounds Imperfect CSIT Ongoing Research
Achievable Secrecy Sum-Rate
We showed that an achievable secrecy sum-rate for our scheme is
Rs (H) =K∑
k=1
log2
1 +ρ|hHk wk |2
1+ρ∑
j 6=k |hHk wj |2
1 + ρ∥∥H
kwk
∥∥2
+
We study Rs in the large-system regime (Random Matrix Theory):
I M,K →∞I The ratio β = K/M is held constant
Jinhong Yuan:
Secrecy in Multiuser MISO Wireless Communications
Outline Introduction System Model Performance Upper Bounds Imperfect CSIT Ongoing Research
Achievable Secrecy Sum-Rate
We showed that an achievable secrecy sum-rate for our scheme is
Rs (H) =K∑
k=1
log2
1 +ρ|hHk wk |2
1+ρ∑
j 6=k |hHk wj |2
1 + ρ∥∥H
kwk
∥∥2
+
We study Rs in the large-system regime (Random Matrix Theory):
I M,K →∞I The ratio β = K/M is held constant
Jinhong Yuan:
Secrecy in Multiuser MISO Wireless Communications
Outline Introduction System Model Performance Upper Bounds Imperfect CSIT Ongoing Research
Large-System Analysis
In the large-system regime
R◦s = K
log2
1 + g (β, ξ)ρ+ ρξ
β[1+g(β,ξ)]2
ρ+[1+g(β,ξ)]2
1 + ρ
(1+g(β,ξ))2
+
with
g (β, ξ) =1
2
sgn(ξ) ·
√(1− β)2
ξ2+
2 (1 + β)
ξ+ 1 +
1− βξ− 1
Jinhong Yuan:
Secrecy in Multiuser MISO Wireless Communications
Outline Introduction System Model Performance Upper Bounds Imperfect CSIT Ongoing Research
Asymptotic Sum-RatesFor large SNR, the secrecy sum-rate Rs simplifies to
Rs ≈
K log2
1−ββ + K log2 ρ for β < 1
K2 log2
2764 + K
2 log2 ρ for β = 1
3K log2ββ−1 − K log2 ρ for β > 1
whereas the sum-rate R without secrecy is[3]
R ≈
K log2
1−ββ + K log2 ρ for β < 1
K2 log2 ρ for β = 1
K log2ββ−1 for β > 1.
[3] V. Nguyen and J. Evans, ”Multiuser transmit beamforming via regularized channel inversion: A large systemanalysis,” in Proc. IEEE Global Commun. Conf. (GLOBECOM), Dec. 2008, pp. 1 4.
Jinhong Yuan:
Secrecy in Multiuser MISO Wireless Communications
Outline Introduction System Model Performance Upper Bounds Imperfect CSIT Ongoing Research
Asymptotic Sum-RatesFor large SNR, the secrecy sum-rate Rs simplifies to
Rs ≈
K log2
1−ββ + K log2 ρ for β < 1
K2 log2
2764 + K
2 log2 ρ for β = 1
3K log2ββ−1 − K log2 ρ for β > 1
whereas the sum-rate R without secrecy is[3]
R ≈
K log2
1−ββ + K log2 ρ for β < 1
K2 log2 ρ for β = 1
K log2ββ−1 for β > 1.
[3] V. Nguyen and J. Evans, ”Multiuser transmit beamforming via regularized channel inversion: A large systemanalysis,” in Proc. IEEE Global Commun. Conf. (GLOBECOM), Dec. 2008, pp. 1 4.
Jinhong Yuan:
Secrecy in Multiuser MISO Wireless Communications
Outline Introduction System Model Performance Upper Bounds Imperfect CSIT Ongoing Research
Asymptotic Sum-RatesFor large SNR, the secrecy sum-rate Rs simplifies to
Rs ≈
K log2
1−ββ + K log2 ρ for β < 1
K2 log2
2764 + K
2 log2 ρ for β = 1
3K log2ββ−1 − K log2 ρ for β > 1
whereas the sum-rate R without secrecy is
R ≈
K log2
1−ββ + K log2 ρ for β < 1
K2 log2 ρ for β = 1
K log2ββ−1 for β > 1.
