UVicSpace | Institutional Repository
UVicSpace is the University of Victoria’s open access scholarship and learning repository. It preserves and provides access to the digital scholarly works of UVic faculty, students, staff, and partners. Items in UVicSpace are organized into collections, each belonging to a community.
For more information about depositing items, see the Submission Guidelines.
Recent Submissions
Malware Detection and Categorization Using ML and LLMs
(2025-12-08) Singh, Damanpreet
The rapid growth of malware attacks has created an urgent need for automated systems capable of accurately detecting and understanding malicious behavior. This project presents a comprehensive work for Malware Detection and Categorization using Machine Learning and Large Language Models (LLMs). The system's goal is to improve cybersecurity by not just detecting malware but also producing concise, intelligible descriptions of every threat it finds. The Microsoft Malware Classification dataset, which comprises approximately 21,000 malware samples grouped into nine primary families with corresponding .byte and.asm files, was adopted for the project. Since only malicious samples were present in the original dataset, roughly 15,000 benign files were added to enable binary categorization of malicious and non-malicious programs. XGBoost, LightGBM, SVM (RBF), and KNN were among the machine learning models that were trained and tested independently on both datasets. By applying the SMOTE technique, the dataset imbalance was reduced, thereby improving classification accuracy and mitigating bias toward the majority ofmalware families. Using the SMOTE technique, class imbalance was addressed. The models were evaluated for both binary classification (malicious vs. benign) and multi-class family prediction, achieving high detection performance. To enhance interpretability, an LLM-based explanation module was integrated. Following classification, the anticipated malware family is sent to an LLM (through Ollama), which produces a natural-language synopsis outlining the traits, actions, and defenses of the malware. Users can upload files, view predictions, and read the generated explanations in real time thanks to an intuitive Gradio interface. In order to provide both technical accuracy and human interpretability, the developed system successfully blends large language models for explainable analysis with machine learning for precise detection. By assisting researchers and security analysts in proactive malware defense, this method advances the field of intelligent cybersecurity by bridging the gap between detection and comprehension.
Non-linear probabilistic load flow of power systems with wind and electric vehicles
(2025) Amid, Pouya; Crawford, Curran
This thesis presents a comprehensive probabilistic framework designed to assess the reliability of power systems increasingly influenced by renewable energy sources and Electric Vehicles (EVs). It addresses the critical need for methodologies that effectively incorporate complex system characteristics and uncertainties without compromising computational efficiency.
Traditional deterministic methods often inadequately capture the probabilistic nature of power systems, which is essential for understanding the impacts of variable renewable energy generation and stochastic loads associated with rising electric transportation adoption. To address these limitations, this study introduces a novel methodology integrating sequential reliability simulations with detailed probabilistic analyses of non-linear load flow equations. The core of this approach employs advanced cumulant-based methods that accurately represent higher-order statistical characteristics and correlations among multiple uncertainties, efficiently modeling non-linear system behavior and fluctuations in renewable energy outputs and demand patterns. This significantly enhances computational efficiency and improves the accuracy of reliability assessments.
Building upon this foundation, the thesis further develops the concept of cumulant-tensor-based Probabilistic Load Flow (PLF) analysis. This innovative methodology extends cumulant approaches to handle higher-dimensional probability distributions, providing deeper insights into system behaviors under various scenarios, particularly those involving large-scale integration of wind generation and extensive EV charging demand.
An indicative real-world case study using the BC Hydro (BCH) power system demonstrates the practical application of these advanced methodologies. Through sequential reliability simulations combined with cumulant-tensor-based PLF analysis, the study examines the effects of wind generation variability and diverse EV charging scenarios, including adoption levels of up to one million vehicles. The results highlight the mixed impacts on system reliability: while increased wind generation capacity offers potential reliability improvements in urban areas with substantial EV integration, it presents challenges for rural areas with limited balancing resources. Generation facilities typically exhibit robustness against such variability; however, critical transmission infrastructure experiences significant stress, underscoring the need for targeted investments to enhance system resilience.
By specifically analyzing critical transmission lines within the BCH system, the study identifies key vulnerabilities and suggests targeted opportunities for infrastructure improvements. The use of hourly PLF analysis to determine confidence margins of power flows facilitates economical infrastructure design by accurately identifying periods of peak stress, thereby preventing unnecessary over-design. While the integration of renewable sources brings clear environmental advantages, it also introduces considerable complexities in economic dispatch and system expansion planning. The developed probabilistic framework provides utility providers with practical tools to effectively and reliably manage these complexities.
