32584 research outputs found
Sort by
Discriminative Regions and Adversarial Sensitivity in CNN-Based Malware Image Classification
No full textThe escalating prevalence of malware poses a significant threat to digital infrastructure, demanding robust yet efficient detection methods. In this study, we evaluate multiple Convolutional Neural Network (CNN) architectures, including basic CNN, LeNet, AlexNet, GoogLeNet, and DenseNet, on a dataset of 11,000 malware images spanning 452 families. Our experiments demonstrate that CNN models can achieve reliable classification performance across both multiclass and binary tasks. However, we also uncover a critical weakness in that even minimal image perturbations, such as pixel modification lower than 1% of the total image pixels, drastically degrade accuracy and reveal CNNs’ fragility in adversarial settings. A key contribution of this work is spatial analysis of malware images, revealing that discriminative features concentrate disproportionately in the bottom-left quadrant. This spatial bias likely reflects semantic structure, as malware payload information often resides near the end of binary files when rasterized. Notably, models trained in this region outperform those trained in other sections, underscoring the importance of spatial awareness in malware classification. Taken together, our results reveal that CNN-based malware classifiers are simultaneously effective and vulnerable to learning strong representations but sensitive to both subtle perturbations and positional bias. These findings highlight the need for future detection systems that integrate robustness to noise with resilience against spatial distortions to ensure reliability in real-world adversarial environments
Smithian Merchant Towns and Good Government in The Wealth of Nations
No full textCommentators on Adam Smith’s Inquiry into the Nature and Causes of the Wealth of Nations often characterize Smith as a proponent of government in multiple areas, including security in person, property, contract, and some public goods. However, Smith understood government more expansively than people today. In volume I, book III, chapters I–IV, Smith describes the evolution of merchant towns in England that led to “good governance” that dismantled the feudal system. This limited government not only included the security provided by the formal laws and institutions. It also aligned the informal elements of individual civic ethics and self-reliance with the formal institutions. This Smithian combination of formal and informal government encouraged liberty and market exchange, which enriched towns and, then, agriculture. It established the foundation for English economic development, which made the nation wealthy
Library Science Literature, 2019–2025: An Exploration Using Critical Bibliometric Methods
No full textThis study applies critical bibliometric methods to the library science literature published from 2019 to early 2025. It combines citation network analysis and text analysis in novel ways to identify clusters of articles focused on trends and practice areas, and to analyze the citation rates and centrality of those clusters within the network of library science articles. The study provides an example of how critical bibliometric methods can be applied to provide context and nuance when assessing the impact of research and interpreting research impact metrics. Themes identified within the literature include the COVID-19 pandemic, diversity, artificial intelligence, and social media, and the study notes differences in citation rates between these trends, with articles in the cluster focused on diversity cited at a lower rate than those focused on technology. While the study employs a novel weighting method to mitigate the impact of journal self-citation, the preliminary results demonstrate the susceptibility of citation-based metrics to gaming by authors, journal editors, and publishers. Critical bibliometric methods, like those used in this study can illuminate flaws and biases in widely-accepted bibliometric approaches, and point towards unanswered questions about our perceptions of prestige, quality, and impact in academic research
Robust Hashing for Improved CNN Performance in Image-Based Malware Detection
No full textThis paper presents a comparative study on the impact of robust hashing in enhancing image-based malware classification. While Convolutional Neural Networks (CNNs) have shown promise when working with image-based malware samples, their performance degrades significantly when obfuscation techniques are taken into consideration to hamper the malware classification or detection. To address this, we apply a robust hashing technique that generates invariant visual representations of malware samples, enabling improved generalization under obfuscation implemented as image salting. Using a custom obfuscation method to simulate polymorphic variants, we evaluate MobileNet, ResNet, and DenseNet architectures across five salting conditions (0% to 40%). The results demonstrate that robust hashing substantially boosts classification accuracy, with DenseNet achieving 89.50% on unsalted data, compared to only 68.00% without hashing. Across all salting levels, models consistently performed better when robust hashing was applied, confirming its effectiveness in preserving structural features and mitigating adversarial noise. These findings position robust hashing as a powerful preprocessing strategy for resilient malware detection
Transcultural Competence and Empathy in Language Education: Imagining the Unimaginable
