21793 research outputs found
Sort by
Assessing the Robustness of HAR Deep Learning Models against Variability
Get PDFDeep learning (DL) Human Activity Recognition (HAR) models using wearable inertial measurement unit (IMU) sensors have shown great promise in applications like continuous healthcare monitoring and early disease prediction. However, most DL HAR models remain untested in real-world scenarios laden with variabilities; rather, they are trained and tested on constrained and closely curated HAR datasets that assume an ideal setting. This thesis explains the effects of real-world variabilities like subject, device, position, and orientation on the performance of DL HAR models. Due to the inability of existing datasets to isolate variabilities, we collect our own, the HARVAR dataset. We isolated the effect of different variabilities and provided a nuanced understanding of how each affects DL HAR models' performance. Maximum Mean Discrepancy (MMD) was used to quantify shifts in data distribution due to each isolated variability and drew a relationship between the drop in performance and the change in data distribution. The REALDISP dataset was used to perform a case study to understand the effects of compounded and unisolated variabilities in the real world. This study found that different variabilities have varying effects on the DL HAR model performance, from insignificant to detrimental. We showed a negative correlation between the MMD and the performance drop of the DL HAR models in the results drawn from both HARVAR and REALDISP datasets. The study emphasizes the need for more robust models and the development of pre-processing methodologies to optimize the IMU data for training robust DL HAR models
Essays on Structural Labour Supply and Government Policies
Get PDFThis thesis uses a structural modeling approach to assess labour supply and evaluate policy programs.
The first chapter compares labour market outcomes for high school dropouts to graduates in Quebec, Ontario, Alberta, and British Columbia. Results show that dropouts face worse outcomes across all provinces, with Quebec having a significantly higher proportion of male dropouts. Simulations aimed at boosting employment incentives for low-skilled individuals emphaise the importance of long-term strategies that enhance skill acquisition and reduce financial barriers. Current welfare eligibility criteria offer limited incentives to transition from welfare to work at modest wages.
The second chapter focuses on modeling individual heterogeneity, particularly unobserved characteristics, using random coefficients. It uses Monte Carlo simulations across six scenarios with varying shapes and variances for the distribution of unobserved characteristics. Findings reveal that methods accounting for heterogeneity perform well when variances are small, but become sensitive to distribution shapes as variances increase, indicating the need for more flexible models in high-variance contexts.
The final chapter examines the labour supply of single mothers, with a focus on childcare utilisation and social assistance participation. Contrary to traditional views, the study finds that childcare costs are no longer a significant barrier to employment, with access to childcare being a more critical issue. Policies targeting direct employment incentives may be more effective in increasing labour force participation. The chapter also highlights the role of unobserved preferences in shaping work decisions, suggesting that current programs may be limited by not fully addressing these behavioural factors
Workforce planning for SMEs under stochastic labour turnover
Get PDFIn today's rapidly changing manufacturing industry, even minor uncertainties can cause significant disruptions for small and medium-sized enterprises (SMEs). Although previous studies on operational supply chain networks have focused primarily on addressing uncertainties related to demand fluctuations, machine breakdowns, and unpredictable events such as natural disasters and geopolitical disruptions, this paper specifically addresses workforce uncertainty due to stochastic turnover rates. As an extension of the multistage workforce capacity planning problem with turnover proposed by \cite{2007Successive}, this study builds on their workforce planning network, which incorporates decisions around transferring, hiring, and firing, by introducing a proficiency ranking system. This system classifies the workforce into three proficiency levels, each with a specific production rate according to the worker's status. Integrating this proficiency ranking system into the planning network allows a more comprehensive evaluation of workforce capabilities to meet the required demand. Results from numerical experiments demonstrate that the modified model offers an optimized workforce planning solution, balancing cost and time effectively, to help SMEs achieve their demand targets under conditions of uncertain workforce turnover
A Comprehensive Analysis of Security Questions in Web Authentication
Get PDFWith the growing prevalence and sophistication of Internet services, user account security has become a critical concern. Security questions, widely adopted as a secondary authentication method, play a pivotal role in various online services. Although research on security questions has a long history, key gaps remain, particularly concerning user perceptions about security questions and the requirements used by websites for selecting and answering security questions. In this thesis, we address these gaps through a two-part study: (1) a comprehensive user survey (N = 292) that captures insights from a diverse and largely representative sample of the US population and (2) an analysis of an extensive set of 26 security requirements across 73 websites, also aiming to uncover security practices and weaknesses in their authentication systems (i.e., answer length restrictions). Additionally, we gather and analyze common online security questions (totaling 1913 questions) across several dimensions, including memorability, consistency, applicability, confidentiality, and specificity.
