21793 research outputs found
Sort by
Security Weaknesses in E-commerce Platforms
Get PDFSoftware as a Service (SaaS) e-commerce platforms for merchants allow individual business owners to set up their online stores without any coding, or procuring any software/hardware. Prior work has shown that the checkout flows of such e-commerce applications are vulnerable to different kinds of logic bugs such as parameter tampering or workflow bypass, with serious financial consequences, e.g., allowing “shopping for free”. In this work, we first present a list of typical operations for such platforms, showing that there are several more functionalities beyond the check-out process, which can also lead to serious security consequences. We then leverage the fact that such platforms now heavily incorporate API requests and GraphQL calls (emerging) to design a semi-automated security analysis framework. We use this framework to analyze 32 representative e-commerce platforms (including 8 open-source ones) for seven different vulnerability categories; such platform host over 10 million stores as approximated through Google dorks. We uncover several previously unknown vulnerabilities with serious consequences, e.g., allowing an attacker to takeover all stores under a platform, and listing illegal products at a victim’s store—in addition to “shopping for free” bugs, without exploiting the checkout/payment process. We found 12 platforms vulnerable to store takeover and 6 platforms vulnerable to shopping for free, affecting thousands of stores (49000+ for store takeover, and 28000+ for shopping for free, as approximated via Google dorks). We have responsibly disclosed the vulnerabilities to all affected parties: two vendors have fixed the issues and four are still working. We have also requested four CVEs (amongst the 8 open source projects), and three CVEs have been assigned
Scaling Local Learning for Supervised and Self-supervised Learning
Get PDFTraditional neural network training methods optimize a monolithic objective function jointly for all the components. This can lead to various inefficiencies in terms of potential parallelization. Local learning is an approach to model-parallelism that removes the standard end-to-end learning setup and utilizes local objective functions to permit parallel learning amongst model components in a deep network. Recent works have demonstrated that variants of local learning can lead to efficient training of modern deep networks. However, in terms of how much computation can be distributed, these approaches are typically limited by the number of layers in a network. Hence, the first study explores how local-learning can be applied at the level of splitting layers or modules into sub-components, adding a notion of width-wise modularity to the existing depth-wise modularity associated with local learning. We investigate local-learning penalties that permit such models to be trained efficiently. Our experiments on various datasets demonstrate that introducing width-level modularity can lead to computational advantages over existing methods and opens new opportunities for improved model-parallel distributed training. The second study focuses on adapting existing local-learning frameworks to self-supervised learning tasks, specifically using the SimCLR method. However, existing local-learning frameworks lack in performance due to task-relevant information collapse in early layers. To address the issue, we propose modifying the local objective functions layerwise to gradually increase the problem difficulty with depth. We found that our method was able to maintain the similar performance as the end-to-end trained model while also increasing parallelization
Anatomical Differences across Cerebellar Neuronal Networks in Valproic Acid (VPA) Induced Rats
Get PDFThe cortico-cerebellar circuits are highly involved in the coordination of sensorimotor processes and are crucial for the expression of sensorimotor and, with more recent literature, cognitively demanding behaviors in early childhood. Through these circuits, cerebellar differences could impact behavioral changes characterized in many neurological disorders, including Autism Spectrum Disorder (ASD). Despite the increasing prevalence of the disorder, insights into the anatomical differences across cerebellar neuronal networks in ASD remain poorly understood. The objective of this thesis is to characterize the cerebellar circuitry in a valproic acid (VPA)-induced animal model of ASD. Tissues were stained with Cresyl violet and captured with a confocal microscope. First, we examined the counts in (1) Purkinje cell (PC) and (2) Deep cerebellar nuclei (DCN) cells while measuring their density. Consistent with our hypothesis, we found that there were: (1) significant losses in PC density in the posterior inferior regions of lobules Crus I, Crus II, PM, and Cop of the hemisphere, and (2) significantly higher DCN density of the medial nuclei, contrary to our hypothesis, in the VPA-exposed rats compared to the control. Then, we examined the granule cell layer (GCL) thickness measurements across the lobules of the vermis and hemisphere and found that GCL thickness of Lobule 6 was lower, while Lobule 7-8-9 and Lobule 10 were higher in the VPA-exposed rats compared to control. Finally, we found a correlation and cross-covariance between the PC and GCL thickness measurements in the lobules of the posterior lobe. These findings suggest alterations in cerebellar anatomy affected by the VPA, and therefore further support the implication of the cerebellum and cerebellar circuits in VPA-exposed rats
Exploring Pedestrian Road Safety in Public Transit Locations
