Abstract
This study suggests an energy-scavenging capable unlicensed relay not only to retain communications between an unlicensed sender-recipient pair in underlay cognitive networks but also to secure these communications against eavesdropping of malicious users. Message-securing capability of such a network configuration is assessed through secrecy outage probability (SOP). For this purpose, a precise closed-form formula of the SOP accounting for interference power restriction, Nakagami-m fading, and maximum transmit power restriction is first proposed. Then, the proposed formula is validated by computer simulations. Ultimately, various results are supplied to contrive that the relay position, the time percentage, and the power percentage of the energy-scavenging technique should be appropriately chosen for achieving the best security performance. Moreover, the SOP decreases with lower severity fading level and is constant in the range of high maximum interference power or high maximum transmit power.
1. Introduction
Thanks to the advantage of efficient radio spectrum utilization, the spectrum-sharing paradigm (SSP) has lately acquired significant interest from both industry and academia [1]. According to the SSP, unlicensed users (UUs) are granted opportunities to send their message in the spectrum of licensed users (LUs) [2]. Nonetheless, UUs must frequently adjust their transmit power in order to conform to the interference power restriction inflicted by the successful signal reception of LUs and the maximum transmit power restriction inflicted by hardware specifications. These restrictions limit the power of unlicensed transmitters, rendering direct message transmission between an unlicensed sender-recipient pair unsuccessful. Additionally, this direct message transmission is presumably blocked due to severe shadowing, heavy path loss, and deep fading. Accordingly, to retain communications between the unlicensed sender-recipient pair, an unlicensed relay between this pair should be exploited to connect the sender with the recipient [3]. Nevertheless, as a volunteer, the relay is probably unenthusiastic to consume its individual power to forward the message of the unlicensed sender. Propitiously, currently advanced technologies make possible for wireless devices to be self-powered by scavenging energy in radio frequency (RF) signals [4]. Accordingly, the energy which the relay scavenges is used for relaying to lengthen the radio coverage of the sender, enhancing the communication possibility between the unlicensed sender-recipient pair. Conditioned on the short supply of the scavenged energy, the question is if the relay is capable of securing message transmission for the sender against the eavesdropping of illegal users in the information-theoretic viewpoint. This viewpoint affirms that the secured message transmission is when the trusted channel has a larger capacity than the wiretap channel [5]. Our paper aims to find a suitable answer to such a question.
1.1. Literature Review
Although various works analyzed the security performance of the direct message transmission (i.e., no relay) in underlay cognitive networks with energy scavenging (e.g., [6–14] and references therein), few publications [15–20] have paid attention to security evaluation for the underlay cognitive network with the energy-scavenging capable relay (UCNwEScR). To be more specific, the authors in [15] investigated the UCNwEScR in which the decode-and-forward unlicensed relay consumes its energy scavenged from the RF signals of the unlicensed sender and the licensed transmitter to forward the unlicensed sender’s message to the unlicensed recipient as the direct message transmission is unreliable. Also, the relay applies the power-dividing method for energy scavenging. However, Raghuwanshi et al. [15] merely presented the simulated SOP results. The problem in [15] was reconsidered in [16] with distinct aspects in which the relay performs the amplify-and-forward operation and scavenges the energy from merely the unlicensed sender’s signals relied on the time-dividing method. For the security capability enhancement, Maji et al. [17] further developed the work in [16] with permitting both the relay and the sender to interfere the eavesdropper. The study in [18] kept improving [17] by selecting an appropriate relay for higher security capability. For further security capability improvement, Hieu et al. [19] suggested to select a multihop relaying link offering the largest spectral efficiency. Nonetheless, Hieu et al. [19] relied on committed beacons for the relays to harvest the energy with the time-dividing method. Moreover, Hieu et al. [19] merely implemented the analysis on the connection outage probabilities (the connection outage probability is the possibility which the signal-to-noise ratio (SNR) does not exceed a