No full textThe 2007 MLA report offered guidelines and competencies for foreign language majors at American universities in the age of globalization. The notion of translingual and transcultural competence suggests that today’s foreign language education is not merely a matter of language acquisition, but humanistic learning. The ultimate goal for foreign language learners should be to gain alternative ways of seeing the world, namely, “imagining the unimaginable” (Ozick, 1987). This underscores the central importance of empathy in foreign language education. Yet the true challenge lies in how to assess such an abstract concept. Even eighteen years since its initial publication, the report remains highly relevant, especially today, as we witness cultural, ideological, political and socio-economic divisions and the accompanying conflicts rooted in a failure to imagine the perspective of “others.” This paper explores a new approach to assessing learners’ transcultural competence, focusing on the role of empathy in understanding the cultural “other.” Discourse analyses of two student final papers were conducted and compared, drawing on poststructuralist theories and a sociolinguistic analytical framework. The results show that one student constructed an imagined Japanese “other” through a process of projection and initiated an empathetic dialogue beyond time and space, whereas the other inadvertently reproduced an Orientalist discourse by negatively stereotyping Japanese people, culture and society. This study advocates for discourse analysis as an effective formative tool for assessing and improving existing syllabi and curricula
Family-Level Impacts of Equine-Assisted Occupational Therapy for Children With Disabilities: An Exploratory Mixed-Methods Pilot Study
No full textEquine-assisted occupational therapy (EAOT) has been shown to enhance functional, social, and quality-of-life outcomes for children with disabilities (McDaniel Peters & Wood, 2017; Prieto et al., 2022). While child-focused outcomes are well documented, significantly less is known about how EAOT participation may influence caregivers and the broader family unit, as existing research largely prioritizes individual outcomes over family experiences (Peters et al., 2017; Tan & Simmonds, 2018; Trzmiel et al., 2019; Xiao et al., 2023). Emerging qualitative studies suggest potential family-level impacts including family engagement, reduced parental stress, and enhanced feelings of pride, empowerment, and connection (Kalmback et al., 2020; Ozyurt et al., 2020; Tan & Simmonds, 2018) however, findings remain inconsistent and limited by small samples and methodological heterogeneity (Zoccante et al., 2021). This study addresses the identified gap by examining the family-level impacts of EAOT for children with disabilities through an exploratory occupation-based, mixed-methods approach
A Hierarchical Two-Stage Adversarial Training Framework for Defending LLMs Against Jailbreak Attacks
No full textLarge language Models (LLMs) are used in many AI tools people rely on today, and they can be tricked into ignoring their safety rules through jailbreak prompts. These jailbreaks happen in multiple ways, mainly in two ways: one involves tweaking the exact wording or token sequence to push the model into unsafe behavior, and the other uses clever phrasing and indirect instructions to slip past the LLM’s content filters. The existing gap is that the defenses focus on only one of these tricks, leaving the model vulnerable to the others. To address both problems, we propose a hierarchical two-stage adversarial training framework that systematically addresses both vulnerability classes. The first stage addresses token-level or word-level robustness through adversarial training with curriculum learning, applying gradient-based perturbations to fortify models against attacks such as Greedy Coordinate Gradient (GCG). The second stage extends protection to semantic-level threats via Adaptive Adversarial Prompt Refinement (AAPR) that synthesizes diverse semantic adversarial training samples. Attack Success Rate comes down for all the models when the model is added with the two-stage defense, compared to the stage before adding any defense. Comparative evaluation against established baselines (Self-Reminder, Unlearning, SmoothLLM) demonstrates that our hierarchical approach surpasses all alternatives across attack categories. The framework achieves effective defense while requiring LoRA fine-tuning on 0.1% of the model parameters
Spartan Daily, March 5, 2025
Get PDFVolume 164, Issue 18https://scholarworks.sjsu.edu/spartan_daily_2025/1017/thumbnail.jp
Performance Testing of Asphalt Binder Modified with Amine-Impregnated Zeolite and Plastic in Hot Mix Asphalt to Reduce Carbon Footprint
Get PDFThe rise in global temperatures, driven in part by significant transportation carbon emissions, necessitate sustainable solutions for infrastructure. Traditional asphalt binders and lime additives significantly contribute to carbon emissions, and conventional liquid amine-based antistrip agents, which are used to reduce moisture damage, lose efficacy over time. This study evaluates the performance of PG 64-16 Low Carbon binder, incorporating 10% post-consumer plastic and amine-impregnated zeolite (AIMZ) as a protective carrier for liquid amines. Researchers compare this low-carbon binder to conventional PG 64-16 binder and evaluate AIMZ against amine and zeolite separately (AZ) and a commercial liquid antistrip (LAS). The study tests three aging levels (3, 5, and 7 days), simulating 4, 8, and 10 years, respectively, of field aging in Southern California. The evaluation of moisture-induced damage uses the Tensile Strength Ratio (TSR), while the Hamburg Wheel Tracking (HWT) test assesses rutting resistance (the wear from tires and loads that occurs on roads). The IDEAL Cracking Test measures cracking resistance, Hello,We have a new project that needs to be published to ScholarWorks and registered with DOI.Attached are the report and RB PDFs.Thank you in advance,
and the Moisture-Induced Shear-Thinning Index (MISTI) and Multiple-Stress Creep Recovery (MSCR) tests analyze moisture susceptibility and rheological properties, all of which are important factors to consider in long-term efficacy. AIMZ demonstrated higher TSR values compared to those with AZ and LAS at both 5 days and 7 days of aging levels for both binders. Rutting resistance is comparable between binders, and low-carbon binder mixtures show improved cracking resistance over time. MISTI values suggest lower moisture susceptibility for the low-carbon binder, though MSCR results suggest it is best suited for low-traffic volumes. This study indicates that AIMZ effectively prolongs liquid amine efficacy and that low-carbon binders, despite some limitations, offer environmental and performance benefits. These findings support the potential for incorporating post-consumer plastics in asphalt pavements, promoting sustainability in infrastructure