Wireless physical layer security is to exploit physical characteristics of the channels between the transmitter and the legitimate user and the eavesdropper to achieve secure communications, such as the randomness, difference, reciprocity, and so on. It guarantees the messages sent by the sources cannot be decoded by the eavesdroppers from an information-theoretic point of view and has broad application prospects on anti-eavesdropping, low-intercepting, and wireless communication system design. Through exploiting the spatial degrees of freedom, multi-antenna techniques can not only provide both reliability and efficiency gains, but also inherently enlarge the difference between the main channel and the eavesdropping channel, providing a wide space for wireless physical layer security research. However, the secrecy performance of multi-antenna systems is critically dependent on the amount of channel state information(CSI) at the transmitter. Especially, secure transmit designs under different CSI assumptions and different system models are the key and challenging problems for physical layer security research. Thus, this thesis mainly focuses on these key and challenging problems.Firstly, considering the case of statistical information on the eavesdropping channel available at the transmitter for a multiple-input single-output(MISO) wiretap channel, the thesis studies the power allocation between the information-bearing signal and artificial noise(AN) for AN-aided scheme and proposes an optimal power allocation algorithm to minimize the secrecy outage probability. The existing results show that the AN should be generated on the orthogonal space of the main channel and the secrecy performance depends highly on how well the power is allocated between the information-bearing signal and AN. The thesis derives the closed-form secrecy outage probability and proposes the optimal power allocation between the information-bearing signal and AN to minimize the secrecy outage probability for the cases of single multi-antenna eavesdropper and multiple single-antenna eavesdroppers, respectively. For the single multi-antenna eavesdropper case, a Golden search-based optimal power allocation algorithm is investigated. Moreover, for the multiple single-antenna eavesdroppers case, a closed-form optimal power allocation algorithm is proposed. Through some derivations, the optimal power allocation algorithm can be transformed into solving a deterministic equation, thus resulting in a closed-form solution. Simulation results for both the aforementioned scenarios show that the secrecy performance of proposed optimal power allocation outperforms naive beamforming and equal power allocation.Secondly, for the case where no eavesdropping channel information is available at the transmitter in the multiple-input multiple-output multi-antenna eavesdropper(MIMOME) model, the thesis proposes a joint transmit antenna selection(TAS) and maximal ratio combining(MRC) transmission algorithm. In the TAS-MRC algorithm, the transmitter has to obtain the optimal transmit antenna index fed back by the legitimate receiver. With the perfect feedback, the achievable diversity order is the same as that of conventional multi-antenna system without secrecy consideration, which is independent of the number of the eavesdropperâ€™s antennas. In practical systems, because of the effects from channel feedback delay and channel estimation errors, perfect feedback is almost impossible. Further, the thesis studies the effects of imperfect feedback on the secrecy performance, including time-delayed feedback, erroneous feedback, and both time-delayed and erroneous feedbacks. Furthermore, the probability of non-zero secrecy rate and secrecy outage probability are presented in closed form. The derived asymptotic secrecy outage probability at high signal-to-noise ratio(SNR) reveals that imperfect feedbacks significantly degrade the secrecy performance. When the feedbacks are outdated and/or erroneous, the diversity gain from the TAS disappears and only the MRC diversity at the receiver remains. The analytical results are more general and encompass the existing results for the perfect feedback case. Simulation results verify the analytical results.Thirdly, assuming that no eavesdropping channel information is available at the transmitter, the thesis proposes an AN-aided transmission algorithm for a MISO cognitive radio network(CRN) in the presence of multiple passive eavesdroppers. The secondary user(SU) and the primary users(PUs) operate in underlay mode. Utilizing the signal-to-interference-plus-noise ratio(SINR) as the quality-of service(Qo S) metric, the goal is to maximize the power of AN available to confuse the eavesdroppers, while maintaining a predefined SINR at the receiver of the SU and keeping the interference to the