Você está na página 1de 4

This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/LCOMM.2017.2671406, IEEE
Communications Letters

On Reducing Intercept Probability for Unsubscribed Video Layers


using Network Coding
Mohammad S. Karim, Mohammad Esmaeilzadeh and Parastoo Sadeghi

Abstract—In this paper, we use random linear network coding Layer 1 Layer 2
(RLNC) to deliver a two layered video sequence to two wireless window 1 window 2
subscribers such that a high-paying subscriber receives the Layer 1 Layer 2 Bob
highest quality and a low-paying subscriber receives the basic Layer 1
²B
quality. We first derive closed-form expressions for the proba-
Alice
bility of the low-paying subscriber intercepting sufficient coded ²E Eve
packets to experience the highest video quality. We consider
both feedback-aided and feedback-free transmissions. We then
propose a resource allocation framework that minimizes the Fig. 1: Alice wants to send the first video layer to Eve and
intercept probability subject to deadline and subscribed video both video layers to Bob.
quality constraints. We finally show the tightness between the
theoretical and simulation results, and illustrate the protection
achieved by RLNC for subscription based layered video delivery.
probability of the eavesdropper by considering the field size
Index Terms—Network coding, layered video, intercept prob- of RLNC, feedback and fixed number of transmissions.
ability, subscription based video delivery. In contrast to [7]–[9], we address the following question:
‘How can we design a network coding framework that allows
I. I NTRODUCTION a low-paying subscriber to view the basic video quality while
Video streaming in smart phones is projected to double preventing it from receiving the highest video quality?’ As a
in each year [1]. This rapid growth of video traffic creates result, we first derive closed-form expressions for the proba-
numerous challenges for the service providers to meet the bility of intercepting the unsubscribed video layers by taking
quality of service requirements of their subscribers. Within into account the hard deadline and in-order layer decoding
this broad problem space, we consider a network topology constraints. We consider both feedback-aided and feedback-
wherein a wireless sender wants to deliver a two layered video free transmissions in order to identify the protection gain
sequence to two subscribers. Such layered video has a hard achieved by a feedback message. We finally propose a resource
deadline and imposes a decoding order on the layers [2]–[4]. allocation framework that minimizes the intercept probability
Moreover, a subscriber experiences the basic video quality if subject to a deadline and subscribed video quality constraints.
it decodes only the first layer and the highest video quality if II. S YSTEM M ODEL AND PARAMETERS
it decodes both layers. Particularly, we address the problem of We consider a wireless sender, Alice, that wants to multicast
secure delivering a two layered video sequence from a service a two layer H.264/SVC video sequence to two subscribers,
provider perspective such that (i) the subscribers decode one or Bob and Eve. This video sequence imposes a hierarchical
two video layers depending on their subscription levels, and order on decoding the layers such that the second layer, i.e.,
(ii) with a high probability a low-paying subscriber cannot enhancement layer, can be decoded only if all the packets of
view the highest quality video. this layer and the first layer, i.e., base layer, are successfully
Secure packet delivery over a wireless network is a chal- received [4]. The base layer provides a basic video qual-
lenging problem due to its broadcast nature. As a result, ity while the enhancement layer further improves the video
a low-paying subscriber who does not have access to the quality. As shown in Fig. 1, Alice wants to send the first
second video layer can overhear the packet transmissions and layer to both subscribers, and the enhancement layer only
potentially experience the highest video quality. To prevent to the higher paying subscriber, Bob. However, Eve tries to
such unauthorized access, physical layer security that exploits eavesdrop the second layer for experiencing a better video
the channel conditions is an emerging solution [5], [6]. Further, quality. Alice can send one packet in each time slot and the
random linear network coding (RLNC) that algebraically packet is subject to an independent Bernoulli erasure at Bob
mixes different packets provides protection of a video trans- and Eve with probability ǫB and ǫE , respectively. Note that
mission as a subscriber needs to collect sufficient number of the analysis of this paper can be extended to a large number of
coded packets before retrieving all packets simultaneously [3]. subscribers and video layers if (i) the unauthorized subscribers
A body of works [7]–[9] showed that network coding are likely to experience worse channel gain compared to
combined with physical layer security provides protection legitimate subscribers and (ii) an additional number of video
of packet delivery, increases robustness to packet losses and layers requested by the legitimate subscribers needs to be
simplifies scheduling. In particular, the authors in [8] exploited protected from the unauthorized subscribers.
fountain coding, channel fading and power control to ensure a Let k = [k1 , k2 ] represent the number of packets in the
higher packet reception rate at the legitimate user and reduce video sequence, with k1 and k2 being the number of packets in
the intercept probability of the eavesdropper. The authors the first and the second layers, respectively. The minimum de-
in [9] continued the work in [8] and derived the intercept lay requirement imposes a deadline constraint of N time slots
for delivering k1 + k2 packets. Similar to the works in [2], [4],
The authors are with the Research School of Engineering, Australian
National University, ACT 2601, Australia ({mohammad.karim, moham- we use a non-overlapping window strategy and, subsequently,
mad.esmaeilzadeh, parastoo.sadeghi}@anu.edu.au). focus on quantifying the intercept probability reduction using

