See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/275635145

Multi-user chaos MIMO-OFDM scheme for physical layer multi-access

security

Article in Nonlinear Theory and Its Applications IEICE · April 2014

DOI: 10.1587/nolta.5.172

CITATIONS READS

7 140

2 authors, including:

Eiji Okamoto
Nagoya Institute of Technology
129 PUBLICATIONS 612 CITATIONS

SEE PROFILE

All content following this page was uploaded by Eiji Okamoto on 18 May 2018.

The user has requested enhancement of the downloaded file.

 NOLTA, IEICE

Paper

Multi-user chaos MIMO-OFDM scheme

for physical layer multi-access security

Yuma Inaba 1 a) and Eiji okamoto 1 b)

1
Department of Computer Science and Engineering, Graduate School of
Engineering, Nagoya Institute of Technology
Gokiso-cho, Showa-ku, Nagoya-shi, Aichi-ken 466-8555, Japan

a)
cju17512@stn.nitech.ac.jp
b)
okamoto@nitech.ac.jp

Received July 12, 2013; Revised November 16, 2013; Published April 1, 2014

Abstract: Wireless multihop networks can be potentially used to realize a smart commu-
nity that comprehensively controls social infrastructures. In a wireless multihop transmission,
personal data are forwarded by a third party, so wireless security is indispensable. In current
wireless systems, security is ensured by encryption in the upper layers. However, this encryption
tends to require a complex protocol or processing, which is not suitable for a multihop protocol
with a simple implementation. To solve this problem, we have proposed a chaos multiple-input
multiple-output orthogonal frequency division multiplexing (MIMO-OFDM) scheme with phys-
ical layer encryption and channel-coding abilities. On the other hand, an effective technique
for wireless multihop transmission is point-to-multipoint (P-MP) communication, and recently,
multi-user (MU)-MIMO has been proposed as an effective P-MP scheme. In MU-MIMO, wire-
less security is also important. However, there are few studies considering MU-MIMO security
in the physical layer. Therefore, in this paper, we propose a multiuser (MU) chaos MIMO-
OFDM scheme that achieves physical layer security and channel-coding gain in MU-MIMO
transmission. Additionally, the user distribution and propagation loss are taken into consider-
ation. The improved performances are shown through computer simulations.
Key Words: multihop communications, encryption, chaos, MU-MIMO

1. Introduction
In wireless communications, frequency bands are a limited resource, and improvement in spectral
efficiency is always desired. Additionally, it is not practical to expand a frequency band for one
user for higher-capacity transmissions. A multiple-input multiple-output (MIMO) technique in which
multiple antennas in both the transmitter and receiver simultaneously communicate using the same
frequency band can increase the channel capacity without expanding the frequency band and is
drawing significant attention. The transmission rate can be increased in proportion to the minimum
numbers of transmit and receive antennas. However, in mobile terminals such as smart phones, the
number of antennas and the signal processing ability are limited by the restriction of the package and
battery sizes, and the capacity enhancement of MIMO is restricted by the mobile terminal limitation