preset value) at the eavesdropper and the recipient. In [20], an analysis framework was proposed to assess the eavesdropping-decoding trade-off in the UCNwEScR where the unlicensed relay only harvests the energy in the unlicensed sender’s RF signals. The eavesdropping-decoding trade-off is represented by the relation between the possibilities which the unlicensed recipient and the eavesdropper unsuccessfully restore the sender’s message. Relied on the information-theoretic viewpoint, the SOP analysis is more vital than the eavesdropping-decoding trade-off analysis. Accordingly, the former should be carried out to evaluate the security capability of the UCNwEScR before practical deployment. It is reminded that all studies in [15–20] merely considered the UCNwEScR over Rayleigh fading channels and under the interference power restriction (it should be emphasized that Ho-Van and Do-Dac [21] provided the SOP analysis for the UCNwEScR with the system model similar to [15–20]; nonetheless, Ho-Van and Do-Dac [21] neglected the interference power restriction but considered the average scavenged energy and the outage probability restriction on the licensed receiver; although the interference power restriction advances the average scavenged energy and the outage probability restriction in ensuring the short-term signal reception quality at the licensed receiver (i.e., the latter ensures the long-term one), the former requires instantaneous channel information which complicates the SOP analysis; accordingly, the literature review [15–20] merely concentrates on the publications related to the interference power restriction which is also considered in this paper) and the maximum transmit power restriction. Also, it is common knowledge that the Nakagami-m fading channels are suitable and general to fit realistic ones (e.g., the Nakagami-m fading is reduced to the Rayleigh fading when m equals 1) [22]. Accordingly, it is of utmost importance to perform the SOP analysis for the UCNwEScR over Nakagami-m fading channels (it is noted that Lei et al. [23] performed the SOP analysis for underlay cognitive networks over Nakagami-m fading channels but did not study energy scavenging). The current paper aims to reach this objective.
1.2. Contributions
The current paper differs from the studies [15–20] in the following matters:(i)The relay is active merely as long as it exactly restores the unlicensed sender’s message. This (this paper and the previous studies [15, 18, 19] exploited the decode-and-forward relay while the studies in [16, 17, 20] employed the amplify-and-forward one) prevents the error propagation (the error propagation is due to the fact that the relay simply forwards the decoded information of the sender whether or not the decoded information is correct; this error propagation used to happen in noncognitive radio networks without energy scavenging) in [15, 18, 19] and the noise enhancement in [16, 17, 20].(ii)This paper investigates the Nakagami-m fading, which generalizes the Rayleigh fading in [15–20].(iii)The time percentage for energy scavenging is a variable, which is convenient for optimization of the SOP. It is reminded that Raghuwanshi et al. [15] fixed the time percentage of 0.5.(iv)The SOP is analyzed in a precise closed form while Raghuwanshi et al., Benedict et al., and Maji et al. [15–18] merely presented the simulated SOP results and Hieu et al. [19] and Ho-Van and Do-Dac [20] merely analyzed the possibilities that the unlicensed recipient and the eavesdropper unsuccessfully restore the unlicensed sender’s message.
Our contributions are listed as follows:(i)Exploiting the unlicensed relay capable of energy scavenging to secure communications between the unlicensed sender-recipient pair as the direct message transmission is unreliable. The relay must assess the reliability of the received signal before agreeing to relay the sender’s message.(ii)Proposing precise closed-form formulas for vital security capability measures such as the SOP, the intercept possibility (IP) (the IP is the possibility which the secrecy capacity is negative [24]), the possibility of positive secrecy capacity (PPSC) (the PPSC is the possibility which the secrecy capacity is positive [25]), under the maximum transmit power restriction, the Nakagami-m fading, and the interference power restriction to quickly assess the security capability of the UCNwEScR without the need of time-consuming computer simulations.(iii)Optimizing key specifications based on the proposed formulas.(iv)Providing numerous results to expose insights of the security capability, for example, the SOP saturation when one of the two power restrictions is ignored; considerable security capability improvement such as locating the relay and selecting the time percentage for energy scavenging and the power percentage for message decoding is appropriate.