1089-7798 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/LCOMM.2017.2671406, IEEE
Communications Letters

such window-based RLNC. We acknowledge that considering For notational simplicity, in addition to k and nT defined
overlapping window (i.e., coding over different video layers) is earlier, we use n̄T = [n¯t1 , n¯t2 ]. Then, the expressions in
another interesting option, whose theoretical performance can (2) and (3) can be equivalently written as F1,i (k, n̄T ) and
be quantified in a similar fashion as this paper but with higher F2,i (k, n̄T ), respectively.
complexity. As shown in Fig. 1, there are two windows for a
two layer video sequence. While the first window ω1 includes A. Feedback-aided Transmission (FT) System
all k1 first layer packets, the second window ω2 includes all Bob can decode both layers and generate a feedback mes-
k2 second layer packets. We employ RLNC to form a coded sage after k1 + k2 ≤ n¯t1 + n¯t2 ≤ nt1 + nt2 time slots. Since the
packet over window ω1 or window ω2 . With the usage of coded packets from windows ω1 and ω2 can be transmitted
large field sizes, we assume that coded packets constructed in a random order, a feedback generation implies that Bob
from packets of a coding window are all linearly independent. decodes all packets of either the first layer or the second layer
Therefore, Bob or Eve need to collect k1 and k2 coded packets in the (n¯t1 + n¯t2 )-th time slot, which allows him to immediately
from windows ω1 and ω2 to decode all packets of the first and decode both video layers. The probabilities of such events are:
the second layers, respectively.
g 1 (k, n̄T ) = PB (k2 , n¯t2 ){PB (k1 , n¯t1 ) − PB (k1 , n¯t1 − 1)}
Definition 1. A transmission schedule nT = [nt1 , nt2 ] defines a
set of N T coded packets such that nt1 ≥ k1 and nt2 ≥ k2 coded
g 2 (k, n̄T ) = PB (k1 , n¯t1 ){PB (k2 , n¯t2 ) − PB (k2 , n¯t2 − 1)}
packets are generated from ω1 and ω2 windows, respectively,
and nt1 + nt2 = N T ; N T ≤ N . Further, NT is the set of all With these two events, we define the probability that Bob
feasible transmission schedules, i.e., nT ∈ NT . decodes both video layers in exact (n¯t1 + n¯t2 )-th time slot as:
Indeed, a schedule nT is feasible when the following three 2
X n¯t
constraints are satisfied: (i) nt1 ≥ k1 , (ii) nt2 ≥ k2 , and (iii) f2,B (k, n̄T ) = g ℓ (k, n̄T ) ¯t ℓ ¯t (4)
n1 + n2
nt1 + nt2 ≤ N . Similar to [7]–[9], we consider a physical layer ℓ=1
security condition of ǫB < ǫE , which can be satisfied by power n¯t
where n¯t +ℓn¯t is the probability that the transmission in the
control, adaptive modulation, artificial noise or channel pre- 1 2