172
Nonlinear Theory and Its Applications, IEICE, vol. 5, no. 2, pp. 172–183 IEICE
c 2014 DOI: 10.1587/nolta.5.172
even if the base station can prepare many antennas. To address this problem, multiuser MIMO (MU-
MIMO), a type of advanced MIMO technique, has been proposed in which spatial multiple access
can be achieved and in which the system capacity is increased even with a single antenna of a mobile
terminal. As a result, many improved MU-MIMO systems have been studied [1–4]. In MU-MIMO,
beamforming is conducted at the transmitter to avoid signals for multiple receivers interfering with
each other. In particular, the transmitter adjusts the transmit weights multiplied by each transmit
symbol of the MIMO system according to the channel matrix between the transmitter and all receivers
so as to make the signal of each user orthogonal. By this orthogonal beamforming, interference-free
simultaneous transmission is achieved.
On the other hand, along with the development of wireless technologies, a smart community system
has been developed with the use of wireless multihop transmission, e.g., smart meters. In those wireless
multihop transmissions, personal data are transmitted via the terminal of a third party, thus ensuring
that security is indispensable. In multihop transmissions, as shown in Fig. 1, data are forwarded
by relay nodes using a forwarding protocol such as amplify and forward (AF), decode and forward
(DF), or detect and forward (DetF) [5]. It is desirable that relay nodes cannot decode the forwarding
data. Usually the encryption is conducted with the upper layer protocols. However, complicated
processes with increased complexity are needed when a complicated secure protocol is used in the
implementation. To solve this problem, we have proposed a MIMO multiplexing transmission with
a physical layer security technique called chaos MIMO (C-MIMO) in [6, 7]. This technique works by
ensuring that, with the physical layer security, the upper layer secure protocols can be omitted and
that the increase in complexity can be restricted. In C-MIMO, the modulated signals are generated by
utilizing the deterministic and irregular characteristics of chaos, introduced by the principle of chaos
communication. In addition, rate-one channel coding is conducted by using chaos signals correlated
with transmit bits. Hence, the physical-layer security and channel coding gain are obtained in a
trade-off with the increase in the decoding search. As stated above, security also has to be ensured in
multihop transmission and wireless multiple access networks, and MU-MIMO realizing one-to-many
communications with the physical layer security is effective in terms of a multiple-access scheme,
spectral efficiency, and a secure protocol. In [8], the application of C-MIMO into AF multihop
transmission has been proposed, and the effectiveness of security and cooperative diversity has been
demonstrated. In a multihop transmission, as in Fig. 1, when only the source and the destination
nodes have a shared secret key, the relay nodes cannot decode the forwarding data, but the destination
can decode the forwarded data and can also take advantage of cooperative diversity. Furthermore, as
shown in Fig. 2, if C-MIMO and MU-MIMO are combined, secure multiuser multihop transmission

Fig. 1. Multihop transmission.

Fig. 2. MU-Multihop transmission.

173
is expected to be successful.
However, the application of C-MIMO into MU-MIMO has not been considered yet. Therefore,
in this paper, to achieve a secure multiuser multihop transmission, we propose a multiuser chaos
MIMO-orthogonal frequency division multiplexing (OFDM) scheme enabling multiuser transmission
and considering propagation loss with physical layer security and channel coding gain by applying
C-MIMO to MU-MIMO. Additionally, the user distribution and the propagation loss are taken into
consideration. Here, to focus on the secure MU-MIMO architecture, it is assumed in this paper
that the receive users of MU-MIMO in Fig. 2 are not the relay nodes but the destination users for
simplicity.
In the following, the chaos MIMO-OFDM transmission scheme and the proposed MU-C-MIMO-
OFDM scheme are introduced in Sections 2 and 3. The numerical results in which security is ensured
and the coding gain is improved according to the number of users are shown in Section 4, and
conclusions are drawn in Section 5.

2. Chaos MIMO-OFDM system

Figure 3 shows the baseband system model of C-MIMO-OFDM [6], where Tx and Rx are the numbers
of transmit and receive antennas, respectively. The chaos encoder and decoder are inserted into
the conventional MIMO-OFDM system. The variables in the figure are introduced in the following
subsections.

Fig. 3. System block diagram of chaos MIMO-OFDM.