Get PDFThis thesis studies the magnitude of pedestrian road collisions in public transit locations and addresses how the road and built-environment elements affect pedestrian safety at public transit access points (PTAPs). Collision count models and hotspot identification methods are utilized to address the research questions. Chapter 1 and Chapter 2 provide an introduction and literature review over pedestrian road safety in general, and specifically in public transit locations. Chapter 3 explains the methodologies that will be utilized in this research study. Chapter 4 establishes a relationship between pedestrian-vehicle collision counts and public transit services. Pedestrian collisions occur more frequently at intersections with the presence of a PTAP and with a higher bus traffic volume, a higher number of bus routes, and a higher public transit accessibility index. Hence, Chapter 5, explores how road geometry and built environment elements affect pedestrian-vehicle collision counts at PTAPs. The analysis shows that strategies such as road narrowing, sidewalk width increase, median refuges, presence of signal’s walk interval, and vehicle stop signs could improve pedestrian safety at PTAPs. Moreover, pedestrians are at more risk in PTAP where there are roads with higher road grades and more two-way streets than one-way streets. Chapter 5 continues with Empirical Bayes collision hotspot identification and examines a couple of collision hotspots in PTAPs in the case study of Montreal City. The study findings point out the need to improve pedestrian road safety at PTAP locations and offer engineering countermeasures for addressing this problem
Empty Spaces: The Concept Album and the Neoliberal Turn
Get PDFThe concept album, a musical form in which the songs on an album are narratively or thematically linked to tell a story, achieved its preeminence at a time of great social, political, and economic change. This project examines the relationship three concept albums bear to the political and economic doctrine of neoliberalism, and traces the ways these works respond to the trickle-down effects of neoliberal ideology within the social sphere. The first album, The Wall, was released in 1979, and engages with the doctrine of Thatcherism, widely regarded as one of the powerful originators of neoliberal governance, as well as its causes and social consequences. The analysis next moves to more contemporary examples, examining the Titus Andronicus’s 2010 album The Monitor and Marillion’s 2016 release Fuck Everyone and Run to map the ways in which neoliberalism, and the discourses it engenders, have shifted and mutated in response to historical and economic stimuli
Impact of the Pandemic on the Housing and Rental Markets: Evidence from U.S. Metropolitan Statistical Areas
Get PDFA distinguishing characteristic of the COVID-19 economic crisis is the unprecedented decline in human mobility. This thesis aims to examine the effect of the pandemic on housing and rental prices in the United States. Using 315 Metropolitan Statistical Areas (MSAs), I calculate their respective real estate market’s exposure to mobility changes, the increase in work-from-home (WFH), and the increase in state mandates to combat the virus’ spread. Additionally, I document whether MSAs dominated by Blacks, Latinos, and Hispanics are affected more than others. I find that as mobility decreases by 1%, there is a 0.07% and 0.3% decrease in housing and rental prices, respectively. The increase in WFH and state stringency positively affect the overall housing prices but does not affect the rental prices. I also find that the minority-dominated MSAs have a significant and negative impact from the pandemic for both prices
Hysterical: Comedy, Comebacks and #metoo
Get PDF“Louis C.K. Is Done,” declared critic Matt Zoller Seitz on November 9, 2017, the same day a New York Times investigation revealed that C.K. had masturbated in front of or on the phone with women in workplace contexts. Three months later, another auteur comedian, Aziz Ansari, faced accusations of pressuring an anonymous woman, Grace, into sexual acts. This thesis looks at the heated public response to both cases and how they came to symbolize aspects of #metoo, a feminist social movement targeting sexual violence that unfolded in the fall of 2017. The thesis conducts a discourse analysis of commentaries on C.K. and Ansari’s cases—when the allegations first became public and when the comedians later embarked on “comebacks”—in major media outlets as well as fringe feminist and right-wing publications in order to identify the ideological assumptions underpinning these responses. Through these responses, #metoo is constituted as a cultural clash wherein dominant norms of sexual violence as bounded and binary and comedy as a protected artistic sphere are challenged by counter-hegemonic feminist positions. At the same time, an anti-feminist backlash works to close down these challenges. The thesis situates this dialectical tension within the “warring, constantly moving contexts” of what Sarah Banet-Weiser calls popular feminism and popular misogyny (Empowered 13). Sexual violence, comedy, and #metoo become discursively linked as they are mobilized in relation to broader cultural contestations between feminism, misogyny, leftism, and anti-progressivism or “anti-wokeism.” Though the radical challenges of #metoo fade, their ripple effects continue
Connecting Physical Literacy and Injury Prevention Strategies: Assessing and Intervening Movement Skills in 8-12-Year-Old Children