compensation. Further, we study systems with feedback-aided (n¯t1 + n¯t2 )-th time slot is from the ℓ-th layer.
transmission (FT) and feedback-free transmission (NT). In the Now, to derive the intercept probability, we consider two
FT system, Bob generates a single acknowledgement message cases. (Case 1:), the probability PBE (k, nT ) that both Bob
using a dedicated feedback channel when it decodes both and Eve decode the second layer (and of course the first
video layers. In the NT system, there is no acknowledgment layer). This is possible if Bob decodes the second layer in
message in N time slots. In both systems, we achieve secure the (n¯t1 + n¯t2 )-th time slot and Eve decoded the second layer
video delivery if Eve cannot decode the second layer. earlier or decodes the second layer concurrently with Bob.
III. P ERFORMANCE A NALYSIS FOR A T WO L AYER V IDEO Using expressions (4) and (3), and considering all feasible
transmissions of n¯t1 and n¯t2 , we define the probability:
We quantify the physical layer security by the intercept
probability that Eve decodes the second layer. To derive this
probability, we first define the probability that a subscriber PBE (k, nT )
i ∈ {B, E} with erasure probability ǫi decodes k packets after n1 t
n2t nt1  nt2 
receiving nr packets from nt coded packet transmissions as:
X X n¯t1 n¯t2
= f (k, n̄T )F2,E (k, n̄T ),
nt1 +nt2  2,B
(5)
(P t
n nt nr nt −nr n¯t1 =k1 n¯t2 =k2 n¯t +n¯t
; nt ≥ k

t
Pi (k, n ) = nr =k nr (1 − ǫi ) (ǫi ) 1 2

0 ; nt < k The probability (5) is computed as the summation of all


(1) feasible cases with n¯t1 from window ω1 and n¯t2 from window
ω2 , where Bob decodes exactly at (n¯t1 + n¯t2 )-th time slot (and
We for now consider point-to-point communications be-
immediately generates a feedback) and Eve decodes within
tween Alice and a subscriber i, where Alice broadcasts
(n¯t1 + n¯t2 ) time slots (and is able to intercept the unauthorized
n¯t1 ≤ nt1 and n¯t2 ≤ nt2 following schedule nT . Here,
layer). (Case 2:) the probability PE (k, nT ) that Eve decodes
n¯t1 ∈ [0, 1, ..., nt1 ], which is used to compute the binomial
the second layer but Bob fails to decode the second layer
distribution, i.e., the probability that a subscriber decodes k1
after transmitting the complete schedule of nt1 and nt2 coded
first layer video packets if it receives k1 or k1 + 1 or,..., or
packets. Using (3), we define the probability:
nt1 coded packets. Using (1), we define the probability that
subscriber i decodes the first layer, i.e., k1 first layer packets, PE (k, nT ) = F2,E (k, nT ) × (1 − F2,B (k, nT )). (6)
from n¯t1 first layer coded packet transmissions as:
Finally, we define the intercept probability as the sum of
F1,i (k1 , k2 , n¯t1 , n¯t2 ) = Pi (k1 , n¯t1 ). (2) the aforementioned two events’ probabilities as follows:
Having the constraint of second layer decoding is subject to PF T T T T
int (k, n ) = PBE (k, n ) + PE (k, n ). (7)
first layer decoding, we define the probability that subscriber
i decodes the second layer, i.e., k1 first layer packets and
B. Feedback-free Transmission (NT) System
k2 second layer packets, from n¯t1 first layer coded packet
transmissions and n¯t2 second layer coded packet transmissions: In NT system, Alice transmits a complete schedule of nt1
and nt2 coded packets. Therefore, the intercept probability is
F2,i (k1 , k2 , n¯t1 , n¯t2 ) = Pi (k1 , n¯t1 ) × Pi (k2 , n¯t2 ). (3) equal to the probability that Eve decodes the second layer

1089-7798 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/LCOMM.2017.2671406, IEEE
Communications Letters


within nt1 + nt2 time slots. Using expression (3), we define the Proof. Let us consider nt1 as the minimum number of first
intercept probability as: layer packet transmissions, within the range [k1 , N − k2 ], for
which constraint (10) holds. We note that the range starts
PN T T T
int (k, n ) = F2,E (k, n ). (8)
with k1 transmissions, which results from the exact number
IV. R ESOURCE A LLOCATION F RAMEWORK (RAF) of the first layer packet transmissions, and ends with N − k2