1.3. Organization
This paper is organized as follows. The description of the system model under consideration is presented in Section 2. Then, the precise closed-form formulas of the SOP, the IP, and the PPSC are derived in Section 3. Ultimately, results and conclusions are sequentially presented in Sections 4 and 5.
2. System Model
Relevant users in the considered system model of the underlay cognitive radio with the energy-scavenging capable unlicensed relay R are denoted in Figure 1. The unlicensed sender S broadcasts legitimate message to the unlicensed recipient D. This transmission is wire-tapped by the eavesdropper E and interferes the signal reception of the licensed receiver P (it is almost assumed in publications related to the SSP (e.g., [26, 27] and references therein) that licensed transmitters may inflict the ignorable amount of interference to unlicensed recipients; this can be justified due to long distance between licensed transmitters and unlicensed recipients or such Gaussian-distributed interferences). Due to unexpected causes (e.g., severe shadowing, deep fading, and heavy path loss), D and E may receive the signals of S not strong enough for successful decoding. Consequently, S should ask R, which is between S and D, for help in relaying its message to D. As such, two transmission stages are required for message from S to reach D as shown in Figure 2(a).
(a)
(b)
To facilitate performance optimization, durations of two stages are different: for Stage 1 and for Stage 2, where is the time percentage and ψ is the time to finish message transmission from S to D. Stage 1 is for S to broadcast its message while Stage 2 is for R to forward its received message to D. To avoid noise enhancement of the amplify-and-forward operation, R chooses to operate in the decode-and-forward manner. Moreover, R has self-powered capability by harvesting energy in RF signals. For system resource saving, R employs the power-dividing method [28, 29] as shown in (the power-dividing model is linear in Figure 2, which was assumed in most recent publications [6–21, 26, 29]) Figure 2(b) for energy scavenging. This method divides the signal received at R into two fractions with different powers which are represented by the power percentage : one fraction for restoring the sender’s message (it is common in existing works (e.g., [6–12, 15] and references therein) that the message decoder consumes the negligible amount of power for its operation; therefore, its energy consumption can be ignored) and the other for scavenging the energy.
Fading channels under consideration are independent and Nakagami-m distributed. Channel coefficients are denoted in Figure 1 where with and stands for the coefficient of the h-k channel. For the Nakagami-m distribution, is represented by a parameter pair where with being the fading power and being the average operator with respect to the random variable X. Under investigation of the path loss, can be represented as where is the h-k distance and ϕ is the path-loss exponent. Accordingly, the channel gain has the cumulative distribution function (CDF) and the probability density function (PDF), correspondingly, aswhere , is the gamma function and is the lower incomplete gamma function ([30], equation (8.350.1)).
In Stage 1, S performs message transmission while R processes the received signal from S. Therefore, R receives the following signal:where is S’s transmit power, is the unity-power transmit symbol of S, and is the additive noise produced by the receive antenna at R.
For the SSP [31], S’s transmit power must comply with the interference power restriction and the maximum transmit power restriction. The former guarantees that S does not interfere the licensed receiver with the interference power exceeding a tolerable threshold, namely, the maximum interference power T, i.e., . In the meantime, the latter limits S’s transmit power by the maximum transmit power K inflicted by hardware specifications, i.e., . Accordingly, these two restrictions establish S’s transmit power as
Processing the received signal at R follows the diagram in Figure 2(b): is divided into two fractions. The first fraction of is for scavenging the energy and the second fraction of is for restoring S’s message. The energy which R scavenges from the first fraction is given bywhere φ, , is an energy conversion efficiency of the energy scavenger at R.