2.1 Transmit and receive symbols

In C-MIMO-OFDM, the transmit data in one OFDM frame is denoted by a complex vector as

d = (d0 , · · · , dTx N −1 ) (1)

where N is the number of subcarriers. The block transmission is adopted to obtain the channel coding
gain. When B is defined as the chaos block length, the j-th data block is given from (1) as

dj = (dj0 , · · · , djt , · · · , djTx B−1 ) (2)

where 0≤ j≤ N/B-1. Then, the complex chaos vector corresponding to the data block is prepared as

c = (c0 , · · · , ct , · · · , cTx B−1 ) (3)

and the multiplied vector at each element

sj = c ∗ dj = (c0 dj0 , · · · , ct djt , · · · , cTx B−1 djTx B−1 )

= (sj0 , · · · , sjt , · · · , sjTx B−1 ) (4)

is the transmit symbol at the j-th block, where each element is allocated to a different subcarrier of
OFDM, and the generation method of c is described in Subsection 2.3. The transmit sequence of the
j-th block from the l-th MIMO transmit antenna is defined as

sjl = (sjB(l−1) , · · · , sjBl−1 )

= (sjl (0), · · · , sjl (q), · · · , sjl (B − 1)) (5)

174
where 1≤ l ≤ Tx . Then, the l-th antenna transmits the sequence of
N/B−1
sl = (s0l , · · · , sjl , · · · , sl ) (6)

Similarly to (5), the receive sequence is defined by

rjm = (rjm (0), · · · , rjm (q), · · · , rjm (B − 1)) (7)

where m is the number of receive antennas, and 1≤ m ≤ Rx .

2.2 Maximum likelihood sequence estimation in decoder

Let us consider the transmit and receive vectors of the q-th symbol 0 ≤ q ≤ B-1 in the j-th block as

sj,q = [sj1 (q), · · · , sjTx (q)]T

rj,q = [r1j (q), · · · , rR
j
x
(q)]T

where T is the matrix transpose. When the channel matrix between the transmitter and the receiver
at the q-th symbol (subcarrier) at the j-th block is given by
⎡ j j ⎤
H11 · · · H1T x
⎢ .. .. .. ⎥
Hj,q = ⎣ . . . ⎦ (8)
j j
HR x1
· · · HR x Tx

Then, the decoded sequence ŝj at the j-th block can be obtained by the maximum likelihood sequence
estimation (MLSE) as
B−1

ŝj = arg min rj,q − Hj,q sj,q  (9)
sj q=0

Thus, the joint block decoding in terms of the chaos decoding and MIMO detection is conducted.

2.3 Generation of chaos symbols

The generation method of the chaos symbols in (3) is described below. To obtain the channel coding
gain, the chaos symbols generated are correlated to the data vector of (2). The t-th chaos symbol of
(3) is generated by

ct = exp(j2π arctan([st ]/[st ]))

Here, 0 ≤ t ≤ Tx B-1, and st is a pseudorandom Gaussian symbol given by

M −1
1 
st = ([gti ] + [gti ]) exp(j8π[[ggi ] − [gti ]])
M i=0

where gti is the i-th of M chaos element symbols at index t. M is the number of chaos element
symbols to make st Gaussian symbols by the central limit theorem. Because ct is a unit vector with
a random phase, the encryption of dj in (4) is executed by the phase shift operation, and the signal
power of sj is not changed by c. Here, let M chaos element symbols be denoted as

gM,t = (gt0 , · · · , gt(M −1) ), 0 < [gti ], [gti ] < 1 (10)

and its initial vector is denoted by

gM,0 = (g00 , · · · , g0(M −1) )

This gM,0 becomes the key vector shared by the transmitter and the receiver. Thus, the proposed C-
MIMO scheme is a type of common key encryption. The chaos element symbols of (10) are iteratively
calculated by the following equations of (11)–(14).

175
 ⎧
⎨ [g(t−1)i ] (bm = 0)
x0 = 1 − [g(t−1)i ] (bm = 1, [g(t−1)i ] > 1/2) (11)
⎩
[g(t−1)i ] + 1/2 (bm = 1, [g(t−1)i ] ≤ 1/2)

y0 = [g(t−1)i ] (12)

xl+1 = 2xl mod1 , yl+1 = 2yl mod1 (13)

[gti ] = xBO+bm , [gti ] = yBO+bm (14)

Here, BO is a positive constant, m = (t+Tx B −1) mod Tx B, and bm is the transmit bit corresponding
to the m-th data symbol djm in the j-th block. Thus, the chaos symbols are correlated with the transmit
bits in the same block, and the convolutional channel coding effect is obtained. Equation (13) is the
Bernoulli shift map and makes gti chaotic.