We determine the optimal schedule nT ∈ NT that transmissions, which results from the number of allowable
minimizes the intercept probability subject to a deadline and first layer packet transmissions given only k2 transmissions
subscribed video quality constraints using the following RAF: are allocated to second layer. If the number of first layer

coded packets is nt1 − 1, intercept probability will decrease as
Pint (k, nT )

min (9) per Proposition 1. However, nt1 − 1 transmissions will fail to
nT ∈NT ∗
satisfy constraint (10). As a result, nt1 ∈ [k1 , N − k2 ] is the
subject to F1,B (k, nT ) ≥ σ1 , F1,E (k, nT ) ≥ σ1 (10) optimum number of first layer packet transmissions. Similarly,
T ∗ ∗ ∗
F2,B (k, n ) ≥ σ2 (11) for a given nt1 , we can show that nt2 ∈ [k2 , N − nt1 ] is the
optimum number of second layer packet transmissions. 
where objective (9) minimizes the decoding probability of the
second layer at Eve over all feasible schedules NT , constraint Remark 1. In the case of FT system, we also select a schedule
∗ ∗
(10) ensures that the decoding probability of the first layer at with nt1 first layer coded packets and nt2 second layer
Bob and Eve meets a pre-defined threshold σ1 separately, and coded packets following Theorem 1. However, if Bob does not
∗ ∗
constraint (11) ensures that the decoding probability of the generate a feedback after nt1 and nt2 transmissions, Alice
second layer at Bob meets a pre-defined threshold σ2 . Here, starts transmitting packets from the successor set of packets.
σ2 < σ1 as the second layer decoding probability is dependent ∗
To determine nt1 in (13), we start with nt1 = k1 and search
on the first layer decoding probability. Note that the resource ∗
until nt1 = N − k2 . Then, to determine nt2 in (14), we start
allocation function is unsolvable when there exists no feasible ∗
with nt2 = k2 and search until nt2 = N − nt1 . We finally
schedule nT ∈ NT satisfying the aforementioned constraints.
provide the following theorem to demonstrate that a FT system
Indeed, the number of allowable time slots N needs to be
offers a protection gain compared to a NT system.
sufficiently large as compared to the required number of coded
packets since the transmissions are subject to channel erasures. Theorem 2. If the system is FT, the probability of intercepting
We now describe a proposition that leads to the solution of the second layer by Eve is no larger than the intercept
the RAF problem. probability of NT system, i.e.,
Proposition 1. The intercept probability PN T T
int (k, n ) is a PF T T NT T
int (k, n ) ≤ Pint (k, n ). (15)
non-decreasing function of the number of first layer packet ∗

transmissions nt1 and the number of second layer packet Proof. Let us consider that nt1 first layer coded packets

transmissions nt2 such as: and nt2 second layer coded packets are selected following

Theorem 1. If there is no feedback arrival at Alice, all nt1
PN T t t NT t t ∗
int (k1 , k2 , n1 , n2 ) ≤ Pint (k1 , k2 ,n1 + θ1 , n2 + θ2 ); and nt2 packets will be transmitted to satisfy constraints
θ1 ≥ 0, θ2 ≥ 0. (12) (10) and (11), respectively. However, if Alice receives an
acknowledgment in the (n¯t1 + n¯t2 )-th time slot, i.e., n¯t1 ≤ nt1

Proof. For a given schedule nT = [nt1 , nt2 ], let us consider and n¯t2 ≤ nt2 , the reduction in the number of first and

additional θ1 ≥ 0 and θ2 ≥ 0 transmissions for the first second layer packet transmissions are nt1 − n¯t1 and nt2 − n¯t2 ,
∗ ∗