The time of Stage 2 is ; hence, the power which R consumes in Stage 2 is given by
The input of the message decoder in Figure 2(b) is represented aswhere is the additive noise inflicted by the passband-to-baseband signal conversion.
Embedding (3) in (7), one can rewrite (7) as
Based on (8), the performance of the message decoder relies on the following SNR:where
The channel capacity which R can obtain is bits/s/Hz where the pre-logarithm factor μ appears since the time of Stage 1 is . Relied on the communication theory, R correctly restores S’s message as its channel capacity exceeds the preset spectral efficiency , i.e., . Equivalently, is successfully recovered at R if where .
In Stage 2, the relay whose transmit power is denoted as sends the restored symbol as long as it correctly recovers S’s message (i.e., and ). Otherwise, it is in the standby mode. As such, D and E receive the following signals, respectively:where and are the additive noises produced by the receive antenna at E and D, respectively, and is R’s transmit power, which is given bybecause R operates in the SSP.
Based on (11) and (12), one infers the SNRs at D and E as
Then, E and D can achieve the following channel capacities, respectively [32]:
In (16) and (17), the time of Stage 2 is which introduces the pre-logarithm coefficient of .
According to [5], the secrecy capacity of the UCNwEScR is the capacity difference between the R-D trusted channel and the R-E wiretap channel:where represents .
3. Security Analysis
This section aims at deriving a precise closed-form formula for the SOP of the UCNwEScR. The SOP is the possibility which the secrecy capacity is smaller than a preset security level . Accordingly, the smaller SOP means more secure message transmission. As such, the communication theory claims the SOP as a vital measure for assessing the security capability of wireless communications. The SOP analysis is helpful not only for assessing message securing capability without exhaustive simulations but also to obtain other useful security capability measures such as the IP and the PPSC.
By representing as the possibility of the event , one expresses the SOP as
As shown in (18), exists for ; hence, (19) can be computed in two different cases as
Furthermore, the required security level is positive (i.e., ); hence, (20) can be simplified aswhere .
In order to analytically evaluate , the term Q must be first computed as
Observing (14) and (15), one recognizes that and have a common term as . Accordingly, correlates . To decorrelate them, Q in (22) should be computed through the conditional probability as
To derive I, one further rewrites it in an explicit form as
Conditioned on , it is apparent that and are statistically independent. Consequently, can be represented as . Using this fact, (24) can be further simplified as
In the sequel, and are computed to completely derive (25). First, based on the definition of the CDF, is derived as
Then, substituting (14) into (26), one achieves
By transforming the variable, can be deduced as
By changing variables reasonably in (27) and (28) and then substituting the result into (25), one achieves
Usingin [30], equation (8.352.1), (29) can be expanded as
Applying the binominal expansion with being the binomial coefficient, (31) can be expanded as
By exerting [30], equation (3.351.3), one computes the last integral in (32) as
Substituting (33) into (23), one obtains
Given Q in (34), one can rewrite (21) as
Plugging in (33) into (35), one obtains
Using (13), the term X in (36) is simplified as
Inserting the PDF of , which has a general form as (2), into (37), one simplifies (37) as
The last integral in (38) is computed with the help of [30], equation (3.351.2); hence, one simplifies (38) as
Using (30) to decompose the function in (39), one obtains
After some manipulations, one rewrites (40) in a compact form aswhere
Obviously, X in (41) is represented in closed form if in (44) is also solved in closed form. Next, the derivation of in (44) is presented.
The scavenged power of R, , in (6) can be rewritten aswhere
Using (45) to rewrite in (44) gives
In order to complete the computation of the last integral in (47), the PDF of must be derived first. Towards this end, the CDF of , , is derived as follows and then the derivative of is taken to result in .
Theorem 1. has a precise closed form aswhere
Proof. See Appendix A.
The PDF of is produced from the derivative of as
Theorem 2. has a precise closed form aswhereand denotes the exponential integral [30].
Proof. See Appendix B.