2.4 Bit error rate performance of C-MIMO-OFDM

Here, the performance of the C-MIMO-OFDM scheme is evaluated through computer simulations
with the parameters of Table I. It is assumed that the channel is an antenna-i.i.d. and OFDM frame-
i.i.d. 1-dB decaying 9-path quasi-static Rayleigh fading channel and that the receiver perfectly knows
the channel. The chaos maps are the Bernoulli shift map of (13) and the Tent map given by

2al (al < 0.5)

al+1 =
2al (1 − al ) (0.5 ≤ al )

The chaos initial value gM,0 is randomly changed on every chaos block to obtain the average perfor-
mance. Figure 4 shows the bit error rate (BER) after the MLSE decoding of (9). It is shown that
the performance of C-MIMO-OFDM is superior to that of unsecured MIMO transmission with MLD.
This improvement is achieved by the convolutional coding effect of chaos. In addition, it is confirmed
that the BER is not affected by the use of different maps because the chaos encoding is conducted by
the random phase shift in C-MIMO, as described in (4) and Subsection 2.3. Therefore, the Bernoulli
shift map whose generation and implementation are simple is used in this paper.

Table I. Simulation conditions of C-MIMO-OFDM.

Modulation BPSK
Num. of 1OFDM symbols 64
Num. of transmit antenna Tx 2
Num. of receive antenna Rx 2
Chaos N/A Bernoulli shift map tent map
Num. of MIMO symbols on 1 block B N/A 4
Num. of chaos signals M N/A 10
Num. of chaos iteration BO N/A 19
Channel 1-dB decaying 9-path quasi-static Rayleigh fading
Receive channel state inf. Perfect
Decoding algorithm MLD MLSE

3. Proposed MU-C-MIMO-OFDM system

Figure 5 shows the system block diagram of the proposed MU-C-MIMO-OFDM system. Based
on C-MIMO-OFDM described in Section 2, MU-MIMO architecture is added, and the multiuser
transmission is achieved. Each user has a different key gM,0 from the others but shares it only with
the transmitter. Consequently, a secure transmission with a channel coding gain for each user is
obtained.

176
 Fig. 4. BER Performance of chaos MIMO-OFDM.

Fig. 5. System block diagram of proposed multiuser chaos MIMO-OFDM.

3.1 Transmitter and receiver design

Consider the downlink MU-C-MIMO-OFDM. Let U be the number of users (1 ≤ k  ≤ U ), Rx be the
number of receive antennas per user, and Tx = U Rx be the number of transmit antennas. Then, the
transmit sequence Sj,n in the j-th block at the n-th subcarrier is defined by
Sj,n = [S1,1 · · · S1,Rx · · · Sk ,1 · · · Sk ,Rx · · · SU,1 · · · SU,Rx ]T
= [(Sj,n T j,n T j,n T T
1 ) · · · (Sk ) · · · (SU ) ]

where Sk ,1 is the transmit symbol to the first receive antenna of the k  -th user, and the vector Sj,n
k
is the transmit sequence to user k  as
Sj,n
k = [Sk ,1 · · · Sk ,Rx ]
  
T

The receive sequence of the k  -th user is defined by

Rj,n
k = [Rk ,1 · · · Rk ,Rx ]
  
T

and the sequence of all users Rj,n

k is given by

Rj,n = [(Rj,n T j,n T j,n T T

1 ) · · · (Rk ) · · · (RU ) ]
= [R1,1 · · · R1,Rx · · · Rk ,1 · · · Rk ,Rx · · · RU,1 · · · RU,Rx ]T

