Research On MISO Spatial Scrambling Techniques

Spatial scrambling is a set of physical layer techniques to realize secure messagestransmission. In order to jam the unknown eavesdroppers, spatial scrambling takes theadvantages of multi-antenna technique to send artificial noise. Due to the rapid variation ofartificial noise, it’s difficult for eavesdroppers to decode messages. As a fundamental physicallayer security technique, it’s necessary and important to study how to suppress artificialinterference and, on the other side, how to enhance the secrecy of scrambling signals. At present,there are mainly three issues concerning spatial scrambling:1) The lack of criterion for weighingsecrecy. The widely used criterion BER is not suitable to measure secrecy;2) The suppression ofinterference. The existing method is susceptible to channe l noise and has an impracticablearithmetic complexity;3) The enhancement of scrambling signals. The existing spatialscrambling signals will lose its efficiency when eavesdropper is equipped with enough antennas.Relying on the National Natural Science Foundation, this article takes these issues from twoopposite perspectives: interference suppression and secrecy enhancement.First, three low-complexity interference suppression algorithms, deriving from thecharacteristics of signal space and stochastic density distribution respectively, are proposed tointercept scrambling signals. Second, in order to enhance the secrecy of scrambling signals, aSpatial Symmetric Scramble (SSS) based scheme is proposed under the criterion ofunidentifiable mixed density. The chief contents of this article are as follows:1. A Principal Projection (PP) algorithm is proposed to intercept scrambling messages. First,it’s proved that a principal signal direction exists when the eavesdropper is equipped with nofewer antennas than the sender. Second, calculate the principal direction from training samplesby the Fisher linear discriminant criterion. Projection along the principal direction wouldsuppress the artificial interference while maintaining the information components. I n this way,scrambling signals can be intercepted. Theoretical analysis and simulations show that the PPalgorithm has the benefits of lower complexity, and better anti-noise performance, comparedwith the existing MUSIC-like algorithms. The tolerance of training samples has also beenverified.2. In order to get rid of training samples from PP, a signal signature based HyperplaneClustering (HC) algorithm is proposed. It is proved that the receiving scrambling signals aredistributed within parallel hyperplanes when the eavesdropper has equal or more antennas thanthe transmitter. The HC algorithm uses parallel hyperplanes in signal space to cluster receivingsignal points, extract signal features and thus to intercept messages. Simulations show that theHC algorithm has the benefits of lower complexity, and better anti-noise performance, comparedwith the existing MUSIC-like algorithms.3. From the characteristic of probability density, a Maximum-Likelihood Estimation (MLE)is proposed. The scrambling signals are regarded as stochastic variables, the distributionparameters of which are determined by both artificial interference and channel noise. Themaximum-likelihood criterion is used to calculate these parameters. Since the solutions satisfy a set of non-linear equations, a Gradient Ascent (GA) method is adopted to solve them. Then,based on maximum a posteriori (MAP), scrambling signals can be decoded at a minimum errorprobability when these parameters are estimated accurately. Simulations show that MLE has abetter interference suppression performance, compared with the other three algorithms.4. A Spatial Symmetric Scrambling (SSS) scheme is proposed to enhance the secrecy ofspatial scrambling. Unlike the traditional way where the white Gauss noise (WGN) is transmittedin the null space of main channel, SSS sends signal-like noise instead. Different from thehyperplane distribution, SSS signals shape like a symmetric pattern with many symmetrical axis.It’s proved that the four existing algorithms cannot intercept SSS signals because theunidentifiable mixed probability density. An adaptive power allocation scheme betweensignal-like noises is proposed to further increase the secrecy. Simulation verifies the conclusion.