TITLE: Secret-Key Agreement and Authentication Under Degraded Channel Reciprocity

ABSTRACT: The widespread deployment of low-cost wireless devices and Internet-of-Things (IoT) devices has intensified the need for security mechanisms that can reliably support cryptographic key establishment under practical hardware constraints and the condition at the environment of deployment. While conventional key agreement protocols provide strong security guarantees, their effectiveness in resource-constrained wireless systems depends on the availability of robust key management infrastructure and reliable shared secrets. Physical-layer secret-key generation (SKG) techniques that exploit properties of the wireless channel, such as reciprocity and temporal variation, offer an additional source of entropy and shared randomness that can assist cryptographic key agreement, but their practical use is limited by imperfect channel reciprocity on low-cost devices, the channel sampling technique, and the severe channel condition. This dissertation develops an end-to-end framework that integrates wireless channel randomness as a supporting component of cryptographic key agreement and authentication for low-cost wireless devices operating under imperfect reciprocity. The central thesis of this dissertation is that, under realistic hardware and channel constraints, reliable key agreement must move beyond instantaneous channel measurements and instead exploit statistical reciprocity, novel signal representations and randomness extraction technique assisted by cryptographic locking mechanisms to tolerate residual mismatch in channel-derived keys and achieve reliability and security in realistic environments. The proposed end-to-end framework achieves this by integrating statistical modeling with cryptographic locking mechanisms that transform imperfectly reciprocal channel observations into consistent cryptographic keys and increase the key generation rate compared to the latest SKG approaches. The dissertation begins with an extensive experimental assessment of channel reciprocity using real Wi-Fi channel state information (CSI) collected from wireless devices. This study identifies the dominant practical factors that degrade reciprocity and evaluates multiple metrics for assessing reciprocity quality. The results demonstrate that wavelet coherence–based time–frequency metrics provide a more informative characterization of reciprocity than conventional correlation-based measures. The CSI datasets collected in this work are released to support reproducible research in practical wireless security. From these experimental findings, a wavelet-based CSI reconstruction and key generation framework is proposed to enhance the consistency of CSI . By leveraging time–frequency alignment, wavelet coherence, and synchronization, the proposed approach produces CSI representations that are more stable across communicating devices, leading to higher key generation rate under diverse propagation conditions compared to other quantization-based techniques that relies on the temporal analysis of the channel features. Recognizing that reciprocity enhancement on the channel-sample level alone cannot fully eliminate mismatch on low-cost devices, the dissertation next introduces a statistical key generation paradigm based on wavelet scattering embeddings, hidden Markov models (HMMs), and digital locks. By modeling channel behavior at the distributional level, this approach avoids reliance on instantaneous measurements and conventional quantization. A sample-then-lock reconciliation technique with high quality bits (HQB) sampling is developed to transform statistically correlated channel features into cryptographically consistent keys, while accounting for helper-data leakage and possible statistically informed eavesdroppers. Extensive experimental evaluations across indoor and outdoor scenarios demonstrate that the proposed framework enables reliable and secure key generation and authentication despite imperfect channel reciprocity and outperforms quantization-based techniques as well the latest quantization-free techniques in terms of the key generation rate, shared keys randomness, and the required communication overhead. Finally, the dissertation extends the use of CSI beyond key generation to assist in devices authentication. A CSI-assisted authentication technique is proposed to detect replay attacks by exploiting temporal variation and time shifts in channel measurements. Overall, this dissertation advances wireless security by positioning physical-layer secret key generation as a practical complement to cryptographic mechanisms strengthening key establishment and authentication in IoT networks without replacing established cryptographic foundations.

MAJOR ADVISOR: Bechir Hamdaoui
COMMITTEE: Weng-Keen Wong
COMMITTEE: Lizhong Chen
COMMITTEE: Eduardo Cotilla-Sanchez
COMMITTEE: Bechir Hamdaoui
GCR: Karl Haapala