The precise closed-form representation of in (54) makes X in (41) to be represented in the precise closed form. Accordingly, in (36) is also derived in the precise closed form, which is favourable in assessing the security performance of the UCNwEScR under the interference power restriction, the Nakagami-m fading, and the maximum transmit power restriction without long computer simulations. Relied on our knowledge, this formula of has not been reported yet. Additionally, some vital security performance measures (e.g., the PPSC or the IP) can be straightforwardly derived from this analysis. Specifically, the PPSC is the possibility which the secrecy capacity is greater than zero, i.e.,
Furthermore, the IP is the possibility which the secrecy capacity is negative [24], i.e.,
4. Illustrative Results
This section evaluates the SOP of the UCNwEScR through vital specifications. For demonstration purpose, the path-loss exponent is exemplified as and the coordinates of S, D, R, E, and P are randomly adopted as , , (d, 0.0), , and , respectively. Thereafter, “Sim.” and “Ana.” stand for the analytical result in (36) and the simulated result, respectively; the energy conversion efficiency is adopted; without loss of generality, equal noise variances and equal severity levels are assumed, i.e., and for and .
The SOP with respect to (w.r.t) the maximum transmit power-to-noise variance ratio is illustrated in Figure 3 for bits/s/Hz, , , , bits/s/Hz, and dB. This figure shows the coincidence of the simulation with the analysis, verifying the exactness of (36). Moreover, the security performance is enhanced with increasing . This is obvious because increasing offers R more opportunities to harvest more energy in S’s transmit signal and to correctly restore S’s message, ultimately declining the outage possibility in Stage 2. Nonetheless, the security capability becomes constant at large . The unchanged security capability is due to the power allotment of R and S (observe (4) and (13)) where the powers of R and S do not depend on for large (i.e., the maximum transmit power restriction can be ignored in the case of large ), rendering the SOP unchanged. Furthermore, the SOP decreases with lower fading severity level (i.e., increasing m) as expected.
For the similar specifications as Figure 3 with the exception of dB, Figure 4 illustrates the SOP w.r.t the maximum interference power-to-noise power ratio . The simulated/analytical results confirm the coincidence of the analysis with the simulation, again justifying the derivation of (36). Additionally, the security performance is improved with increasing and is constant at large . These results are comprehended from the power allotment of R and S, identical to Figure 3. Furthermore, the SOP reduces with lower fading severity level.
Figure 5 shows the SOP w.r.t the S-R distance d for bits/s/Hz, dB, dB, , , and bits/s/Hz. The results expose the coincidence of the analysis with the simulation, proving the exactness of (36). It is recalled that as R does not exactly restore S’s message (i.e., the S-R distance is long) or R does not reliably forward S’s recovered message to D (i.e., the R-D distance is long), the secrecy outage happens. Accordingly, an optimal relay position that well balances the possibility which R can correctly recover S’s message and the possibility that R can reliably forward S’s recovered message to D to minimize the SOP will exist. Figure 5 illustrates this reasoning where the security performance is optimized when R is away from S for , respectively. Furthermore, lower fading severity level improves the security performance as expected.
Figure 6 demonstrates the SOP w.r.t the time percentage μ for , bits/s/Hz, dB, , bits/s/Hz, and . The results demonstrate the coincidence of the analysis with the simulation, confirming the accuracy of (36). Additionally, optimal μ (e.g., corresponding to as illustrated in Figure 6) for the best security performance exists. The presence of is comprehended as follows. The increment of μ stretches out the time of Stage 1; therefore, R harvests more energy and restores S’s message more exactly. Nonetheless, the increment of μ also decreases the secrecy capacity in Stage 2; therefore, the security capability reduces. As such, μ should be optimized to balance the times of two stages for the best security capability. Moreover, the best security performance is enhanced with lower fading severity level.
