Jammer-Assisted Resource Allocation in Secure OFDMA With Untrusted Users

01/11/2019 ∙ by Ravikant Saini, et al. ∙ Indian Institute of Technology Delhi 0

In this paper, we consider the problem of resource allocation in the orthogonal frequency division multiple access system with single source and M untrusted users in presence of a friendly jammer. The jammer is used to improve either the weighted sum secure rate or the overall system fairness. The formulated optimization problem in both the cases is a mixed integer non-linear programming problem, belonging to the class of NP-hard. In the sum secure rate maximization scenario, we decouple the problem and first obtain the subcarrier allocation at source and the decision for jammer power utilization on a per-subcarrier basis. Then, we do joint source and jammer power allocation using primal decomposition and alternating optimization framework. Next, we consider fair resource allocation by introducing a novel concept of subcarrier snatching with the help of jammer. We propose two schemes for jammer power utilization, called proactively fair allocation (PFA) and on-demand allocation (ODA). PFA considers equitable distribution of jammer power among the subcarriers, while ODA distributes jammer power based on the user demand. In both cases of jammer usage, we also present suboptimal solutions that solve the power allocation at a highly reduced complexity. Asymptotically optimal solutions are derived to benchmark optimality of the proposed schemes. We compare the performance of our proposed schemes with equal power allocation at source and jammer. Our simulation results demonstrate that the jammer can indeed help in improving either the sum secure rate or the overall system fairness.



There are no comments yet.


page 1

This week in AI

Get the week's most popular data science and artificial intelligence research sent straight to your inbox every Saturday.

I Introduction

Rapid deployment of wireless communication systems has always been engraved with the issue of security of the transmitted data. The basic reason is the broadcast nature of transmission which makes the signals vulnerable to tapping by malicious users [1]. Security of the transmitted signal is generally considered as a responsibility of the higher layers which employ cryptographic techniques. This strategy relies on the basic assumption that the enciphering system is unbreakable by the malicious users [2]. With growing computing capabilities, such measures prove to be insufficient and it has motivated the research community to explore security at the physical layer. Physical layer security finds its basis from independence of the wireless communication channels and has a low implementation complexity [3]. This added layer of security is considered as the strictest kind of security, not requiring even any kind of key exchange [1, 2].

Orthogonal frequency division multiple access (OFDMA) is a potential physical layer technology for the next generation access networks such as WiMAX, LTE, and beyond. Hence, the study of physical layer security in OFDMA has gained considerable attention in recent years [4, 5, 6, 7, 8, 9, 10, 11, 12]. The subcarrier allocation and power optimization in multi-node secure OFDMA system has been studied in two cases: in one scenario transmitter assumes all users to be trusted and there exists an external eavesdropper which tries to decode the data of trusted users [4, 5, 6, 7, 8], while in another scenario there is a mutual distrust among the users and a particular user may demand the source to transmit its data considering all other users as the potential eavesdroppers [11, 10, 12].

In trusted users case, the authors in [5] considered maximizing the minimum of weighted sum secure rate of all the users. They proposed a complex MILP based solution and less complex suboptimal schemes. However, the issue of resource scarcity of a user that is very near to the eavesdropper was not highlighted. Maximization of energy efficiency with multi-antenna source, eavesdropper, and single antenna users was studied in [6]

with bounds on per-user tolerable secrecy outage probability. The average secrecy outage capacity maximization in the same setup with a multi-antenna Decode and Forward relay was studied in

[7]. The authors in [13] studied secure rate maximization under fixed quality of service (QoS) constraints for secure communication among multiple source-destination pairs in the presence of multi-antenna external eavesdropper, with the help of multiple single antenna Amplify and Forward (AF) relays. An extension of the above proposed schemes to multiple eavesdroppers case may be straight forward, but with untrusted users the problems present an interesting challenge with multiple unexplored facets. Looking at another domain of untrust, subcarrier assignment and power allocation problem was studied in [9] in an AF untrusted relay aided secure communication among multiple source-destination pairs.

