PublisherThe University of Arizona.
RightsCopyright © is held by the author. Digital access to this material is made possible by the University Libraries, University of Arizona. Further transmission, reproduction, presentation (such as public display or performance) of protected items is prohibited except with permission of the author.
EmbargoRelease after 12/08/2024
AbstractIncreasing the performance of a security system to reduce risk is an important task in many applications. In cybersecurity, physical unclonable functions (PUFs), with their unique identity and random property, are excellent candidates to replace the vulnerable pseudo-random signatures. Optical PUFs, consisting of non-integrated circuit with high complexity, provide non-replicable strong signatures for the authentication method. Here, we analyze a novel and low cost architecture of an optical PUF using nano-dendrite to demonstrate a potentially secure and stable system for encryption applications. In this work, we utilize two different optical simulation methods to analyze the speckle pattern at micro-scale and macro-scale at normal incidence and validate simulation results with experiments. We also systematically explore the sensitivity of our nano-dendrite optical PUF by selecting the thresholds of fractional Hamming distance (FHD) values for different challenges. The results show the thresholds we set for FHD is 0.1 in wavelength challenge, 0.075 in polar-angle challenge and 0.018 in polarization challenge.Aviation security is another area in security with broad implications. The X-ray transmission machines in airports utilize the dual-energy transmission system to detect non-allowable matters. However, some materials cannot be reliably detected with such methods. Therefore, developing a more accurate X-ray transmission based system for the material classification has become an urgent concern. In this work, we review the Alvarez-Macovski coefficients for the new feature space and estimate them by multi-energy X-ray configuration. The mean and variance of the estimated coefficients are the inputs of the supervised learning models for classification. We inspect the performance of 4 energy bins system is better than dual-energy bins system using neural network classifier, indicating only two energy parameters are not enough for high-fidelity material characterization. The results show the true positive rate (TPR) value in receiver operating characteristic (ROC) curves from dual-energy bins to 4 energy bins is increased from 87.46% to 94.94% in higher source flux images and increased from 87.35% to 89.44% in lower source flux images when the false positive rate (FPR) value is 5%.
Degree ProgramGraduate College