Graduation Year

2026

Document Type

Dissertation

Degree

Ph.D.

Degree Name

Doctor of Philosophy (Ph.D.)

Degree Granting Department

Electrical Engineering

Major Professor

Gokhan Mumcu, Ph.D.

Co-Major Professor

Huseyin Arslan, Ph.D.

Committee Member

Stavros Vakalis, Ph.D.

Committee Member

Yasin Yilmaz, Ph.D.

Committee Member

Attila Yavuz, Ph.D.

Keywords

Directional Modulation, Eavesdropping, Phased Array, Spoofing

Abstract

Modern wireless networks support a wide range of applications including consumer devices, healthcare monitoring, industrial automation, smart grids, and military platforms, in which transmitted information is often highly sensitive and subject to stringent confidentiality and integrity requirements. As wireless communication operates over an open broadcast medium it exposes systems to physical layer (PHY) threats such as eavesdropping, in which an adversary passively intercepts transmitted information, and spoofing, in which a forged transmission misleads the receiver about device identity. Cryptographic defenses address both threats but require computational assumptions and key management infrastructure that may not scale to dense or resource-constrained deployments. PHY security offers a complementary approach by exploiting hardware and wireless channel to degrade an adversary's ability to intercept or impersonate, independently of computational bounds.

This dissertation investigates the use of chaotic antenna arrays (CAAs) for PHY security through RF fingerprint authentication and anti-eavesdropping. A CAA is a phased array in which each element is displaced from its nominal half-wavelength position by random three-dimensional offsets and assigned a randomized feed line length. These intentional perturbations produce direction-dependent phase errors that are unique to each manufactured device and are embedded in the radiated waveform.

This dissertation establishes an analytical foundation for the CAA RF fingerprint by deriving the far-field phase response from the array geometry, modeling its coupling to a multipath channel, and defining the fingerprint as the ordered sequence of per-element phases observed at a receiver. Uniqueness conditions are derived, showing that two arrays are indistinguishable only if all four structural parameters of every element match exactly. Fingerprint capacity grows with the number of array elements, the phase estimation resolution, and the number of observation directions. Three-dimensional element position randomization is shown to provide a multiplicative increase in fingerprint diversity and to introduce an irreducible bias in blind structural parameter estimation, preventing an adversary from recovering feed-line lengths regardless of how many phase observations are collected.

Practical fingerprint extraction is developed through a maximum-likelihood phase estimator paired with two equalization schemes that remove predictable channel contributions while preserving the device-specific signature. Both are validated on a software-defined radio (SDR) testbed with fabricated CAA hardware. Convolutional neural network (CNN) classifiers trained under line of sight (LOS) conditions exceed 99\% authentication accuracy on similar data and maintain above 87\% accuracy in non-line of sight (NLOS) environments with a single dominant reflected path. A cooperative authentication protocol combining fingerprint observations from multiple spatially separated receivers rejects both replay and spoofing attacks with probability approaching unity as the number of receivers increases, exploiting the fact that an attacker can simultaneously reproduce the fingerprint seen from at most one direction.

Two anti-eavesdropping methods are then developed for use in wireless communication scenarios once device authentication is completed. In the point-to-point wireless communication scheme, phase distortion based on the CAA's array factor (AF) is digitally embedded symbol-by-symbol into the transmitted signal. A legitimate receiver with knowledge of the CAA parameters acquires the AF for the particular direction through a brief pilot exchange, uses it to remove the deliberately added distortion, and demodulates correctly. In contrast, an eavesdropper without CAA knowledge cannot remove the distortion to demodulate the symbols, even when located in the same direction as the legitimate receiver. The second anti-eavesdropping scheme uses joint directional modulation (DM) and encoding for distributed receiver systems, where a CAA-equipped transmitter applies DM together with inter-access point (AP) phase difference encoding based on the perceived phase difference observed at spatially distributed APs. The phase shifters are frequently varied to produce deliberate pattern distortion, and each symbol is additionally encoded according to the sign of the instantaneous phase difference between two distributed APs. Correct decoding requires joint processing of both AP signals, which no single-location eavesdropper can achieve. The second scheme incurs marginal performance loss compared to conventional communication schemes, while providing PHY security against eavesdropping.

Download

Share

COinS