In an effort to tackle the untrusted users, the authors in [10] proposed to allocate a subcarrier to its best gain user in a two-user OFDMA system and presented the optimal power allocation. The scheme can be trivially extended to more than two users for sum rate maximization. In a scenario having heterogeneous demand of resources by users, the authors in [11] proposed a joint subcarrier and power allocation policy for two classes of users: secure users demanding a fixed secure rate and normal users which are served with best effort traffic. In the cognitive radio domain, secure communication over a single carrier, between multi-antenna secondary transmitter and a fixed user considering other secondary and primary users as eavesdroppers was studied in [14]. Similarly, precoder design to maximize the sum secrecy rate to achieve confidential broadcast to users with multiple antennas was studied in [15, 16]. The authors in [12] solved the resource allocation problem among multiple source-destination pairs in a relay-aided scenario with untrusted users. However, in secure OFDMA fair resource distribution among untrusted users poses new challenges which have not yet been investigated.

I-a Motivation

The study in [11] raised the feasibility issue of resource allocation problem because of the channel conditions, i.e., some users may not achieve the required secure rate due to tapping by untrusted users. This motivates us to investigate if the secure rate over a subcarrier could be any how improved. The strategy proposed in [8] utilizes interference to improve the secure rate of a user. But this scheme cannot be used in multi-user untrusted scenario because it relies on a strong assumption that the jammer affects the eavesdropper only. It was shown in [17] that, for single user multi-antenna system, secrecy rate can be improved by a multi-antenna source by jointly transmitting message in the range space and interference in the null space of the main channel. Since this solution is difficult to apply for single antenna systems, the authors in [18] showed the possibility of achieving a positive rate even when eavesdropper’s channel is stronger compared to main channel by using jamming power control only. The studies in [8, 17, 18] are based on single user system. The implications of utilizing interference in multi-user scenario are yet to be investigated. With regard to fair resource allocation without the help of interference, the strategy proposed in [5], which balances the secure rate among users, can limit the maximum achievable rate of the system because of a poor user facing strong eavesdropping by other users. This leads to system resource wastage. Effectively utilizing jammer power control in multiple untrusted user scenario for sum rate maximization or fair resource allocation raises challenges which to the best of our knowledge has not yet been investigated in the literature.

In this paper, we intend to explore the role of the jammer power in alleviating the following issues which plague single antenna secure OFDMA system with untrusted users:

  • Secure rate achievable by the users can be very low as compared to the single external eavesdropper case, because now there are wiretappers instead of one;

  • A large number of users can starve for subcarriers, because a group of users with very good channel gains may prohibit secure communication to the other users.

Assuming the jammer to be a node affecting all the users, we intend to use the jammer power for individual and independent jamming over the subcarriers. The resource allocation problem in presence of jammer is a complex Mixed Integer Non Linear Programming (MINLP) problem belonging to the class of NP-hard [19, 20]. Due to the requirement of decision on subcarrier allocation at the source as well as to use jammer power over a subcarrier, the problem exhibits combinatorial nature having exponential complexity with number of subcarriers [21]. Hence, instead of attempting a global optimal solution, we study the resource allocation problem to improve either the sum secure rate or the overall system fairness. While extending our preliminary work in [22]

, where we presented two suboptimal solutions based on sequential source and jammer power allocation, we conduct a deeper study on the behavior of secure rate with jammer power and solve the joint source and jammer power optimization problem. For max-min fair resource allocation, two novel methodologies of jammer power utilization are introduced. We also present the asymptotic analysis of the algorithms with

and , and discuss the computational complexity of all the proposed schemes under both the usages of jammer.

I-B Contribution

Our contributions can be summarized as answers to the following two questions:

(1) How to improve sum secure rate in OFDMA systems with untrusted users? To address this question, we consider the possibility of using jammer power over a subcarrier to improve the secure rate beyond what can be achieved by source power only. The features of the proposed solution are as follows:

  • We obtain the constraints of secure rate improvement over a subcarrier and show that the secure rate in this constrained domain is a quasi-concave function of jammer power, thereby offering a unique maxima.

  • The jammer aided approach introduces a new challenge - referred as SNR reordering in the rest of the paper - which makes the problem combinatorial even after subcarrier allocation and jammer utilization decision. We introduce the concept of constrained-jamming by developing jammer power bounds to handle this challenge.

