21793 research outputs found
Sort by
Blackbox Security Auditing for Network Functions Virtualization (NFV)
Get PDFOver the past decade, Network Functions Virtualization (NFV) has revolutionized networking by leveraging virtualization to separate Network Functions (NFs) from dedicated physical hardware. However, this architecture introduces unique security risks, such as stealthy attacks causing discrepancies between tenant-level NF specifications and cloud provider-level deployment. To safely utilize NFV, robust security auditing mechanisms are crucial to ensure compliance and detect breaches. Yet, existing methods face challenges: NFV tenants have limited access to cloud infrastructure, and providers are hesitant to share data due to confidentiality concerns. Relying solely on providers for auditing may overlook tenant-specific requirements and legitimate modifications by attackers. Furthermore, current solutions often require unrealistic infrastructure modifications. This thesis introduces novel auditing solutions for both tenants and providers of NFV, addressing these limitations. Firstly, an interactive anonymization tool called iCAT facilitates selective, privacy-preserving data sharing between tenants and providers. It utilizes an anonymization space to model various anonymization techniques, translating requirements from both parties into suitable primitives using NLP and ontology modeling. Secondly, a tenant-based, two-stage solution enhances auditing autonomy. The first stage utilizes tenant-side information to detect integrity breaches, while the second stage anonymizes provider-level data for tenant verification, offering control, transparency, and accuracy in breach identification. Additionally, a cryptographic approach is combined with side-channel watermarking to bolster tenant security. This lightweight solution enables continuous detection and classification of cloud-level attacks on service function chains, encoding cryptographic trailers as side-channel watermarks. This approach ensures verifiable attack detection without significant overhead, overcoming challenges such as limited side channel capacity and packet delay. By addressing these issues, the proposed solutions aim to enhance the security of NFV deployments and enable safer utilization of this innovative networking architecture
A Tableau-based Algebraic Calculus for Description Logic SHOIQ
Get PDFThe growing demand for efficient knowledge representation and reasoning in the era of interconnected systems and extensive data collection motivates this thesis. Addressing the limitations of existing Description Logic (DL) reasoners, we focus on the challenges posed by qualified cardinality restrictions (QCRs), nominals, and inverse roles. These constructs, though crucial for expressive ontologies, often hinder computational efficiency in traditional reasoning approaches. Consequently, real-world ontologies either exclude them or employ very small numerical values. This motivates our exploration of a novel reasoning approach, employing algebraic methods, aiming to enhance DL reasoning with a focus on large numerical restrictions.
In this thesis, a novel algebraic tableau calculus for SHOIQ is presented for deciding ontology consistency. This hybrid approach integrates standard tableau-based reasoning with algebraic reasoning to handle a large number of nominals, QCRs, and their interaction with inverse roles. The algorithm extends the previously presented algebraic tableau algorithm for SHOI. Numerical restrictions imposed by nominals and qualified number restrictions are encoded into a set of linear inequalities. The knowledge about other axioms, such as universal restrictions, role hierarchy, subsumption and disjointness, is also embedded in order to get a more informed mapping of QCR satisfiability to feasibility. Column generation and branch-and-price algorithms are used to solve these inequalities. The feasibility test for the linear inequalities can be computed in polynomial time. Rigorous proofs ensure soundness, completeness, and termination of the reasoning procedure.
In practice, the proposed reasoning approach, implemented in the Cicada prototype, demonstrates its effectiveness against existing state-of-the-art reasoners. Empirical evaluations using synthetic and ORE 2014 datasets reveal Cicada's robust performance against increasing numerical values, showcasing its viability for handling expressive ontologies. Despite its more focused optimization techniques, Cicada outperforms other reasoners in certain scenarios, providing a promising avenue for practical applications requiring efficient DL reasoning
Image Moment-based Visual Servoing for Satellite Target Tracking using a Robotic Manipulator
Get PDFRobotic manipulators have become indispensable tools in space operations. Over the past few decades, manipulators like Canadarm2 have played significant roles, ranging from repair tasks to the complex process of capturing servicing satellites. On this basis, the main objective of this research is to automate the satellite-catching process on the International Space Station (ISS). The study employs an image moment-based visual servoing to fulfill the task. Various image features have been introduced to control the manipulator’s movements using image moments. Yet, these features often face the challenging issue of coupling between six degrees of freedom movements, a problem that many researchers have aimed to solve. Nevertheless, previous literature has not completely addressed decoupling, which reveals the need for alternative approaches.