I For β < 1 secrecy can be achieved at no costI For β = 1 secrecy can be achieved at a cost of less than 4dBI For β > 1 the secrecy requirements result in a poor sum-rate
Jinhong Yuan:
Secrecy in Multiuser MISO Wireless Communications
Outline Introduction System Model Performance Upper Bounds Imperfect CSIT Ongoing Research
Simulations vs Large-system analysis
0 5 10 15 200
1
2
3
4
ρ [dB]
Per-antennasecrecysum-rate
Simulations, M = 10
Simulations, M = 20
Simulations, M = 40
Large-system analysis
β = 0.8
β = 1.2
β = 1
The analysis is accurate even for finite-size systems
Jinhong Yuan:
Secrecy in Multiuser MISO Wireless Communications
Outline Introduction System Model Performance Upper Bounds Imperfect CSIT Ongoing Research
Outline
Introduction: Physical Layer Security
System Model and Proposed Scheme
Performance
Upper Bounds
Imperfect Channel State Information
Ongoing Research
Jinhong Yuan:
Secrecy in Multiuser MISO Wireless Communications
Outline Introduction System Model Performance Upper Bounds Imperfect CSIT Ongoing Research
Upper Bounds
We consider two upper bounds:
I Sum-rate R without secrecy requirementsI The gap R − Rs represents the secrecy lossI i.e. the loss due to the secrecy requirements
I Secrecy capacity Cs,SU of a single-user systemI The gap Cs,SU − Rs represents the multi-user lossI i.e. the loss due to the requirement of serving multiple users at
the same time
Jinhong Yuan:
Secrecy in Multiuser MISO Wireless Communications
Outline Introduction System Model Performance Upper Bounds Imperfect CSIT Ongoing Research
Upper Bounds
We consider two upper bounds:
I Sum-rate R without secrecy requirementsI The gap R − Rs represents the secrecy lossI i.e. the loss due to the secrecy requirements
I Secrecy capacity Cs,SU of a single-user systemI The gap Cs,SU − Rs represents the multi-user lossI i.e. the loss due to the requirement of serving multiple users at
the same time
Jinhong Yuan:
Secrecy in Multiuser MISO Wireless Communications
Outline Introduction System Model Performance Upper Bounds Imperfect CSIT Ongoing Research
Upper Bounds
We consider two upper bounds:
I Sum-rate R without secrecy requirementsI The gap R − Rs represents the secrecy lossI i.e. the loss due to the secrecy requirements
I Secrecy capacity Cs,SU of a single-user systemI The gap Cs,SU − Rs represents the multi-user lossI i.e. the loss due to the requirement of serving multiple users at
the same time
Jinhong Yuan:
Secrecy in Multiuser MISO Wireless Communications
Outline Introduction System Model Performance Upper Bounds Imperfect CSIT Ongoing Research
Cost of Secrecy in Multi-User Systems [bits]
Users per antenna Secrecy Loss Multi-User Loss
β < 1 0 ∼ 0
β = 1 0.62 0.62
1 < β < 2 2− log2 β 2− 2 log2 β
β ≥ 2 log2ββ−1 0
Jinhong Yuan:
Secrecy in Multiuser MISO Wireless Communications
Outline Introduction System Model Performance Upper Bounds Imperfect CSIT Ongoing Research
Proposed Scheme vs Upper Bounds
0 5 10 15 20 250
2
4
6
8
ρ [dB]
Per-userrate
Proposed scheme
Upper bound (no secrecy)
Upper bound (single user)
β = 1
β = 1.2
β = 0.8
No multiplexing gain loss!