  • For the known subcarrier allocation and jammer decision, the joint source and jammer power allocation problem is solved optimally using primal decomposition (PD)[23] and alternating optimization (AO)[24] techniques.

  • We analyze the complexity of the proposed algorithm and show its convergence in finite steps.

  • We also propose less complex solution which allows a trade-off between performance and complexity.

  • Asymptotically optimal solution is derived to assess optimality of the proposed scheme.

(2) How to remove subcarrier scarcity and have fair distribution? In secure OFDMA with untrusted users and in absence of jammer, the best gain user over a subcarrier is the strongest eavesdropper for all other users. Thus, a subcarrier should normally be given to its best gain user, which may cause some users to starve for channel resources. To attain a better fairness, we propose to take away a subcarrier from its best gain user and allocate it to another user with poor channel gain with the help of jammer. The features of the proposed solution are:

  • Secure rate over the snatched subcarrier is positive when the jammer power is above a certain threshold and attains a maximum value at an optimal jammer power.

  • The conventional max-min fairness algorithm cannot be employed in secure OFDMA because a user with poor channel gains over all the subcarriers may force the algorithm into a deadlock. The proposed max-min fair scheme offers a graceful exit with such users.

  • Two variants of max-min fairness algorithms are proposed where the jammer power can be utilized either by preserving it for possible snatching over each subcarrier, or by allocating it based on the demand of snatching user.

  • The power allocation is done by PD and AO process while the issue of signal to noise ratio (SNR) reordering is handled similarly as in sum rate maximization case.

  • Low complexity solutions have been presented for both the variants of max-min fairness algorithms.

  • We obtain an asymptotic upper bound of the fairness achievable by our proposed scheme.

The remainder of the paper is organized as follows: The system model is introduced in Section II. The sum secure rate maximization problem is studied in Section III, followed by fair resource allocation study in Section IV. Asymptotic analysis has been presented in Section V. Simulation results are discussed in Section VI. Section VII concludes the paper.

Ii System model

We consider the downlink of an OFDMA system with a single source (base station), a friendly jammer, untrusted mobile users (MUs) which are randomly distributed in the cell coverage area, and subcarriers. The jammer, which is considered as another node in the network (e.g., in LTE-Advanced it could be an idle relay node), is controlled by the source to help improve the overall system performance. Jammer is assumed to be capable of collecting channel state information (CSI) in the uplink and sending jamming signal which is unknown to users on the downlink. Since all users request secure communication with the source, they share their CSI with the source as well as the jammer. In the present study we assume the jammer to be located randomly in the cell area.

Source-to-MUs and jammer-to-MUs channels are considered to experience slow and frequency-flat fading, such that the channel parameters remain constant over a frame duration but vary randomly from one frame to another. Perfect CSI of source-to-MU and jammer-to-MU channel pairs for all the MUs is assumed to be available at source [4, 11, 6, 7, 5, 8, 9, 10]. Source utilizes this information for subcarrier allocation, and source and jammer power allocation. We consider that a subcarrier is exclusively allocated to one user only, which has been proved to be optimal for sum secure rate maximization [10].

Fig. 1: System model

The secure OFDMA system with untrusted users is a multiple eavesdropper scenario, where for each main user there exist eavesdroppers. Out of these eavesdroppers, the strongest one is considered as the equivalent eavesdropper (hereafter referred as eavesdropper). The secure rate of the main user over a subcarrier is defined as the non-negative capacity difference between the main user and the eavesdropper. Let be the source to th user channel coefficient and be the jammer to th user channel coefficient over subcarrier as shown in Fig. 1. Then, the secure rate of user on th subcarrier is given as [10, 11]:


where ,

is the AWGN noise variance,

and are respectively source and jammer powers over subcarrier . is an indicator of absence or presence of jammer power on subcarrier such that , and is the eavesdropper. The non-linearity of secure rate in source and jammer powers along with the operator complicate the optimal resource allocation which will be discussed in the Section III and IV.

Example: Let us consider a symbolic 3 user 5 subcarrier OFDMA system, with the respective channel gains and given in Tables I and II. This system will be used to illustrate the key concepts developed further in the paper.      