layer and the second layer, respectively, and the resulting respectively. According to Proposition 1, a reduction in the
nt1 + θ1 and nt2 + θ2 transmissions. Using expression (1), we first or second layer packet transmissions results in a reduction
can infer that P(k1 , nt1 ) ≤ P(k1 , nt1 + θ1 ) and P(k2 , nt2 ) ≤ in the intercept probability, and concludes the proof. 
P(k2 , nt2 + θ2 ). According to expression (3), the increase
of probabilities P(k1 , nt1 + θ1 ) and P(k2 , nt2 + θ2 ) also V. S IMULATION R ESULTS OVER A R EAL V IDEO
increases the second layer decoding probability at Eve, i.e.,
We use H.264/SVC Foreman video sequence, which is in
F2,E (k1 , k2 , nt1 , nt2 ) ≤ F2,E (k1 , k2 , nt1 + θ1 , nt2 + θ2 ). Indeed,
common intermediate format, i.e. 352 × 288, and has 300
a higher probability F2,E (k1 , k2 , nt1 + θ1 , nt2 + θ2 ) causes a
frames with rate 30 frames per second [4]. We encode the
higher intercept probability. 
sequence using JSVM H.264/SVC codec while considering
With Proposition 1, we now describe the solution of the temporal scalability and a group of pictures (GOP) structure
RAF problem for NT system using the following theorem: ‘IBBBPBPP’ of 8 frames. Each GOP is encoded into two
video layers. We consider each packet consists of 1500 bytes,
Theorem 1. The optimal solution of the RAF problem in NT
where 100 bytes are used for header information and 1400
system, if admissible, is achieved by first minimizing the num-
bytes are used for video data. The average number of packets
ber of first layer packet transmissions nt1 and then minimizing
in the first and the second layers over 38 GOPs are 10.66 and
the number of second layer packet transmissions nt2 subject to
3.71, respectively. For a GOP, given that it has 8 frames, the
deadline and subscribed video quality constraints as follows:
video rate is 30 frames per second, the transmission rate is

nt1 = arg min{nt1 ∈ [k1 , N − k2 ] | F1,B (k, nT ) ≥ σ1 , α bits per second and a packet length is 1500 × 8 bits, the

F1,E (k, nT ) ≥ σ1 }. (13) deadline is calculated as: N = 1500×8×30 .
∗ ∗ Fig. 2 shows the optimum number of first and second layer
nt2 = arg min{nt2 ∈ [k2 , N − nt1 ] | F2,B (k, nT ) ≥ σ2 }. ∗ ∗
coded packets nt1 and nt2 determined by Theorem 1 against
(14) different channel erasure probabilities ǫB and ǫE . Here, the

1089-7798 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/LCOMM.2017.2671406, IEEE
Communications Letters

probability increases when both Bob and Eve have similar


30 4.4 channel erasure probabilities. However, the channel gains of
25 4.2
both Eve and Bob can be adjusted by power control, adaptive
modulation, artificial noise or channel pre-compensation [5],


nt1

nt2
20 4
[8]. Moreover, if Alice can estimate the instantaneous channel
15 3.8 quality, she can transmit packets from the first layer when Eve
10 3.6 has better channel condition than Bob. On the other hand, if
0.1 0.1 Bob has better channel condition than Eve, she can transmit
0.5 0.5
0.05 0.05 packets from the second layer. As expected, [9] suffers from
ǫB 0.3 ǫB 0.3
0 ǫE 0 ǫE high intercept probability due to encoding all packets together
0.1 0.1
∗ ∗ and transmitting these coded packets in a rateless fashion to
Fig. 2: nt1 and nt2 , determined by Theorem 1, against meet the threshold σ1 of decoding the first layer at Eve.
different channel erasure probabilities ǫB and ǫE . Finally, we can see the benefit of using a single feedback
message in reducing the intercept probability over feedback-
9:; <6=>75-63, free transmissions. For example, for ǫB = 0.05 and ǫE = 0.2,
?:; <6=>75-63,
9:; 5,578-6057 the intercept probability reduces from 0.48 to 0.27 when the
?:; 5,578-6057
?:; 5,578-6057 @(A
system changes from NT to FT. We also consider peak signal
"
B,03C.C; 5,578-6057 to noise ratio for measuring the intercepted video quality
!) and observe that Eve enhances its video quality by only
+,-./0.1- 2/3454676-8

