Ambient backscatter communication (AmBC) is a revolutionary technology in the Internet of Things (IoT) due to its spectrum-sharing capability and ultra-low energy

consumption. The multiple-antenna tags of AmBC using spatial modulation (SM) can enhance spectrum efficiency and communication reliability. However, due to the

inherent broadcast nature of wireless communications in IoT and the use of simple modulation and coding schemes, AmBC systems are vulnerable to security threats

like jamming and eavesdropping attacks. Therefore, how to protect AmBC from such attacks becomes a worthy concern. The current research mainly considers unilaterally enhancing the security of AmBC through physical layer security (PLS) methods such as artificial noise (AN) or unilaterally improving the efficiency and reliability of AmBC through advanced modulation techniques such as SM. Enhancing the reliability and efficiency of AmBC while ensuring its security has not been investigated.
 

In this thesis, we propose an adaptive PLS enhancement (APSE) framework for generalized SM (GSM) based AmBC systems to guarantee the security and reliability of communication  dequately. Specifically, we first consider the security enhancement for a simplified version of SM, i.e., space shift keying (SSK), and propose to apply AN and antenna selection techniques to SSK-based IoT systems for enhancing PLS. Second, we propose a machine learning-based blind signal detection method for AmBC systems. We use the variational Bayesian inference approach of the Gaussian mixture model to obtain the real constellation information and use an improved expectation maximization algorithm in machine learning to recover the signals. The proposed scheme can effectively reduce the bit error rate of signal detection in AmBC systems and ensure communication reliability. Finally, we propose the APSE framework with GSM scheme for AmBC systems in IoT. The experimental results show that the proposed method can effectively improve the security and reliability of communication and outperform the state-of-the-art methods.