Iii Resource allocation for sum secure rate maximization

In this section we discuss the joint source and jammer resource optimization problem for weighted sum secure rate maximization. The problem can be stated as:

subject to

where is a binary allocation variable to indicate whether subcarrier is given to user or not, is the priority weight allocated by the higher layers to user , and and are source and jammer power budgets, respectively. and are budget constraints, is jammer allocation constraint, and are subcarrier allocation constraints, and denotes source and jammer power boundary constraints.

The optimization problem in (III) has total four variables per subcarrier: ,

as binary variables; and

and as continuous variables. Since the problem is a non-convex combinatorial problem belonging to the class of NP-hard, there is no polynomial time optimal solution possible [19, 20]. We tackle the problem by breaking it in parts and attempt to find a near-optimal solution, which approaches asymptotically optimal solution (discussed in Section V) as and increases. For this, first we perform subcarrier allocation without considering the jammer. Next we distinguish those subcarriers over which jammer can improve the secure rate, and finally we complete the joint power allocation.

1.1027 0.3856 0.6719 1.2101 0.7043
0.7423 1.0735 0.6558 1.0006 0.8943
0.7554 1.4772 0.2498 1.3572 3.5391
TABLE I: Source-users channel gains
3.3624 6.0713 3.4125 3.0584 0.4987
8.1741 7.0607 4.1047 0.9860 1.6860
0.9028 2.0636 0.5605 3.0277 4.5346
TABLE II: Jammer-users channel gains

Iii-a Subcarrier allocation at source

In presence of jammer, subcarrier allocation at source, i.e., decision is difficult. In secure OFDMA with untrusted users a subcarrier can only be given to the user having maximum SNR over the subcarrier. Any change in may force the decision of subcarrier allocation to change because of the possible reordering of updated SNRs (cf. (II)). Since is in numerator, any change in does not affect the SNR ordering. This makes the problem combinatorial, as we may need to check subcarrier allocation for every update of . In order to bypass this step, we initially assume that the jammer is not present. In absence of jammer , and the secure rate definition in (II) changes to


As observed in (III-A), is required to have positive secure rate over subcarrier , irrespective of . Thus, the subcarrier allocation policy, allocating a subcarrier to its best gain user can be stated as:

  • Note that the subcarrier allocation (4) depends only on users’ gains. It is indifferent to users’ priority imposed through weights which will play their role in power allocation.

Iii-B Subcarrier allocation at jammer and jammer power bounds

In the presence of jammer, even after subcarrier allocation at source is done, the problem is still combinatorial due to . Any change in may even jeopardize the earlier decision on as described in the previous section. In-order to solve this issue, we first introduce the concept of rate improvement, which will help us decide retaining the decision on .

Iii-B1 Selective jamming for secure rate improvement

Since jammer affects all the users, it appears that using jammer over a subcarrier may degrade the secure rate. But it is interesting to note that, with jammer power the secure rate can be improved beyond what is achieved without jammer. For the proof of concept, let us consider a simple OFDMA scenario having four nodes: source, jammer, and two users and . Let a subcarrier be allocated to user , and be the eavesdropper. The following proposition describes the possibility of secure rate improvement and the existence of optimal jammer power achieving maximum secure rate over subcarrier .

Proposition III.1.

The secure rate over a subcarrier having can be improved if and the source and jammer powers and are constrained as

and (5)

where and .

In the constrained domain of rate improvement, the rate is a quasi-concave function of having a unique maxima.


See Appendix A. ∎

Example (continued): Observing the source gains in Table I, the best gain users of the subcarriers are , respectively. Users are their corresponding eavesdroppers. Subcarriers satisfy the condition of secure rate improvement, i.e., . Considering the unit of transmit power is in Watt, let , , and . Source power thresholds for the three subcarriers are , respectively. Assuming equal source power allocation over all subcarriers () and comparing with , subcarriers can be utilized for secure rate improvement while cannot be used. The corresponding jammer power thresholds for are respectively found as . Variation of SNRs of the users and the secure rate of user on subcarrier with jammer power are presented in Table III. As observed, with is higher compared to the value when , till . has a maxima between and , at .