177
Then, Rj,n is obtained by
Rj,n = Hj,n Vj,n Sj,n = H̃j,n Sj,n (15)
where Vj,n is a precoding matrix in the j-th block at the n-th subcarrier, and H̃j,n is a block
diagonal matrix. By multiplying the precoding matrix Vj,n by the transmit vector Sj,n , the inter-user
interference is eliminated. The derivation of Vj,n is described in the next subsection. Equation (15)
can be described in matrix form as
⎡ j,n ⎤ ⎡ j,n ⎤ ⎡ j,n ⎤
R1 H̃1 0 S1
⎢ . ⎥ ⎢ ⎥⎢ . ⎥
⎢ .. ⎥ ⎢ . .. ⎥ ⎢ .. ⎥
⎢ ⎥ ⎢ ⎥⎢ ⎥
⎢ j,n ⎥ ⎢ j,n ⎥ ⎢ j,n ⎥
⎢ Rk ⎥ = ⎢ H̃k ⎥ ⎢ Sk ⎥ (16)
⎢ . ⎥ ⎢ ⎥⎢ . ⎥
⎢ . ⎥ ⎢ .. ⎥⎢ . ⎥
⎣ . ⎦ ⎣ . ⎦⎣ . ⎦
Rj,n
U 0 H̃j,n
U Sj,n
U

Rj,n j,n j,n

k = H̃k Sk

where H̃j,n   
k is an Rx × Rx equivalent channel matrix for user k .

3.2 Derivation of precoding matrix

First, a Tx × Rx precoding matrix V̂kj,n
 for the k  -th user in the j-th block in the n-th subcarrier is
j,n 
derived [9, 10]. When Hk is an Rx × Tx channel matrix between the transmitter and the k-th user,
V̂kj,n
 has to satisfy
Hj,n j,n
k V̂k = 0 , (k  = k)
Next, a (U − 1)Rx × Tx subchannel matrix of all users except the k  -th user is defined by
Ĥj,n j,n T j,n T j,n T j,n T T
k = [(H1 ) · · · (Hk −1 ) (Hk +1 ) · · · (HU ) ]

Then, when Ĥj,n

k is decomposed to singular values, we obtain

Ĥj,n j,n j,n j,n H

k = Ûk D̂k (V̂k ) (17)
where Ûj,n
k and V̂kj,n
are (U − 1)Rx × (U − 1)Rx and Tx × Tx unitary matrices, respectively, and D̂j,n
k
is a (U − 1)Rx × Tx diagonal matrix whose diagonal and non-diagonal elements are the square roots
of the eigenvalues of (V̂kj,n H j,n
 ) V̂k  and zero, respectively. Here, H denotes the Hermitian transpose,
and the column vectors of D̂k from (U − 1)Rx to Tx become zero vectors. From (17), Ĥj,n
j,n
k satisfies

Ĥj,n j,n j,n j,n

k V̂k = Ûk D̂k (18)
and when the column vectors from (U − 1)Rx + 1 to Tx of the left term in (18) are extracted, we
obtain
⎡ ⎤
Hj,n j,n 
1 v̂k ((U − 1)Rx + 1) ··· Hj,n j,n
1 v̂k (Tx )
⎢ .. ⎥
⎢ . ⎥
⎢ ⎥
⎢ j,n j,n  j,n j,n ⎥
⎢ k −1 k
H v̂ ((U − 1)R x + 1) · · · Hk −1 v̂k (Tx ) ⎥
⎢ j,n j,n ⎥=0 (19)

⎢ Hk +1 v̂k ((U − 1)Rx + 1) · · · Hj,n j,n
k +1 v̂k (Tx ) ⎥
⎢ ⎥
⎢ .. ⎥
⎣ . ⎦
j,n j,n  j,n j,n
HU v̂k ((U − 1)Rx + 1) · · · HU v̂k (Tx )

where v̂kj,n j,n

 (i) is the i-th column vector of V̂k  . It is confirmed from (19) that the column vectors from
j,n
(U − 1)Rx to Tx of V̂k form a null space for all users other than user k. Therefore, the precoding
matrix of the k  -th user V̂kj,n
 of Tx × Rx is given by
V̂kj,n j,n  j,n  j,n
 = (v̂k  ((U − 1)Rx + 1)v̂k  ((U − 1)Rx + 2) · · · v̂k  (Tx ))

and finally a comprehensive precoding matrix Vj,n for all users is composed as follows:
Vj,n = (V̂1j,n V̂2j,n · · · V̂U
j,n
)