Get PDFPhysical activity is associated with positive health outcomes in children and adolescents. The physical literacy model aims to promote lifelong physical activity by focusing on affective, cognitive, behavioral, and physical factors. Promoting physical activity in youth is a global objective; however, participating in physical activity and sports is associated with an increased risk of lower-limb musculoskeletal injuries, which is a barrier to physical activity participation. Injury prevention strategies use multicomponent injury prevention programs and screening tools that target modifiable risk factors of injury. The physical literacy model and injury prevention strategies use similar movement-related constructs but are rarely connected in the literature and practice. A screening tool that assesses movement competence and injury risk may be a valuable source of information to fit interventions in different contexts. An intervention based on the physical literacy model and multicomponent injury prevention programs may help enhance physical literacy constructs and neuromuscular performance and reduce the risk of lower-limb injuries. Addressing these related elements and constructs may favor adopting and maintaining physical activity. This dissertation consists of five chapters. Chapter one introduces the concepts used in the dissertation and states the hypotheses and objectives. Chapter two describes a systematic review with six meta-analyses that studied the characteristics and effects of multicomponent injury prevention programs on various fundamental movement skills in children and adolescents. Chapter three describes the evaluation of the concurrent and construct validity of the Children Focused Injury Risk Screening Tool (ChildFIRST). The ChildFIRST is a process-based assessment of movement skills that aims to identify 8-12-year-old children with poor movement competence and increased risk of lower-limb musculoskeletal injury. Chapter four was based on the evidence from the second chapter and the literature, and it describes the development, implementation, and feasibility testing of a neuromuscular warm-up for 8-12-year-old children. The neuromuscular warm-up was based on the physical literacy model and multicomponent injury prevention programs. The intervention positively affected physical literacy constructs, neuromuscular performance, movement competence, and injury risk profile, which was assessed using the ChildFIRST. Chapter five discusses the findings from chapters two, three, and four and offers a general conclusion and recommendations for future research
Hél čhaŋkú kiŋ ȟpáye (There lies the road): How to Make Art in a Good Way
Get PDFThis dissertation explores the possibilities of the ethical creation of artwork using artificial intelligence (AI) through the development of protocols based on Lakȟóta ontologies, resulting in collaborative artworks collectively titled Hél čhaŋkú kiŋ ȟpáye (There lies the road). It investigates how Lakȟóta ontologies are well-suited for creating ethical and reciprocal relationships with the nonhuman and articulates that perspective in the context of creating performance artworks and collaborative art-making processes. This project is conducted in collaboration with the Lakȟóta communities diasporically, the Oglála Lakȟóta community in Pine Ridge Reservation, and diasporic Indigenous communities internationally. Through these collaborations, this dissertation has involved recordings, installations, performances, and written documents considering our current and future relationships to nonhumans with special consideration to digital technology in general and AI specifically—while articulating ethical processes which define who and what is in relation. It considers the acute necessity for engagement with Indigenous ontologies through research-creation in order to develop artworks and technologies in ethical ways. This dissertation 1) seeks to understand Indigenous ontologies, specifically Lakȟóta ontology, by collaboratively defining relationships to nonhuman beings through the process of art-making; 2) develops culturally-grounded methodologies to guide the creation and refinement of AI wearable and digital technologies through performance, sound art, and visual art practice; 3) develops the foundations for a specifically Lakȟóta Artificial Intelligence system; and 4) suggests Indigenous (Lakȟóta) research methods for research and creation in collaboration with Indigenous communities
The Effects of Task Repetition on Chinese EFL University Students’ Task Accomplishment, Communicative Competence, and Willingness to Communicate
Get PDFEnglish has played an important role in China, and it has been given an important status in the school curriculum. However, only less than 1% of Chinese EFL learners are conversational (Smith, 2017). How to improve learners’ task accomplishment, communicative competence, and L2 Willingness to Communicate (WTC) are the main challenges in the Chinese EFL context. Previous studies have examined the effects of task repetition on L2 oral performance in terms of accuracy, fluency, and complexity. Fewer studies, however, have examined the effects of task repetition on task accomplishment, communicative competence, and L2 WTC.
Using data collected in four EFL classes in a Chinese university over nine weeks, this study examined the impact of task repetition on task accomplishment, communicative competence, and L2 WTC. Pre-test, post-test, and delayed post-test assessing students’ task accomplishment, communicative competence, and L2 WTC were performed in week 1, week 5, and week 9 outside class. Four classes of Chinese university students were divided into four groups: procedural repetition (n = 27), content repetition (n = 23), identical task repetition (n = 28), and a control group (n = 29). Students in the three repetition groups performed a treatment task once a week for three weeks outside class. Students in the control group followed the regular curriculum without carrying out any tasks. Four trained raters rated students’ task accomplishment and communicative competence based on two rubrics. L2 WTC data consisted of two parts: trait WTC measured by a WTC questionnaire and state WTC operationalized by self-rated WTC immediately after each test. Interviews with students were conducted to learn their perception of L2 learning and WTC in week 1 and week 9. A mixed ANOVA using SPSS was carried out to see if students have changed in terms of task accomplishment, communicative competence, and L2 WTC over time.
Results indicated that task repetition did not significantly affect task accomplishment, communicative competence, and L2 WTC. However, students in all four groups improved their task accomplishment and L2 WTC over time. Implications are discussed in terms of how to teach and assess with tasks in the Chinese EFL context