0.1487 0.0317 0.0178 0.0033 0.0030
1.1524 0.1925 0.1050 0.0189 0.0175
2.1821 1.5304 1.1784 0.3571 0.3339
0.6988 1.5518 1.4720 0.7239 0.6882
TABLE III: Users’ SNRs and secure rate versus

Based on the results described in Proposition III.1, we create a prospective set of subcarriers, over which the jammer power can be used for secure rate improvement. The subcarrier allocation policy at jammer can be summarized as:


While extending the result of Proposition III.1 to , there is an inherent challenge associated with the allocation of , which we refer as SNR reordering. Without loss of generality, let the channel gains from source-to-MUs be sorted as over a subcarrier such that user 1 is assigned the subcarrier and user 2 is the eavesdropper. While optimizing source and jammer powers jointly, any update in does not disturb the resource assignments, but as is updated it raises the following two concerns:
(i) User 1 may not have the maximum SNR.
(ii) User 2 may not remain the corresponding eavesdropper.
Because of this SNR reordering challenge, for every update a new main user and the corresponding eavesdropper are to be determined. This is the problem introduced by the operator appearing in the rate definition (II), which makes the joint source and jammer power allocation a tedious task even after and allocation. In order to address this challenge we develop a strategy as described below.

Iii-B2 Bounds on jammer power to avoid SNR reordering

In order to handle the operator, we enforce certain constraints over jammer power so as to retain the same main user and the same eavesdropper throughout the jammer power allocation. In order to retain the eavesdropper, the rate of user should be larger than all other possible eavesdroppers, i.e.,


Similarly, to preserve the main user we need to have:


The constraints in (7) and (8) depend on channel conditions only, which result in lower and upper bounds over such that . Since (III.1) enforces another upper bound on , the final lower and upper bounds are given as

  • Since , hence (7) and (8) may result only in upper bounds, and lower bounds are zero. In case , there is no feasible region for and subcarrier cannot be considered for rate improvement.

Example (continued): The SNR reordering issue can be explained with the help of jammer power variation over subcarrier . The respective SNRs of the users with variation are presented in Table IV. From to , and are respectively the main user and the eavesdropper. At , the eavesdropper changes from to and this reduces the secure rate from to . When increases to , even the main user changes from to , and now is the corresponding eavesdropper. Now, the secure rate of is , while that of is . Though the jammer power threshold guarantees that, if , the secure rate remains greater than that without jammer (), this assertion relies on the basic assumption that the main user and the eavesdropper do not change. The jammer power threshold, optimum jammer power, and its upper bound to avoid SNR re-ordering on are respectively , , and . As soon as , the SNR order changes and is also reduced. This issue did not arise in as there was no finite jammer power upper bound .

After decision of and , combinatorial aspect of the problem is resolved if remains within bounds. Next we consider the joint source and jammer power allocation.

0.4514 0.2086 0.0798 0.0662 0.0493
0.4301 0.1602 0.0556 0.0456 0.0336
0.0624 0.0605 0.0554 0.0539 0.0512
0.0328 0.1020 0.0616 0.0315 0.0000
TABLE IV: Users’ SNRs and secure rate versus

Iii-C Joint optimization of source and jammer power

After the decision on utilization of jammer power based on Proposition III.1, all subcarriers can be categorized in two sets: - the ones that do not use jammer power (), and - the others that use jammer power (). Let and respectively denote the SNRs of the user , without jammer power and with jammer power over subcarrier , i.e.,


The joint power allocation problem can be stated as

subject to

where is the weight of the main user over the th subcarrier. and are the power budget constraints. Constraints and are imposed to tackle the SNR re-ordering challenge (cf. Section III-B2), while the boundary constraints for source power are captured in .