Jinhong Yuan:
Secrecy in Multiuser MISO Wireless Communications
Outline Introduction System Model Performance Upper Bounds Imperfect CSIT Ongoing Research
Outline
Introduction: Physical Layer Security
System Model and Proposed Scheme
Performance
Upper Bounds
Imperfect Channel State Information
Ongoing Research
Jinhong Yuan:
Secrecy in Multiuser MISO Wireless Communications
Outline Introduction System Model Performance Upper Bounds Imperfect CSIT Ongoing Research
Imperfect CSIT
H = H + E, var(hij
)= 1− τ2, var(eij) = τ2
τ2 can significantly degrade Rs , especially at high SNR.
Jinhong Yuan:
Secrecy in Multiuser MISO Wireless Communications
Outline Introduction System Model Performance Upper Bounds Imperfect CSIT Ongoing Research
Imperfect CSIT
H = H + E, var(hij
)= 1− τ2, var(eij) = τ2
τ2 can significantly degrade Rs , especially at high SNR.
Jinhong Yuan:
Secrecy in Multiuser MISO Wireless Communications
Outline Introduction System Model Performance Upper Bounds Imperfect CSIT Ongoing Research
Imperfect CSIT
H = H + E, var(hij
)= 1− τ2, var(eij) = τ2
τ2 can significantly degrade Rs , especially at high SNR.
Jinhong Yuan:
Secrecy in Multiuser MISO Wireless Communications
Outline Introduction System Model Performance Upper Bounds Imperfect CSIT Ongoing Research
Secrecy Sum-Rate under Imperfect CSIT
In the large-system regime
R◦s = K
log2
1 + g(β, ξ)ρ+ ξρ
β [1+g(β,ξ)]2
ρ+[1+g(β,ξ)]2
1 + ρ
[τ2 + 1−τ2
(1+g(β,ξ))2
]
+
with
ρ4=ρ(1− τ2
)ρτ2 + 1
and ξ4=
ξ
1− τ2
Jinhong Yuan:
Secrecy in Multiuser MISO Wireless Communications
Outline Introduction System Model Performance Upper Bounds Imperfect CSIT Ongoing Research
Minimum Required CSIT quality
I For β ≤ 1, a CSI distortion τ2 = Cρ , with
C =
{12
(√4b − 3− 1
)for β < 1
23
(√3b − 2− 1
)for β = 1
produces a high-SNR rate gap of log2 b bits.
I For β > 1, if limρ→∞ τ2 = 0,
then the high-SNR rate gap is zero.
Jinhong Yuan:
Secrecy in Multiuser MISO Wireless Communications
Outline Introduction System Model Performance Upper Bounds Imperfect CSIT Ongoing Research
Minimum Required CSIT quality
I For β ≤ 1, a CSI distortion τ2 = Cρ , with
C =
{12
(√4b − 3− 1
)for β < 1
23
(√3b − 2− 1
)for β = 1
produces a high-SNR rate gap of log2 b bits.
I For β > 1, if limρ→∞ τ2 = 0,
then the high-SNR rate gap is zero.
Jinhong Yuan:
Secrecy in Multiuser MISO Wireless Communications
Outline Introduction System Model Performance Upper Bounds Imperfect CSIT Ongoing Research
Simulations
0 5 10 15 200
1
2
3
4
5
ρ [dB]
Per-usersecrecyrate
τ2 = 0
τ2 = Cρ , C for log2 b = 1
β = 1.2
β = 1
β = 0.8
Jinhong Yuan:
Secrecy in Multiuser MISO Wireless Communications
Outline Introduction System Model Performance Upper Bounds Imperfect CSIT Ongoing Research
Outline
Introduction: Physical Layer Security
System Model and Proposed Scheme
Performance
Upper Bounds
Imperfect Channel State Information
Ongoing Research
Jinhong Yuan:
Secrecy in Multiuser MISO Wireless Communications
Outline Introduction System Model Performance Upper Bounds Imperfect CSIT Ongoing Research
Ongoing Research
Jinhong Yuan:
Secrecy in Multiuser MISO Wireless Communications
Outline Introduction System Model Performance Upper Bounds Imperfect CSIT Ongoing Research
Imperfect CSIT in FDD and TDD systems
I We want a constant rate gap, compared to the case withperfect CSIT (no multiplexing gain loss)
I In a Frequency Division Duplex (FDD) system:I How many feedback bits are required by each user?