In this research, two novel approaches, function-based visual servoing and deep neural network (DNN)-based visual servoing, were developed to address this challenge. In the first approach, we introduce a novel general image feature function whose numerator and denominator are the polynomials with terms consisting of various image moments and adjustable parameters. Through the optimization process, two distinct rotational features about the x and y axes are formulated with the optimally tuned parameters. The second approach integrates DNN to estimate the 6D pose of the camera, yielding six decoupled image features.
Experimental results from the Denso manipulator confirm that decoupling image features can improve the controlling performance of the manipulator for capturing servicing satellites. The DNN-based visual servoing method could potentially enhance the performance of Canadarm2 in catching satellites, achieving a 32.04% average reduction in pose error and enhancing the velocity’s precision by 21.67% over traditional methods
Basic Ecclesial Communities: Fertile Ground for Religious & Social Harmony
Get PDFReligious intolerance and blatant polarization along the lines of caste, colour, and racial differences are on the rise in India, a trend that has been compounded by the explicit religious nationalism of the incumbent ruling party, the Bharatiya Janata Party (BJP). In the face of this rising intolerance and polarization, a substantial number of Indians from different demographics have begun to envision and participate in people-centred initiatives to nurture religious and social harmony. The present article proposes that Basic Ecclesial Communities – a practical, resourceful, and sustainable means of fostering community and religious harmony through interfaith sharing, hospitality, and a welcoming mindset – represent an impactful contribution to this effort
Exploring the spatiotemporal patterns of theta-band activity during rapid-eye movement sleep: a magnetoencephalography analysis
Get PDFTheta oscillations (4-8 Hz) are a prominent electrophysiological feature of rapid-eye move- ment (REM) sleep. Theta activity during REM sleep has been linked to memory consolidation; however, the role of cortical theta oscillations in this process remains unclear. Interestingly, theta rhythms are not exclusive to REM sleep but also appear in frontal regions during resting wakefulness and working memory tasks. To advance our understanding of human REM sleep and the mechanisms that support memory processing, a spatially resolved, whole-brain char- acterisation of REM oscillatory activity is essential. Magnetoencephalography (MEG) offers high temporal and spatial resolution, making it ideal for examining the topographic distribu- tion of theta oscillations in REM sleep. In this study, we recorded electroencephalography (EEG)/MEG data during overnight sleep in 10 healthy subjects. We also analysed a separate MEG/EEG data of 17 healthy subjects who performed a working memory task. Our aims were to characterise the spatio-temporal patterns of theta-band activity during REM sleep by 1) dis- tinguishing theta from the overlapping alpha (8-12 Hz) network, 2) comparing REM and non- rapid-eye movement (NREM) sleep oscillatory activity, and 3) evaluating similarities between REM and working memory task theta patterns. Our results show theta activity in frontal mid- line regions is best observed within a focused 5-7 Hz range, separating it from occipital alpha activity. Theta-band activity was greater in REM sleep compared to NREM in frontal-central, parietal, temporal, and subcortical regions. Theta topographies during the working memory task correlated positively with phasic REM sleep. These results enhance our understanding of REM sleep physiology and suggest future research targets for learning and memory roles
Detection of Counterfeit Coins Using Multimodal GPT-4 and Vision Transformer
Get PDFThe proliferation of counterfeit coins poses a substantial threat to the integrity of monetary systems and the stability of financial markets. Advanced counterfeiting techniques allow these fraudulent coins to closely mimic genuine ones, complicating the detection process and necessitating robust methods capable of discerning minute differences between genuine and fake coins. This thesis addresses the problem of counterfeit coin detection by introducing a diverse dataset comprising high-resolution images of both Danish and Chinese coins, categorized into genuine and counterfeit sets across multiple years.
To tackle the detection task, we employ two advanced approaches: a Vision Transformer (ViT) model and a multimodal GPT-4 model. The ViT model leverages its self-attention mechanisms to capture intricate patterns and details within the coin images, while the GPT-4 model integrates both visual and textual data, utilizing various prompting techniques to enhance its performance. Our results show that the ViT model outperforms previous methods and the state-of-the-art in terms of accuracy and robustness, achieving a remarkable 99.31% accuracy. The GPT-4 model, although primarily designed for natural language processing, demonstrates promising capabilities in counterfeit detection, particularly with advanced prompting strategies like Chain-of-Thought and Generated Knowledge.