The source power has to be shared among the subcarriers’ set and , while the jammer power has to be allocated on the subcarriers of set only. We observe that source power is a coupling variable between the power allocation problem over the complementary sets and , and is the corresponding complicating constraint. Since the secure rate over a subcarrier is an increasing function of source power, the problem has to be solved at full source power budget . The optimization problem (III-C) can be solved using primal decomposition (PD) procedure by dividing it into one master problem (outer loop) and two subproblems (inner loop) [23]. The first subproblem is source power allocation over set , and second is joint source and jammer power allocation over set . The subproblem-1 can be stated as:

subject to

where , representing the coupling variable, is the source power budget allotted to subproblem-1. The source power budget to subproblem-2 is , as described in the following:

subject to

The two subproblems are solved independently for a fixed . Once the solutions of both the subproblems are obtained, the master problem is solved using subgradient method by updating the coupling variable as , where is an appropriate step size [25] and , are the Lagrange multipliers [23] corresponding to the source power constraints in subproblems (III-C) and (III-C), respectively. This procedure of updating is repeated until either the sum secure rate saturates or the iteration count exceeds a threshold.

Iii-C1 Solution of subproblem-1

Since each subcarrier is allocated to the best gain user (cf. Section III-A), . Thus, the objective function in (III-C) is a concave function of . The optimal obtained after solving the Karush-Kuhn-Tucker (KKT) conditions is given as


where is defined as follows:


, associated with , is found such that .

Iii-C2 Solution of subproblem-2

In order to solve the subproblem-2, we observe that the secure rate over the subcarriers of set is a concave function of source power for a fixed jammer power , and also a quasi-concave function of for a fixed as shown in Proposition III.1. This motivates us to use the method of alternating optimization (AO) [24] for joint power optimization, which alternates between source and jammer power optimizations. The AO method starts with , i.e., equal power over all the subcarriers of set , and for the known optimal allocation is done. For a fixed source power, the jammer power allocation problem can be stated as:

subject to

Since the secure rate over a subcarrier achieves a maxima at a unique jammer power (cf. Proposition III.1). So, first we evaluate optimal for achieving maximum secure rate, respecting the jammer power bounds as follows:


where is a small positive number. Note that, in secure rate improvement and , so first case does not arise. If the condition is satisfied, then optimum jammer power is allocated over all subcarriers, i.e., . Otherwise the problem (III-C2) is solved under jammer power budget constraint. The partial Lagrangian of the problem (III-C2) is given in (18).


After setting first-order derivative of the Lagrangian in (18) equal to zero, we obtain a fourth-order nonlinear equation in having following form


The coefficients of the equation (III-C2) are tabulated in Table V. In the domain , the secure rate is a concave increasing function (cf. Proposition III.1) and since , depending on , there exists a single positive real root of (III-C2). After obtaining for a fixed value of , is updated using the subgradient method [25]. For constrained jammer power budget, optimal jammer power is obtained after constraining between jammer power bounds as in (17). Thus, allocation for fixed can be written as:


For a known , the AO method now obtains based on (14) with the source power’s coefficient being , instead of . This completes one iteration of the AO method. This procedure of optimizing source and jammer powers alternatively continues until either the secure rate saturates or the iteration count exceeds a threshold. A PD and AO based Joint Power Allocation (JPA) scheme for weighted sum secure rate maximization is presented in Algorithm 1.

  • During power optimization in AO, if the jammer power over some of the subcarriers in is zero, then these subcarriers are taken out of and added back into for source power allocation in the next iteration of PD procedure.

= ; = ;
= ;
= ; = ;
= ; = ;
TABLE V: Coefficients of the non linear equation (III-C2)

Iii-C3 Convergence of joint power allocation

The source power allocation is done through PD with as the subgradient. The secure rate is a concave increasing function of source power in subproblem-1 and for a fixed jammer power, in subproblem-2. Thus, and are positive and bounded. Hence, is bounded and the method converges to an -optimal value in finite number of steps [25].

The authors in [26] have shown with the help of Bolzano-Weirerstrass theorem and [27] that, AO method converges for a problem which is concave in one set of variable and quasi-concave with unique maxima in another set. Hence, the joint power allocation based on PD and AO converges.

2:     Subcarrier Allocation:
3:     for every subcarrier  do
4:          find
5:     end for
6:     Subcarrier set creation:
7:     Init and
8:     for every subcarrier  do
9:          if  then
10:               compute and
12:          else
14:          end if
15:     end for
16:     Source and Jammer Power Allocation:
17:     primal decomposition begins
18:     Init