Towards intelligent spectrum awareness in wireless networks

In the 6G era, efficient spectrum utilization is crucial due to the increasing number of wireless devices. Future networks must handle extreme spectrum congestion, necessitating novel strategies beyond 5G. Integrated sensing and communication (ISAC) systems will enable simultaneous data transmission and environmental sensing, supporting applications like traffic monitoring and urban safety. However, ISAC introduces security risks, particularly radar-related threats such as deceptive jamming, which can compromise sensing functions. Artificial intelligence (AI) will play a key role in network orchestration and spectrum management, but it also introduces new vulnerabilities, including signal jamming and denial-of-service attacks. AI-enabled adversaries can dynamically adapt their attacks, making anomaly detection more challenging. To secure 6G networks, robust spectrum monitoring is essential to detect and mitigate these threats. This thesis presents a novel spectrum patrolling framework that secures wireless networks by leveraging distributed sensors to monitor the radio-frequency (RF) environment. Advanced signal processing techniques extract meaningful analytics to detect and counteract jamming and unauthorized spectrum usage. A blind source separation (BSS) technique isolates malicious signals, while causal inference using transfer entropy (TE) identifies jamming attacks. In an ISAC context, a variational autoencoder (VAE)-based approach enhances jamming detection. Furthermore, cooperative wideband spectrum sensing (WSS) is explored using factor analysis and variational inference to assess spectrum occupancy, estimate user count, and measure noise power. A statistical meta-analysis technique is introduced to improve the synthesis of independent tests.