178
3.3 Decoding of MU-C-MIMO-OFDM
The decoding algorithm of MU-C-MIMO-OFDM is almost the same as (9), expect for the channel
matrix in creating the symbol replica. The k  -th user conducts the MLSE decoding by
B−1

Ŝjk = arg min Rj,q j,n j,q
k − H̃k Sk  (20)
sjk q=0

Here, Ŝjk is the transmit block of the k  -th user in the j-th block, and Ŝjk is the decoded block.

4. Numerical results
4.1 Performances with identically distributed user channel
The performances of the proposed schemes were evaluated by computer simulations under the con-
ditions in Table II. It is assumed that the channels are assumed to be i.i.d 1-dB decaying 9-path
quasi-static Rayleigh fading channels in terms of each antenna and MIMO-OFDM block and that the
channels are perfectly known to all receivers. The block length is four, and the numbers of users are
1, 2, 4, and 8. Each user has a different chaos initial key, which is shared only with the transmitter.
In addition, to obtain the average performance, the initial keys are randomly changed at each block.
Figure 6 shows the BER performances decoded by the MLSE of (20) at each block. It is shown
that the BER of the proposed scheme is improved for the conventional unencrypted MIMO with a
maximum likelihood decoding (MLD) with an average Eb/N0 over 5 dB, regardless of the number of
users. However, it is noted that the decoding complexity is increased in the proposed scheme because
of the block decoding. It is also found that the BER becomes 0.5 when the initial key is not identical
(denoted as ‘unsync’ in Fig. 6), that is, physical layer security is ensured for every user.

Table II. Simulation conditions.

Conventional scheme Proposed scheme
Modulation BPSK
Num. of 1OFDM symbols 64
Num. of transmit antenna Tx 2 2,4,8, 16
Num. of receive antenna per user Rx 2
Num. of users U 1 1,2,4, 8
Chaos N/A Bernoulli shift map
Num. of MIMO symbols on 1 block B N/A 4
Num. of chaos signals M N/A 10
Num. of chaos iteration BO N/A 19
Channel 1-dB decaying 9-path quasi-static Rayleigh fading
Receive channel state inf. Perfect
Decoding algorithm MLD MLSE

Figure 7 shows the BER performance versus the number of users at an average Eb/N0 of 15 dB.
Figures 6 and 7 show that the BER slightly improves according to the increase in the number of users.
In MIMO-OFDM, if the channel coding is used in the frequency direction, BER is improved because
of the frequency diversity effect. When the number of delay paths and the delay spread is increased,
the frequency diversity effect is enlarged, and the performance is further improved. In the proposed
scheme, this effect happens according to the increase in the number of users. Because the frequency
diversity is obtained by the channel coding function of C-MIMO, the BER is improved compared to
MIMO-MLD regardless of U. Furthermore, as shown in Fig. 8, the equivalent channel (H̃j,n k of (16))
is changed according to the number of users.
Figure 8 is calculated from the ensemble average of the squared impulse response transformed from
one subcarrier channel of an OFDM channel matrix H̃j,n k in the frequency domain. We can see that
the delay spread effect is enlarged according to the increase in the number of users. Thus, as a result,
a strong frequency diversity effect is obtained, and the BER is improved in proportion to the number

179
Fig. 6. BER performance comparison of single-user and multiuser MIMO in
downlink.

Fig. 7. BER performance versus number of users at Eb /N0 = 15 dB.

Fig. 8. Average squared impulse response of H̃j,n

k in time domain.