Our findings reveal previously unreported user misconceptions, such as users' believing that websites already possess correct answers to personal security questions. We also find that many websites allow insecure practices, such as accepting single-character, offering limited question choices, or identical answers for multiple security questions. By addressing both user perceptions and website security requirements, we provide a comprehensive understanding of the weaknesses in current security question practices and contribute to the discourse on improving authentication methods
Exploring the Potential Trade-Offs of Canada’s Participation in a Global Emissions Trading System Using A Multi-Sector, Multi-Region CGE Model
No full textLike many nations, Canada faces challenges stemming from climate change. Therefore, it aims to reduce overall emissions, measured in megatonnes of CO2 equivalents (MTCO2eq), by 40-45%, relative to 2005 levels by 2030 and achieve net zero emissions by 2050. This paper introduces Environment and Climate Change Canada’s Multi-Sector, Multi-Region (EC-MSMR) recursive dynamic computable general equilibrium (CGE) model, capable of evaluating multiple pathways to reaching emissions-based targets using market-based policies. CGE models capture direct and indirect effects in response to a policy change, making them excellent tools for evaluating economy-wide environmental policies. The EC-MSMR model delivers granular insights into Canada’s emissions and economic activity on the global stage by incorporating data from seventeen aggregated regions and twenty-three commodity-producing sectors, along with three final demand sectors: Consumption, Investment, and Government Spending. With this model, this paper analyzes Canada’s participation in a global Emissions Trading System (ETS) with perfect commitment versus a domestic carbon pricing schedule that adjusts itself to the shadow price that achieves the 2030 target. Results indicate that participating in a global ETS provides slightly greater economic growth and welfare while reducing reliance on fossil fuels to domestic carbon pricing alone. Although both policy experiments meet Canada’s 2030 target, both scenarios experience lower GDP and welfare outcomes than a baseline consisting solely of existing policies and no additional action
Early Childhood Educators’ Perspectives on Love and Care
Get PDFABSTRACT
Early Childhood Educators’ Perspectives on Love and Care
Katherine Pauls, Ph.D.
Concordia University, 2024
The concept of love in early childhood education is complex and challenging to define. However, understanding early childhood educators' views on love is essential for supporting educators and children in childcare settings. This phenomenological study examines early childhood educators’ beliefs and expressions of love and how they differentiate between love and care in their interactions with children. Their perspectives include, amongst others, influences from parents, coworkers and children. Eight early childhood educators from Montreal, Canada, who work with children under five years of age were recruited for this study. They participated in in-depth individual and focus group interviews. During the interviews, the participants were asked questions about their perspectives on love and care in their childcare settings as well as inspirations and challenges to showing love. They each gave concrete examples of how they showed love, including physical affection, curriculum actions and emotional support. The focus group discussions allowed participants to share their thoughts on personal and traditional definitions of love, deepening the dialogue around these complex concepts. The data was analyzed using In Vivo and Emotion coding and four themes were identified: (1) How Educators Believe They Show Love, (2) Factors That Educators Believe Shape the Way They Show Love, (3) Factors Educators Believe Act as Barriers and Challenges to Showing Love, and (4) How Educators Understand Love and Care in Their Work. These themes provide valuable insights into the educators' experiences and interactions within early childhood settings, highlighting the emotional intricacies of their roles, specifically as they alter their expressions of love and care for the children. This research has important implications for early childhood education training and support, helping pre-service teachers better understand their work's emotional dimensions and navigate the challenges of forming and maintaining meaningful relationships in childcare environments
Crowd Counting with Wi-Fi Probe Requests: A Selective Information Elements-based Approach Supported by Generative Data Augmentation
Get PDFCrowd monitoring is essential for smart city applications, particularly for optimizing public transit systems. To address this need, we propose a privacy-conscious crowd-counting pipeline using Wi-Fi probe requests. This pipeline is designed to adapt to the challenges posed by the randomization of Media Access Control (MAC) addresses, which serve as unique identifiers for devices on a network. Our approach leverages a random forest-based feature selection process to identify key Information Elements and frame attributes then applies DBSCAN clustering with adaptive parameter optimization for device counting. A diffusion model generates synthetic tabular data to mitigate the limited availability of labelled data, enhancing model robustness. Experimental results demonstrate improved accuracy in device counting, achieving a V-measure of 0.952, an average silhouette score of 0.789, and reliable clustering counts
MEASURING IMPROPER TOKEN INVALIDATION IN REAL-WORLD WEB LOGINS
Get PDFIn this thesis, we examine token invalidation flaws in real-world web applications.