Overall, the methodologies introduced in this study represent advancements in power system reliability assessment and are applicable to system adequacy studies and long-term planning and risk management. As the global energy landscape continues to evolve, the deployment of such advanced probabilistic frameworks is increasingly essential to ensure the resilience, efficiency, and sustainability of future power infrastructure.
Adapting academic integrity policies to incorporate generative AI tools
(2025) Sehgal, Manish; Pelton, Tim
The rapid rise of Generative AI (GAI) presents both challenges and opportunities for higher education institutions seeking to uphold academic integrity while embracing technological innovation. This dissertation investigates how top U.S. universities are adapting their academic integrity policies and practices in response to GAI. Through document analysis of 20 institutional policies, surveys of students, faculty, and policy makers, and an autoethnographic reflection on the researcher’s use of ChatGPT, the study provides a multi-faceted view of institutional responses to GAI. The findings reveal alignment across institutions on core ethical principles, but wide variation in policy clarity, specificity, and educational integration. Survey data highlight tensions between stakeholder groups, with students eager to adopt GAI tools but seeking clearer guidance, faculty expressing cautious openness and the need for support, and policy makers prioritizing risk management. The autoethnographic reflection offers insight into the practical and ethical complexities of using GAI in academic leadership. The study concludes that successful integration of GAI requires a holistic approach that combines adaptable policy frameworks with educational initiatives, dialogue, and ongoing review. It calls for higher education institutions to engage in collaborative stewardship of GAI technologies to ensure their responsible and inclusive use.
Juvenile smokescreens: Softening the harm of zoos, aquaria, and prisons through (human) children
(Cambridge University Press, 2022) Deckha, Maneesha
This chapter explores how human children soften the abusive edge of carceral spaces. Prisons, immigration detention centres, and zoos and aquaria are institutions that attract sustained public scrutiny from prisoner rights, migrant rights, anti-racist, and animal rights movements. Critics and scholars note the entwined nature of race,
gender, and species logics that shape and unite these spaces and object to the shortand long-term incarceration these institutions make possible as well as the conditions residents confined within experience. Prisoner rights, migrant rights, and animal rights critics also contest the messaging that these institutions and their proponents use to assure the public of the need for confinement and the ethical acceptability of the conditions captive animals and humans experience. These discourses, depending on the specific institution, highlight the larger public “law and order” interests of safety and border control, but also “progressive” interests of rehabilitation, conservation, and education.
In highlighting these latter “progressive” interests, carceral institutions seek to humanize themselves and their work to bolster their social credibility. This “humane-washing” occurs through long-standing rationales about rehabilitation for offenders in the prison context, and more recent rationales about the conservation of nature and conservation education in the zoo and aquarium context. It also, I will argue, occurs through a specific type of marshaling of the human child. I seek to add to the literature on “humane-washing”4 as well as contestations and uses of “childhood” and “family” narratives in general in this analysis. I apply a multispecies lens to consider how the real and imagined human child in the zoo and aquaria context, and narratives about what is in the best interests of human children in the immigration and prison context, figure into characterizing such carceral institutions as legally and socially legitimate spaces. The argument acknowledges that these carceral spaces can yield positive benefits for some, such as rehabilitation or rescue of a specific individual or even conservation of a specific species. However, it accepts the existing critical scholarly literature against such spaces overall to focus on the question of how carceral spaces mask their problematic and oppressive nature by integrating the presence of human children.
Animalization and dehumanization concerns: Another psychological barrier to animal law reform
(Psychology of Human-Animal Intergroup Relations, 2023) Deckha, Maneesha
Legal systems across the world classify animals as property. There is growing global momentum asking courts in anthropocentric legal systems to revisit this position through test-case litigation. This has resulted in a few discrete victories for animals, but not much more. An ongoing issue is general legal conservatism and the belief in human exceptionalism that judges exhibit in these and related cases. In addition to general human exceptionalism, this article argues that a further psychological block for judges can arise from concerns about exacerbating racism and other intra-human prejudices given histories and legacies of animalizing and dehumanizing certain human groups. The first aim of this study is to illustrate this psychological phenomenon impacting judicial decision-making in relation to race. The article discusses the 2022 decision by the New York Court of Appeals with respect to the ongoing captivity of Happy, an elephant at the Bronx Zoo. This decision is selected given its recent and landmark status in North America. The second aim of the study is to outline why the dissociation of humans from animals is counterproductive to eliminating racism and other intra-human prejudices and inequities. The third aim of the study is to explain why affirming human proximity and kinship to animals—and thus putting a positive spin on animalization—in the legal system would be a more effective anti-racist and decolonizing gesture.