PUs. With perfect CSIs for the SU and PUs at the SU transmitter, it is proved that the optimal transmission scheme for the information-bearing signal is beamforming, i.e., the optimal input covariance matrix is rank-one. However, when the CSIs are imperfect, the AN designed for the perfect CSI case will cause noise leakage problem, significantly degrading the legitimate receiverâ€™s reception. Furthermore, the thesis assumes that both CSIs for the SU and PUs are imperfect at the SU transmitter, and the errors are modeled as channel vector uncertainty model, channel covariance uncertainty model, and stochastic channel uncertainty model, respectively. Under the channel vector uncertainty model and channel covariance uncertainty model, the thesis proposes a worst-case AN-aided transmission algorithm, where the optimization problem is a non-deterministic polynomial hard(NP-hard) one. Using S-Procedure lemma and convex optimization theory, the intractable SINR and interference constraints are derived in the equivalent forms and then, the original problem can be converted into a semi-definite program(SDP) one, thus resulting in an optimal solution for the AN covariance. After that, an outage-constrained robust formulation is proposed for the stochastic channel uncertainty model. By some manipulations and simplifications, the original optimization problem is transformed into a chance-constrained non-convex one. With the aid of two kinds of Bernstein-type inequalities, the complicated probabilistic constraints can be converted into deterministic forms, thus resulting in an approximation solution for the AN covariance. Simulation results show that proposed robust AN-aided transmission algorithms under three aforementioned channel uncertainty models can efficiently reduce the sensitivity to CSI errors.Finally, for the case where perfect channel information is available at the transmitter, the thesis proposes an optimal relay beamforming algorithm for the one-way relay eavesdropping system, a relay chatting based transmission algorithm for the two-way eavesdropping system, and two kinds of destination aided cooperative jamming algorithms for the multiple-input multiple-output(MIMO) untrusted relay eavesdropping system, respectively. In the one-way relay system in the presence of a single-antenna eavesdropper, there exist no tractable algorithms to the optimal relay beamforming design for maximizing the secrecy rate using amplify-and-forward(AF) scheme in the literature. Thus, the thesis firstly proposes a relay beamforming design based on branch-and-bound algorithm and proves it can obtain the globally optimal solution. Considering a high computational load for the optimal solution, a suboptimal beamforming algorithm based on general power iteration(GPI) is further proposed. The suboptimal algorithm only calculates three matrices in each iteration, efficiently reducing the complexity. In the two-way relay eavesdropping system, as the transmit power increases, the secrecy outage probability of the existing joint relay and jammer selection algorithm approaches to a constant. To solve this problem, a new relay chatting(RC) based transmission algorithm is proposed, which combines both advantages from opportunistic relay selection and cooperative jamming. Through distributed beamforming, it can eliminate interferences to the legitimate receivers and only confuse the eavesdropper. Theoretic analysis and simulation results show that at high SNR, the secrecy outage probability of proposed relay chatting algorithm converges to zero. Moreover, the relay may be untrusted in practical systems(i.e., the relay tries to interpret the messages sent by the sources) and thus, the MIMO relay system without direct link can achieve no secure communications. To circumvent that, the thesis proposes two kinds of destination aided cooperative jamming transmission algorithms, i.e., joint source, relay, and destination precoding and TAS at untrusted relay. For the joint source, relay, and destination precoding algorithm, the thesis proposes an alternating iterative algorithm to jointly design the source, relay, and destination precoding matrices for maximizing the secrecy rate. Especially, the optimal source beamforming vector and destination precoding matrix can be obtained by solving a convex problem, respectively, and optimal relay precoding matrix is presented in closed form. For the TAS at untrusted relay algorithm, several antenna selection schemes are investigated for different scenarios, including optimal scheme(Optimal), untrusted relayâ€™s SINR maximization scheme(Suboptimal I), and untrusted relayâ€™s SINR minimization scheme(Suboptimal II). Simulation results show that proposed joint source, relay, and destination precoding algorithm and TAS at untrusted relay algorithm can greatly enhance secrecy. |