While previous research has focused on cookie invalidation upon user logout, we investigate
the issue by examining token-based authorization in various contexts (e.g., user logout,
user verification, password recovery). Specifically, we focus on JSONWeb Tokens (JWTs)
to look for invalidation flaws, as JWTs are stateless and require web servers to implement
a separate invalidation mechanism for instant invalidation. To audit invalidation flaws, we
conduct a large-scale study on the top 1 million websites from Google’s Chrome User Experience
Report (CrUX). We develop an automated tool, LoginPlus that handles tasks such
as user registration, login and password recovery. Utilizing LoginPlus, first we identify
JWTs used to fetch resources from a web server while the user is authenticated and determine
whether these tokens are invalidated upon user logout. LoginPlus also handles all
emails sent during account creation and password recovery, identifying whether tokens included
in URLs sent to users for email verification or password recovery are invalidated
after a single use, checking if these URLs remain reusable over time. Finally, we evaluate
whether the websites invalidate all previously active sessions after a user resets their
password through the password recovery mechanism (e.g., ‘forgot password’).
Our analysis provides a comprehensive overview of the current state of invalidation issues
in token-based authorization schemes across real-world websites and reveals several
significant findings regarding token invalidation mechanisms in web authorization. Firstly, 85% of the websites using JWTs for authorization do not implement an explicit invalidation
mechanism upon user logout. Additionally, 2.67% of websites log users in into the
system via verification URLs directly, and 48% of these sites do not invalidate the tokens
in the verification URLs even after 24 hours. Furthermore, 13.8% of websites that allow
password recovery through email URLs do not invalidate the tokens after they have been
used once. Lastly, 54% of the websites where we successfully reset a password do not
invalidate previously active sessions
Data-driven Security Monitoring System for Cyberattacks on SSDCs in DFIG-Based Wind Parks
Get PDFThe massive integration of wind parks (WPs) in the modern power grid resulted in a significant concern regarding the security of the entire grid. These concerns are more important in the presence of WPs with inherent stability issues, e.g., when doubly-fed induction generators (DFIGs) are connected to series-compensated transmission systems. This thesis presents a novel real-time, data-driven security monitoring system to detect false data injection (FDI) and denial of service (DoS) cyberattacks targeting the subsynchronous damping controller (SSDC) in DFIG-based WPs.
A detailed and realistic electromagnetic transient (EMT) model of a DFIG-based WP is developed, along with the design of an SSDC to mitigate the subsynchronous control interaction (SSCI) phenomenon. The cyber vulnerabilities within the WP system, based on IEC 61400 standards, are analyzed to identify potential attack vectors in its cyber layer. It is demonstrated that such attacks can render the performance of SSDC ineffective, resulting in instability and sustained oscillations. To counter these issues, a real-time security monitoring system leveraging a customized recurrent neural network (RNN)-long short-term memory (LSTM) networks model is proposed to identify FDI and DoS attacks against the SSDC. The performance of the developed RNN-LSTM model is benchmarked against well-known classifiers, including random forest (RF), k-nearest neighbors (KNN), and multilayer perceptron (MLP), demonstrating superior detection accuracy. The effectiveness of the proposed model is further validated using unseen data, ensuring its effectiveness and generalization capability. Additionally, the proposed model exhibits low latency, making it suitable for near real-time operations in WPs
Assessing the Contribution of the Global Fund in the fight against Malaria: A case of sub-Sahara Africa and Southeast Asia Economies.
Get PDFThis thesis investigates the effectiveness of Global Fund financing in malaria control in sub-Sahara Africa and Southeast Asia regions using a sixteen year dataset from 2005 to 2020. Employing quantitative methods and non-experimental descriptive research design, the study found that Global Fund exerts significant negative impact on incidence of malaria in sub-Sahara Africa and Southeast Asia. In addition, the findings show that education, country policy and institution assessment, and gross domestic product are key factors influencing the effectiveness of the Global Fund financing for malaria control within sub-Sahara Africa and Southeast Asia regions. In relation to sub-Sahara Africa the study found that Global Fund, education, country policy and institution assessment, and gross domestic product exert significant negative impact on malaria incidence. However, in Southeast Asia, Global Fund, education, and gross domestic product failed to exert significant statistical impact on malaria incidence with the exception of country policy and institution assessment which showed negative impact on malaria incidence. These findings have implications for the international donor community and the health sector authorities of countries in the territories used for this study, as this would contribute to the ongoing efforts to combat malaria and improve global health outcomes. In conclusion, the study highlight the need for region-specific approaches that consider local socio-economic and political contexts, suggesting that a “one-size-fits-all” model is less effective in global malaria control