35967 research outputs found
Sort by
LIBIHT: A Hardware-Based Approach to Efficient and Evasion-Resistant Dynamic Binary Analysis
No full textSenior Honors Thesis, Department of Computer Sciences, University of Wisconsin-MadisonDynamic program analysis is invaluable for malware detection, debugging, and performance profiling. However, software-based instrumentation incurs high overhead and can be evaded by anti-analysis techniques. In this paper, we propose LIBIHT, a hardware-assisted tracing framework that leverages on-CPU branch tracing features (Intel Last Branch Record and Branch Trace Store) to efficiently capture program control-flow with minimal performance impact. Our approach reconstructs control-flow graphs (CFGs) by collecting hardware generated branch execution data in the kernel, preserving program behavior against evasive malware. We implement LIBIHT as an OS kernel module and user-space library, and evaluate it on both benign benchmark programs and adversarial anti- instrumentation samples. Our results indicate that LIBIHT reduces runtime overhead by over 150× compared to Intel Pin (7× vs 1,053× slowdowns), while achieving high fidelity in CFG reconstruction (capturing over 99% of execution basic blocks and edges). Although this hardware-assisted approach sacrifices the richer semantic detail available from full software instrumentation by capturing only branch addresses, this trade- off is acceptable for many applications where performance and low detectability are paramount. Our findings show that hardware-based tracing substantially improves performance while reducing detection risk and enabling dynamic analysis with minimal interference
In The Shadow of Giants: Persistent Revitalization Principles in the Old Northwest, 1813-1832
No full textSenior Thesi
Unveiling Bias in Multimodal Models
No full textSenior Honors Thesis, Department of Computer Sciences, University of Wisconsin-MadisonVision Language Models (VLMs) have significantly advanced multimodal understanding by effectively combining visual and textual modalities for various applications, including image captioning, visual question answering, and video summarization. However, despite their capabilities, these models exhibit pronounced modality biases, predominantly relying on textual inputs over visual data. This thesis systematically evaluates unimodal biases in state-of-the-art VLMs, highlighting the impacts on performance and proposing innovative strategies for bias mitigation, including Prompting, Interleaved Vision-Text Projection (IVTP), and Cross-Attention Projection. Our experimental evaluations using the MMWorld dataset demonstrate that targeted mitigation strategies substantially enhance modality balance and model robustness. The findings underscore the importance of architectural adjustments and training methodologies to ensure equitable multimodal integration, paving the way for more reliable and robust multimodal AI systems
Exploring the dynamic radio sky using all-sky radio surveys
No full textThe timescales over which radio emission from astrophysical sources varies span from ~nanoseconds (in pulsars) to years (e.g., in active galactic nuclei and afterglows of extragalactic transients). However, not all these emission timescales have been uniformly explored to discover new transients (and/or sources) at radio wavelengths. All sky-surveys at other wavelengths (X-ray/ultraviolet/ optical/infrared) in the past two decades have demonstrated that searches targeting different emission timescales (and different luminosities) have resulted in the discovery of new transients. Such new discoveries will amount to probing new emission mechanisms in sources that were previously unexplored. At radio wavelengths (~millimeters to meters), such multi-epoch large-scale radio surveys required to independently discover radio transients were missing due to the narrow field of view of interferometers until recently. This meant that most transient studies at radio wavelengths were limited to targeted follow-up observations. Hence, population properties of different radio transients can be biased since it is impractical to follow up every discovery. Still, more importantly, we are only probing the subpopulation of sources that are bright at shorter wavelengths. It is with the advent of wide-field interferometers in recent years that such surveys have begun. Such surveys have also started to explore the previously unexplored minute-to-day varying radio emission, and this has resulted in the discovery of new classes of radio transients. With the availability of this high cadence data, it is now possible to perform unbiased searches for radio transients to understand the nature of newly discovered radio transients. In this work, using one such large survey, I will describe how such unbiased searches in known classes of radio transients can uncover new phenomena that were previously unexplored. In addition, I will also describe our efforts in discovering more minute--day timescale radio transients and use this sample to comment on the nature of these newly discovered classes of radio transients
Words & Science Matter: Vetting Algae Management Claims
No full textWords & Science Matter: Vetting Algae Management Claim
A Watershed Approach to Reducing Runoff Entering the Waters of Lake Wissota
No full textA Watershed Approach to Reducing Runoff Entering the Waters of Lake Wissot
Generation of CRISPR-Mediated EPAC2 Expression Reporter Melanoma Cell Line for High Throughput Screening for Anti-Melanoma Agents
No full textSenior Honors Thesis, Department of Molecular & Cell Biology, University of Wisconsin-MadisonThis thesis investigates the regulatory role of EPAC2, a cAMP-responsive guanine nucleotide exchange factor, in melanoma cells and explores its potential as a target for therapeutic intervention. Using CRISPR/Cas9 genome editing, HiBiT-tagged EPAC2 constructs were engineered to enable real-time fluorescence-based quantification of endogenous protein levels. These modified cells will then be exposed to a library of natural compounds to identify agents that inhibit EPAC2 expression. The study incorporates molecular cloning, transfection, and antibiotic selection to establish a robust platform for compound screening. This work lays the foundation for further exploration of EPAC2 as a pharmacological target and introduces a versatile tool for functional screening of signaling proteins in cancer cells