I In a Time Division Duplex (TDD) system:I How much channel training is required by each user?
Jinhong Yuan:
Secrecy in Multiuser MISO Wireless Communications
Outline Introduction System Model Performance Upper Bounds Imperfect CSIT Ongoing Research
Imperfect CSIT in FDD and TDD systems
I We want a constant rate gap, compared to the case withperfect CSIT (no multiplexing gain loss)
I In a Frequency Division Duplex (FDD) system:I How many feedback bits are required by each user?
I In a Time Division Duplex (TDD) system:I How much channel training is required by each user?
Jinhong Yuan:
Secrecy in Multiuser MISO Wireless Communications
Outline Introduction System Model Performance Upper Bounds Imperfect CSIT Ongoing Research
Imperfect CSIT in FDD and TDD systems
I We want a constant rate gap, compared to the case withperfect CSIT (no multiplexing gain loss)
I In a Frequency Division Duplex (FDD) system:I How many feedback bits are required by each user?
I In a Time Division Duplex (TDD) system:I How much channel training is required by each user?
Jinhong Yuan:
Secrecy in Multiuser MISO Wireless Communications
Outline Introduction System Model Performance Upper Bounds Imperfect CSIT Ongoing Research
SDoF under Hybrid CSIT
Transmission scheme that achieves the optimal SDoF region?
Jinhong Yuan:
Secrecy in Multiuser MISO Wireless Communications
Outline Introduction System Model Performance Upper Bounds Imperfect CSIT Ongoing Research
SDoF under Hybrid CSIT
Transmission scheme that achieves the optimal SDoF region?
Jinhong Yuan:
Secrecy in Multiuser MISO Wireless Communications
Outline Introduction System Model Performance Upper Bounds Imperfect CSIT Ongoing Research
Acknowledgements (in alphabetical order...)
I Dr. Iain B. CollingsCSIRO ICT Centre, Sydney, Australia
I A/Prof. Romain CouilletDept. of Telecommunications, Supelec, Gif-sur-Yvette, France
I Prof. Merouane DebbahAlcatel-Lucent Chair on Flexible Radio, Supelec, Gif-sur-Yvette, France
I Malcolm EganUniversity of Sydney, Australia
I A/Prof. Mari KobayashiDept. of Telecommunications, Supelec, Gif-sur-Yvette, France
I A/Prof. Adeel RaziNED University of Engineering and Technology, Karachi, PakistanUniversity College London
Jinhong Yuan:
Secrecy in Multiuser MISO Wireless Communications
Outline Introduction System Model Performance Upper Bounds Imperfect CSIT Ongoing Research
PublicationsConference Proceedings
I G. Geraci et al., “Secrecy sum-rates for multi-user MIMO linearprecoding,” Proc. IEEE ISWCS, 2011
I G. Geraci et al., “Large system analysis of the secrecy sum-rates withregularized channel inversion precoding,” Proc. IEEE WCNC, 2012.
Journal PapersI G. Geraci et al., “Secrecy sum-rates for multi-user MIMO regularized
channel inversion precoding,” IEEE Trans. on Commun., November 2012I G. Geraci et al., “Large system analysis of linear precoding in MISO
broadcast channels with confidential messages”, IEEE JSAC, 2012,submitted.
Book ChaptersI G. Geraci and Jinhong Yuan, “Physical Layer Security for Multiuser
MIMO Communications,” InTech Publisher, 2012, submitted.
In PreparationI G. Geraci et al., “Secrecy Sum-Rates with Regularized Channel Inversion
Precoding under Limited Feedback”, for IEEE ICASSP, 2012.
Jinhong Yuan:
Secrecy in Multiuser MISO Wireless Communications
Outline Introduction System Model Performance Upper Bounds Imperfect CSIT Ongoing Research
Thank you!
Jinhong Yuan:
Secrecy in Multiuser MISO Wireless Communications