The fast development of wireless networks, which are intrinsically exposed to eavesdropping, has created a growing concern for confidentiality. While classical cryptographic schemes require a key provided by the end-user, physical-layer security leverages the randomness of the physical communication medium as a source of secrecy. The main benefit of physical-layer security techniques is their relatively low cost and their ability to combine with any existing security mechanisms. This dissertation provides an analysis including the theoretical study of the two-way wiretap channel to obtain a better insight into how to design coding mechanisms, practical tests with experimental systems, and the design of actual codes. From a theoretical standpoint, the study confirms the benefits of combining several multi-user coding techniques including cooperative jamming, coded cooperative jamming and secret key generation. For these different mechanisms, the trade-off between reliability, secrecy and communication rate is clarified under a stringent strong secrecy metric. Regarding the design of practical codes, spatially coupled LDPC codes, which were originally designed for reliability, are modified to develop a coded cooperative jamming code. Finally, a proof-of-principle practical wireless system is provided to show how to implement a secret key generation system on experimental programmable radios. This testbed is then used to assess the realistic performance of such systems in terms of reliability, secrecy and rate.

The onset of quantum computing threatens commonly used schemes for information secrecy across wireless communication channels, particularly key-based data-level encryption. This calls for secrecy schemes that can provide everlasting secrecy resistant to increased computational power of an adversary. One novel physical layer scheme proposes that an intended receiver capable of performing analog cancellation of a known key-based interference would hold a significant advantage in recovering small underlying messages versus an eavesdropper performing cancellation after analog-to-digital conversion. This advantage holds even in the event that an eavesdropper can recover and use the original key in their digital cancellation. Inspired by this scheme, a flexible software-defined radio receiver design capable of maintaining analog cancellation ratios consistently over 40 dB, reaching up to and over 50 dB, is implemented in this thesis. Maintaining this analog cancellation requires very precise time-frequency synchronization along with accurate modeling and simulation of the channel effects on the interference. The key sources of synchronization error preventing this test bed from achieving and maintaining perfect interference cancellation, sub-sample period timing errors and limited radio frequency stability, are explored for possible improvements. To further prove robustness of the implemented secrecy scheme, the testbed is shown to operate with both phase-shift keying and frequency-modulated waveforms. Differences in the synchronization algorithm used for the two waveforms are highlighted. Interference cancellation performance is measured for increasing interference bandwidth and shown to decrease with such. The implications this testbed has on security approaches based on intentional interference employed to confuse eavesdroppers is approached from the framework proposed in the motivating everlasting secrecy scheme. Using analog cancellation levels from the hardware testbed, it is calculated that secrecy rates up to 2.3 bits/symbol are gained by receivers (intended or not) performing interference cancellation in analog rather than on a digital signal processor. Inspired by the positive gains in secrecy over systems not performing analog cancellation prior to signal reception, a novel secrecy scheme that focuses on the advantage an analog canceller holds in receiver amplifier compression is proposed here. The adversary amplifier is assumed to perform linear cancellation after the interference has passed through their nonlinear amplifier. This is accomplished by deriving the distribution of the interference residual after undergoing an inverse tangent transfer function and perfect linear cancellation. Parameters of this scheme are fit for the radios and cancellation ratios observed in the testbed, resulting in a secrecy gain of 0.95 bits/symbol. The model shows that larger message powers can still be kept secure for the achieved levels of cancellation, thus providing an even greater secrecy gain with increased message transmission power.