180
of users. When the number of antennas is increased, antenna diversity is usually obtained. This effect
is the same as that shown in Fig. 8. Note that the computational complexity of precoding in the
transmitter increases as O(n3 ) as the number of users n increases [11].

4.2 Performances with Non-identically distributed user channel

In addition to the conditions in Table II, the path loss and shadowing parameter were considered, as
shown in Table III, and a simulation was conducted. It was assumed that each user was equally and
randomly distributed within a single circle cell whose radius was normalized to 1. Figure 9 shows
the average BER performance versus the average SNR per receive antenna at a cell edge user using
the MLSE of (20). It was found that the channel coding effect was achieved compared to single-user
MIMO, and frequency diversity in proportion to the number of users is still obtained even when the
user location is distributed. The reason for this is the same as what was stated in Subsection 4.1.
Table III. Channel parameters of path loss and shadowing.
Path loss exponent α 3.5
Standard deviation of
Shadowing loss β [dB] 7

Fig. 9. BER performance comparison of single-user and multiuser MIMO

considering propagation loss in downlink.

4.3 Security evaluation of MU-C-MIMO-OFDM

4.3.1 Information theoretic security
In general, the safety of wireless communications is evaluated from the information theoretical or
computational point of view. First, the information theoretic security is considered. The information
theoretic security can be evaluated by the channel capacity of an eavesdropper [12]. When the channel
is an i.i.d. Rayleigh fading channel, the equivalent channel can be modeled as a binary symmetric
channel (BSC), and the channel capacity of an eavesdropper CR is calculated by

CR = Rx [1 + PeR log PeR + (1 − PeR ) log(1 − PeR )] (21)

where PeR is the bit error rate of the eavesdropper. The communication can be recognized as safe
when CR is near zero.
In MU-MIMO transmission, beamforming is conducted at the transmitter based on the feedback
channel state information (CSI) from legitimate users. When the eavesdropper pretends to be a
legitimate user and sends his CSI, the data for the legitimate user are transmitted to the eavesdropper.

181
However, in MU-C-MIMO-OFDM, the data cannot be decoded unless the eavesdropper has the initial
key gM,0 of the legitimate user. Thus, it is assumed that the eavesdropper obtained a close key in
terms of the squared Euclidean distance by any means and that decoding was attempted. The security
of this case was evaluated by a computer simulation. The channel capacity of the eavesdropper is
calculated under the conditions in Tables II and III, and the average SNR for the cell edge user is 20
dB. The results in Fig. 10 show that the security is ensured when the difference in the key distance
is more than 10−11 , which is quite small, and that the security is the same as that of single-user
C-MIMO. Consequently, the proposed scheme provides the secure multiuser transmission.

Fig. 10. Channel capacity of eavesdropper versus squared Euclidean distance

from correct key when SNR of cell edge user is 20 dB.

4.3.2 Computational security

To evaluate the computational security, the number of decoding search patterns in the MU-C-MIMO-
OFDM scheme is calculated. The secret key of the proposed scheme

gM,0 = (g00 , · · · , g0(M −1) )

has M complex values. Then, when the 64-bit double floating-point value is used in the transmitter
and the receiver, the secret key becomes 128 × M bits long, so the eavesdropper needs 2128M trials
to decode it. Furthermore, when the number of transmit antennas and the block length are Tx = 2
and B = 4, respectively, the number of multiplication operations in MLSE decoding becomes 6,144.
Hence, the total number of decoding calculations is

6144 × 2128M ∼
= 1038.5M +3.79

If a reference PC with a processing speed of 4.4 × 109 [FLOPS/sec] is used for the security evalua-
tion [13], decoding requires

1038.5M +3.79 /(4.4 × 109 × 31536000) ∼

= 1038.5M −13.4 [years]

Compared with the upper layer protocol of RSA-2048, which requires 25 × 1016 years for decoding,
the above complexity of the proposed scheme with M = 10 can be recognized as sufficiently safe.