Figure 7 shows the SOP w.r.t the power percentage ρ for , , bits/s/Hz, , bits/s/Hz, and . The results expose the coincidence of the analysis with the simulation, certifying the accuracy of (36). In addition, optimal ρ (e.g., corresponding to as demonstrated in Figure 7) for the best security capability exists. The presence of is explained as follows. The increment of ρ creates more opportunities for R to harvest higher energy; therefore, R increases the signal reception quality in Stage 2, eventually improving the security capability. Nonetheless, the increment of ρ also declines the energy for the message decoder, thus mitigating the possibility which R exactly recovers S’s message and leading to more secrecy outage in Stage 2. Accordingly, ρ should be optimized to trade off the transmission reliability of R and S in two stages. Moreover, the best security performance is enhanced with lower fading severity level.
Figure 8 illustrates the SOP w.r.t the preset spectral efficiency for , , , , , and bits/s/Hz. This figure exposes the coincidence of the analysis with the simulation, confirming the accuracy of (36). Additionally, the SOP rises with . This makes senses because the higher , the lower the possibility which R successfully restores S’s message, leading to the higher SOP. Furthermore, the SOP is reduced with lower fading severity level.
For the similar specifications as Figure 8 with the exception of bits/s/Hz, Figure 9 plots the SOP w.r.t the preset security level . The results confirm the coincidence of the simulation with the analysis, again validating the derivation of (36). Additionally, the security performance is degraded with increasing . This is reasonable because under fixed system parameters, the network only obtains a certain security level. Therefore, the higher , the more the outage events. Moreover, the security performance is improved with lower fading severity level.
5. Conclusion
The precise closed-form formula of the SOP for the underlay cognitive network with the energy-scavenging capable relay under the maximum interference power restriction, the Nakagami-m fading, and the transmit power restriction was proposed in the current paper. The proposed formula was certified by computer simulations. Multiple results demonstrated that the deployment of the self-powered relay can drastically enhance the security capability even when the sender-recipient channel is inapplicable due to heavy shadowing, deep fading, severe path loss (i.e., the direct communication suffers the SOP of 1). Additionally, the best security performance can be achieved by appropriately locating the relay and selecting the parameters of energy-scavenging and message transmission processes (i.e., the time percentage and the power percentage). Moreover, the security capability is unchanged at high maximum interference power or high maximum transmit power. Furthermore, the security performance is improved with lower severity fading level.
Appendix
A. Proof of Theorem 1
The CDF of , , is defined as
Using (30) and the PDF of , which has a general form as (2), (A.1) can be simplified as
The last two integrals in (A.2) is computed by using equation (3.351.1) in [30] and equation (3.351.2) in [30]. As such, (A.2) is expressed in closed form as
Using compact notations in (49)–(52), respectively, one can simplify (A.3) as (48). This finishes the proof.
B. Proof of Theorem 2
Embedding (53) in (47), one obtains
By rearranging the terms in (B.1), one achieves
By denotingit is apparent that (B.2) coincides with (54). Therefore, the last step to finish the proof of Theorem 2 is to justify that and are written in closed form as (55) and (57), respectively. Towards this end, using the series expansion for , (B.3) can be rewritten as
By changing the variable, (B.5) is simplified as
Using the binomial expansion, one obtains
Letwhich is expressed in closed form as (56) by exerting [30], equation (358.4).
Based on (B.8), the last integral in (B.7) is computed as . This makes (B.7) coincide with (55).
Following the derivation of , one derives
Performing the partial fraction decomposition to , one simplifies (B.9) aswhere and are defined in (58) and (59), respectively.
The integrals in (B.10) can be expressed in the function, making (B.10) coincide with (57). This completes the proof.
Data Availability
The data used to support the findings of this study are included within the article.
Conflicts of Interest
The authors declare that there are no conflicts of interest regarding the publication of this paper.
Acknowledgments
This research was funded by Vietnam National University Ho Chi Minh City (VNU-HCM) under grant no. B2019-20-01.