Remote attestation guarantees secure computing environments by authenticating the integrity of software running remotely. Among various solutions, methods based on hardware-software co-design (hybrid) are particularly popular due to their low overhead yet effective approaches. In this project, we implement a new hybrid attestation framework on NEORV32 RISC-V processors. NEORV32 is an open-source, customizable RISC-V processor. Users can personalize the core according to their specific needs by enabling/disabling relevant features. In our threat model, we assume the adversary gains complete device control, modifying program and configuration data maliciously. Our solution is to utilize a software-hardware approach employing hash functions to identify these attacks. We present the implementation of the hybrid attestation on the NEORV32 RISC-V Processor, integrating the hash function hardware module, and ensuring its tampering resistance. The proposed hybrid attestation mechanism strengthens NEORV32-based device security and presents a valuable learning opportunity for developers and researchers interested in embedded system security. The code and data from this work are open source.

Infrared Light Sheet Microscopy has emerged as a powerful imaging technique for researching biological specimens with enhanced penetration depth. In traditional visible light sheet microscopy, imaging depth is limited due to significant light scattering. To address this limitation, we developed a near-infrared open-top light sheet microscopy, utilizing approximately 800 nm excitation, enabling non-invasive, three-dimensional imaging of biological tissues. To assess the system’s capabilities, we imaged various samples of fluorescent beads, zinc agarose, and cross sections from an adult mouse heart. The current microscope system uses a light sheet controlled by a galvanometer to capture multiple images at different focal planes along the z-axis. The current working set-up allows for data collection with an exposure time of 200 ms and within a 400 x 400 μm field of view. Following image acquisition, MATLAB was used to crop the image stack to highlight areas of interest within the samples. Once aligned, the images are stacked into a three-dimensional reconstruction in ImageJ for sample visualization and analysis. Future areas of improvement include decreasing exposure time and maximizing sample field of view to capture morphological nuances in greater detail. Successful integration of light sheet and infrared microscopy in the system will eventually allow for high resolution, 3D imaging of dynamic tissues to research mechanisms of congenital heart disease.

Eye tracking is a sensor technology that uses one or more cameras, infrared light sources, and computing capabilities to measure the movement of a person's eyes for applications such as: gaze estimation, disease diagnosis (ADHD,OCD, ASD, Parkinsons), eye health research, and surgical robotics. Particularly for gaze estimation, the performance of eye tracking critically depends on how well the test data distribution is covered by the calibration data. In practice recording calibration data that covers the full distribution is challenging and has been addressed by rendering photo-realistic images of the eye through the use of eye models generated using computer graphics techniques. However, the use of simplified eye models is one major limitation of previous work. We hypothesize that eye tracking is a function of eye anatomy and that using a more anatomically correct and more anatomically diverse eye model result in improved eye tracking performance. An eye model that can be varied in eyeball size, corneal shape, corneal thickness, and iris size was created in Blender 3.6. Combined with a dense 3D head scan of 10 subjects, images of this eye model are rendered from different perspectives under different lighting conditions. We successfully modeled the refractive caustics of the cornea, created geometrically accurate sclera and cornea, and automated scene setup and eye model creation using scripting.We trained a learning-based gaze estimator on the rendered images of our model and evaluated the change in performance. In the future, we would like to iterate on our current eye model design to make