Advanced Physical Layer Security Techniques for Non-Terrestrial Communications

The broad use of wireless communications makes it necessary to investigate specifically targeted security solutions. Moreover, differently from wired communications, wireless communications are by nature vulnerable to many threats: of course, since the medium is easily accessible, any malicious user can disrupt the communication by using a jamming attack, intercept the signal to disclose its content or information about the transmitter or lead a spoofing attack by generating a counterfeit signal or by tampering the transmitted signal. In many applications, it is not possible to rely on cryptography: for instance, cryptography-based solutions have a considerable computational cost, thus they may not be suited for many wireless applications, where we want to reduce the user energy consumption, such as in a wireless sensor network (WSN) used in an Internet of things (IoT) context. Thus, we resort to physical-layer security (PLS) approaches. Physical layer authentication relies on the collection of the observation about the channel characteristics (e.g., features of the channel impulse response) to tell apart transmissions by legitimate network members from the ones by impersonating attacker. Moreover, PLS mechanisms are also unconditionally secure, since the security is not provided by a computationally hard problem. On the other hand, since theses techniques rely on the channel model, it may be complex to generalize solutions and each context need to be separately analyzed. This Thesis focuses on the development of physical layer authentication for global navigation satellite systems (GNSSs) and underwater acoustic networks (UWANs). GNSS services are used to provide positioning and timing. However, these services do not (necessary) rely on the data content of GNSS but on the properties of the signals themselves, i.e., phase and Doppler frequency. Indeed PLS can be used to provide authentication at signal level by making the spreading code (or part of it) unpredictable. The contributions of this Thesis are multiple. We propose a novel network-aided authentication protocol, proposing also a verification based on the generalized likelihood ratio test (GLRT). To show its robustness, the scheme is tested against several attacks: among others, we also consider the security code estimation and replay (SCER) attack and the internal code attack. Next, we focus on the problem of position, velocity, and time (PVT) assurance, where we propose a series of consistency checks to enlarge the set of trusted signals to be used for the PVT. We focus then on the problem of providing an authenticated but robust timing service, relying only on Galileo’s commercial authentication service (CAS). Finally we address the problem of message scheduling in GNSS: considering, for instance, an authentication service that need to disseminate a digital signature over the GNSS channels, we study both single and multi-round scheduling solutions that aim at minimizing the maximum and the average latency. In the last part of the Thesis, we tackle the problem of physical layer authentication for UWANs: underwater acoustic channels (UWACs) are known to decorrelate easily in space, and to have a limited time coherence, thus by extracting relevant channel features, it is possible to distinguish a packet transmitted by a legitimate transmitter from the one sent by a potential attacker. Indeed by having multiple (trusted) cooperating sensors, it is possible to improve the classification procedure. We address this problem by using machine learning (ML) techniques. We will investigate multiple aspects: for instance, for training, each receiver may have at disposal observations only from the legitimate or from both legitimate and attacker channels; the influence of the amount of the information shared by each user; strategies to deal with mobility and time-varying channels.