THE EFFECTS OF ACTIVE SITE MUTATIONS ON THE CHEMICAL AND PHYSICAL PROPERTIES OF THE ENZYME CYTOCHROME C NITRITE REDUCTASE
No full textCytochrome c nitrite reductases (ccNiRs) catalyze reduction of nitrite to ammonium in a six-electron, eight proton process. This work describes the preparation and preliminary characterization of three active site variants, R103Q, H257Q, and Y206F, from the Shewanella oneidensis ccNiR homologue. The mutated residues are strictly conserved across ccNiR homologues, and the long-term goal of their study is to shed light on their catalytic function. Preliminary characterization of the variants’ nitrite and hydroxylamine reductase activities in comparison to those of ccNiRWT using a strong reducing agent will be presented. By using electrochemically applied potentials, or weak reductants as electron sources, it is possible to accumulate partially reduced enzymatic intermediates, which can then be characterized using various spectroscopic techniques. In one study, UV-vis spectropotentiometric titrations of the Y206F ccNiR variant were performed, first in the presence of nitrite, and second in the presence of cyanide. The variant’s electrochemical behavior was then compared to that of the wild-type enzyme. In a second study, the reactivity of Y206F ccNiR with the weak reductant hexaammineruthenium (II), in the presence of nitrite, was investigated using steady-state and stopped-flow UV-visible spectroscopy. This reactivity was compared with that of the wild-type and the R103Q variant ccNiRs. Finally, an initial EPR experiment using the Y206F variant is described
DETERMINISTIC METHODS ANALYSIS OF A GAS-COOLED INTERMEDIATE-FAST SPECRUM REACTOR
No full textIn this work a gas-cooled fast reactor (GFR) with a long fuel cycle being developed by General Atomics (GA) was modeled using the Argonne National Lab (ANL) code system PyARC which
couples together deterministic solutions from the Argonne Reactor Computation (ARC) codes with Monte-Carlo solutions using OpenMC. PyARC allows for a unified geometrically approximated input which facilitates neutronic code to code benchmarking and comparison. The deterministic solvers have great success when modeling sodium-cooled fast reactors (SFR), and it is standard practice to alter calculational parameters to allow the deterministic solution to converge on the Monte- Carlo one, which hastens neutronic calculations to facilitate efficient core design. In this analysis extending the efficacy of this modeling method to the GFR core will be examined.
The mass by region breakdown for the core geometry built in PyARC was compared to reported mass inventories from a GA report successfully. Multi-group cross sections for the core mesh are generated using 2-3, and the size and energy groups of the mesh were varied to ensure the softened energy spectrum of the GFR is properly resolved. TWODANT is an optional higher order transport solver for core flux calculations that 2-3 solves in first order. Furthermore, due the presence of a SiC guide tube in the center of the fuel assembly, the mesh was altered further with decreased mesh separation to 5mm in the fuel regions of the core to resolve this structure. Using TWODANT, detailed fuel, and increasing the energy group structure to 70 from the default 33 used in SFR calculations provided small corrections toward OpenMC for the power profile.
For the core flux calculation, VARIANT was implemented to provide a higher order transport solution for the core within the mesh generated. After these efforts to converge on the OpenMC
calculation, had discrepancies near 500 pcm in steady state. The power profile was matched more accurately, with an average root mean squared error (RMSE) of 0.6%, with a maximum
difference of 1.8% found in the central six fuel assemblies. The GA report calculates power peaking on a pin level which is a limitation of this analysis, as here assembly powers are calculated instead.
After settings in steady state were investigated, the analysis moved to depletion where the
equilibrium fuel cycle was converged on. This process began with a fully fueled core, and by Cycle 3 of a standard IN-OUT refueling scheme the cycle was close to the target GA cycle. From this point, Zones 1 and 3 were de-enriched to mimic once and twice burned fuel to good effect in Cycle 1 and
3. Cycle 2 proved challenging due to reduced reactivity at BOC 2 regardless of initial enrichment zoning. The solution proposed here was to lengthen Cycle 1 such that it was refueled when =
1.007 and to refuel every cycle from that point whenever it reached this target multiplication factor.
By using 19.25% uranium-235 enrichment in Zone 1 and 16.75% in Zone 3, the final average cycle length was just over 15 years.
The last proposed solution to increase Cycle 2 reactivity and average cycle length was reshuffling Zone 1 and Zone 3 assemblies to reduce initial reactivity to allow for increased initial fissile inventory. This failed to be an effective strategy for either issues, but did reduce assembly power peaking factors by around 3%. Overall, using the ARC codes coupled to OpenMC with PyARC is a sufficient method to study the GFR to acceptable errors in steady state. Furthermore, a start-up to equilibrium fuel loading pattern and refueling scheme for the GFR was proposed that allows the core to operate nearly at the 16 year long cycle target and maintain reduced power peaking factors.by the U.S. Department of Energy, Office of Nuclear Energy, under Contract No. (DE-NE0009052), and the Subaward to the University of Wisconsin-Madison, GA- 4500093436
Integrating Programs for Impact on Green Lake
No full textIntegrating Programs for Impact on Green Lak