5. Conclusions
In this paper, we extended a chaos MIMO-OFDM scheme to a multiuser MIMO system and proposed
a multiuser chaos MIMO-OFDM scheme with physical layer security for every user and channel coding

182
 gain. According to the increase in the number of users, the frequency diversity effect was enlarged,
and the BER performance of the proposed scheme was improved in the cases of undistributed and
distributed users. In addition, security was evaluated by the leak capacity with close initial keys, and
the robustness of the proposed scheme was confirmed.

Acknowledgments
This work is partially supported by Strategic Information and Communications R&D Promotion
Programs (SCOPE) in the Ministry of Internal Affairs and Communications, Adaptable and Seamless
Technology Transfer Program through the target-driven R&D, JST, and KDDI foundation. The
authors wish to thank these entities for their support.

References
[1] F. Yu, C. Tellambura, and W.A. Krzymien, “Limited-feedback precoding for closed-loop mul-
tiuser MIMO OFDM systems with frequency offsets,” IEEE Trans. Commun., vol. 7, no. 11,
pp. 4155–4165, November 2008.
[2] R. Holakouei, A. Silva, and A. Gameiro, “Transmit power allocation for precoded distributed
MIMO OFDM systems,” IEEE International Conference on Advanced Information Networking
and Applications (AINA) 2010, pp. 190–197, April 2010.
[3] W.W.L. Ho and Y.C. Liang, “Optimal resource allocation for multiuser MIMO-OFDM systems
with user rate constraints,” IEEE Trans. Vehicular Tech., vol. 58, no. 3, pp. 1190–1203, March
2009.
[4] M. Schellmann, T. Haustein, and V. Jungnickel, “Spatial transmission mode switching in mul-
tiuser MIMO-OFDM systems with user fairness,” IEEE Trans. Vehicular Tech., vol. 59, no. 1,
pp. 235–247, March 2010.
[5] M. Benjillali and L. Szczecinski, “Quantization of channel state information for detect-and-
forward relaying schemes,” Proc. 2009 IEEE Global Telecommunications Conference (GLOBE-
COM 2009), Hawaii, December 2009.
[6] E. Okamoto, “A chaos MIMO transmission scheme for channel coding and physical-layer-
security,” IEICE Trans. Commun., vol. E95-B, no. 4, April 2012.
[7] E. Okamoto, “Chaos MIMO-OFDM transmission scheme achieving physical-layer security in
mobile channel environments,” IEICE Technical Report, vol. 112, no. 239, RSC2012-125, pp. 1–
6, October 2012.
[8] E. Okamoto, “A secure cooperative relay transmission using chaos MIMO scheme,” International
Conference on Ubiquitous and Future Networks (ICUFN), pp. 374–378, July 2012.
[9] S. Umeda, S. Suyama, H. Suzuki, and K. Fukawa, “PAPR Reduction Method for Block Diago-
nalization in Multiuser MIMO-OFDM Systems,” IEEE Vehicular Technology Conference (VTC
2010-Spring), pp. 1–5, May 2010.
[10] R. Holakouei, A. Silva, and A. Gameiro, “Precoded multiuser distributed MIMO OFDM sys-
tems,” IEEE Intl. Symp. on Wireless Communication Systems (ISWCS) 2009, pp. 605–608,
September 2009.
[11] Y. Wang, K. Cunningham, P. Nagvajara, and J. Johnson, “Design and prototype of singular
value decomposition hardware in IEEE 802.11n MIMO standards for software defined radio,”
International Conference on ReConFigurable Computing and FPGAs, pp. 400–405, December
2010.
[12] F. Oggier and B. Hassibi, “The secrecy capacity of the MIMO wiretap channel,” Proc. Int’l
Sym. on Information Theory (ISIT), pp. 524–528, July 2008.
[13] T. Kleinjung, “Evaluation of complexity of mathematical algorithms,” CRYPTREC technical
report No. 0604 in FY2006, 2007.

183

View publication stats