0.64 and 1.4 dB in FT and NT systems respectively, for


!*
ǫB = 0.09, ǫE = 0.5. This is acceptable given the highest
!% video quality is 34.67 dB with decoding both layers.
!# VI. C ONCLUSION
!"
We used RLNC to deliver a layered video to two wireless
!#
! (
!$
subscribers with different quality of subscription. We derived
! '
!
! &
! $ !% closed-form expressions for the intercept probability of the
! " !& low-paying subscriber in experiencing the highest video qual-
Fig. 3: Intercept probability against different channel erasure ity in both FT and NT systems. We proposed a framework that
probabilities ǫB and ǫE . minimizes the intercept probability subject to a deadline and
subscribed video quality constraints. Simulation and analytical
results showed that RLNC is sufficient to achieve a high level
deadline is N = 45 time slots, the threshold for decoding the of protection for layered video delivery given the legitimate
first layer at Bob and Eve separately is σ1 = 0.9, and the subscriber has better channel condition than the eavesdropper.
threshold for decoding the second layer at Bob is σ2 = 0.75. Future research direction is to extend the framework to a

From this figure, we observe that nt1 is only dependent on ǫE network, where Alice can estimate the instantaneous channel
because of the condition ǫB < ǫE and both Bob and Eve have quality and, Bob and Eve have similar channel conditions.

the same σ1 for decoding the first layer. Moreover, nt2 is only
R EFERENCES
dependent on ǫB as Bob needs to decode the second layer with
σ2 . As expected, the number of transmissions increases with [1] CISCO, “Cisco visual networking index: Global mobile data traffic
forecast update, 2015–2020,” 2016.
the increase in erasure probabilities. [2] A. Tassi, I. Chatzigeorgiou, and D. Vukobratovic, “Resource-allocation
Fig. 3 shows the intercept probability achieved by the frameworks for network-coded layered multimedia multicast services,”
derived analytical expressions and simulation results for both IEEE J. Sel. Areas Commun., vol. 33, no. 2, pp. 141–155, 2015.
[3] N. Thomos, E. Kurdoglu, P. Frossard, and M. Van der Schaar, “Adaptive
FT and NT systems against different erasure probabilities ǫB prioritized random linear coding and scheduling for layered data delivery
and ǫE . For the aforementioned N , σ1 and σ2 values, Alice from multiple servers,” IEEE Trans. Multimedia, vol. 17, no. 6, pp. 893–
∗ ∗
broadcasts nt1 first layer coded packets and nt2 second layer 906, 2015.
[4] M. Esmaeilzadeh, P. Sadeghi, and N. Aboutorab, “Random linear
coded packets as shown in Fig. 2. To obtain the simulation network coding for wireless layered video broadcast: General design
results, we generate 103 realizations and count the number of methods for adaptive feedback-free transmission,” 2014. [Online].
occurrences when Eve decodes the second layer and average Available: http://arxiv.org/abs/1411.1841
[5] Y.-S. Shiu, S. Y. Chang, H.-C. Wu, S. C.-H. Huang, and H.-H. Chen,
over all realizations to calculate the intercept probability. Fig. 3 “Physical layer security in wireless networks: a tutorial,” IEEE Wireless
also shows the intercept probability obtained by the analytical Commun., vol. 18, no. 2, pp. 66–74, 2011.
expression of [9] which considers a single layer video, and an [6] J. K. Tugnait, “Using artificial noise to improve detection performance
for wireless user authentication in time-variant channels,” IEEE Wireless
uncoded broadcast scheme which transmits an uncoded packet Commun. Lett., vol. 3, no. 4, pp. 377–380, 2014.
in each time slot following the order of packets. [7] M. Adeli and H. Liu, “On the inherent security of linear network coding,”
From this figure, we can see the tightness between analytical IEEE Commun. Lett., vol. 17, no. 8, pp. 1668–1671, 2013.
[8] H. Niu, M. Iwai, K. Sezaki, L. Sun, and Q. Du, “Exploiting fountain
expressions of (7) and (8) and simulation results. We can also codes for secure wireless delivery,” IEEE Commun. Lett., vol. 18, no. 5,
observe that intercept probability reduces when Bob channel pp. 777–780, 2014.
condition gets better compared to Eve channel condition. For [9] A. S. Khan, A. Tassi, and I. Chatzigeorgiou, “Rethinking the intercept
probability of random linear network coding,” IEEE Commun. Lett.,
example, for ǫE = 0.2, the intercept probability PN T
int reduces vol. 19, no. 10, pp. 1762–1765, 2015.
from 0.6 to 0.4 when the erasure probability of Bob channel
changes from ǫB = 0.09 to ǫB = 0.05. Note that the intercept

1089-7798 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

Você também pode gostar