This research advances the current state-of-the-art in counterfeit coin detection and highlights the potential of few-shot learning and transfer learning in achieving high accuracy with limited training data
Forecasting Electricity Load and Wind Generation: A Comparative Analysis of Machine Learning Models Enhanced by Bayesian Optimization under Different Sampling
Get PDFThe study explores the application of advanced machine learning techniques to forecast electricity load and wind generation data, focusing on the optimization and comparative analysis of various models. Given the critical importance of accurate energy forecasting in managing power grids and integrating renewable energy sources, this research seeks to enhance forecasting precision through the application of Bayesian optimization for hyperparameter tuning across multiple models. Utilizing time-series data, this study systematically evaluates the performance of several predictive models. Each model's parameters were meticulously optimized using Bayesian techniques to identify the most effective configurations for handling the complex dynamics of energy data. The research methodology involved a comparison within single datasets to identify the best model. Subsequently, the best-performing models were further analyzed across different datasets to validate their robustness and generalizability. The primary evaluation metric is the Root Mean Squared Error (RMSE), complemented by additional metrics to provide a comprehensive assessment of model accuracy and effectiveness. Key findings demonstrate that while some models excel in capturing overall trends, challenges remain in addressing the volatility and variability inherent in the data. The insights derived from this study not only advance the field of energy forecasting but also offer practical implications for energy policymakers and stakeholders in optimizing grid performance and renewable energy integration
What do Learners See in ChatGPT? Challenges, Benefits, and Writing in the Era of Generative AI Literacy
Get PDFThis thesis examines the rapid, uncritical adoption of ChatGPT, a widely used generative AI model, by focusing on students' perceptions and use of the tool for writing tasks. The study explores three main questions: how learners use ChatGPT, the advantages and disadvantages they identify, and their awareness of its ethical implications. Using a qualitative approach, semi-structured interviews were conducted with 31 university students and recent graduates, followed by thematic analysis to uncover recurring patterns. Findings show that participants mainly used ChatGPT for drafting, research, editing, and brainstorming, valuing its efficiency and usability. However, they expressed concerns over generic, repetitive outputs, limited research capabilities, and the risk of deskilling, fearing dependence on the tool might erode their own skills. While participants recognized some ethical issues, particularly in education, disinformation, and privacy, awareness of bias, transparency, and sustainability was limited. Despite these drawbacks, participants generally maintained a positive attitude towards ChatGPT, with a strong interest in maximizing its benefits while attempting to manage its limitations. This study underscores the need for promoting critical engagement with AI through Human-Centred AI (HCAI), which emphasizes ethical considerations in technology use. The findings lay a foundation for future research aimed at addressing misconceptions about generative AI and creating educational strategies for ethical AI integration
Efficient Fine-Tuning Strategies for Federated Learning: Optimizing Model Performance Across Distributed Networks
Get PDFFederated Learning (FL) allows a global model to be trained collaboratively by a number of clients without sharing data. This setting is often characterized by resource-constrained clients con- nected over a low-bandwidth network. Hence, algorithms designed for the setting must account for important factors such as computer and memory requirements, robustness under changing data distributions and communication. Recent works, have started demonstrating the benefits of using pretrained models over random initialization on these considerations. We cover these recent ad- vancements before introducing methods conceived along the same lines. We show that in the FL setting, fitting a classifier using the Neurest Class Means (NCM) can be done exactly. We demon- strate its efficiency and combine it with full fine-tuning to produce stronger performance. Then, we introduce an adapted zeroth-order method capable of bringing a model to convergence with a mini- mal per-round compute budget while reducing the memory burden for clients during training down to that of inference. This work presents several experiments demonstrating the effectiveness of the proposed methods and highlights the importance for additional work into the application pretrained models in the FL setting
Thermodynamic and environmental analysis of heat and power generation using renewable fuels.
Get PDFThis study evaluates compact cogeneration technologies for urban areas with varying energy demands, focusing on renewable fuels such as hydrogen and biogas, compared to fossil fuels like methane. Among various options, proton exchange membrane fuel cells (PEM fuel cells) and recuperated micro-gas turbines (MGTs) are identified as the most promising technologies. Their performance under diverse scenarios, including control strategies, fuel choices, and operational conditions, is thoroughly modeled.
For PEM fuel cells, detailed electrochemical and thermal models simulate electricity and heat production, while for MGTs, a comprehensive model optimizes heat recovery and control strategies. The developed control strategy involves precise bypass valve adjustments to regulate mass flow distribution, improving efficiency. Heat management is further enhanced by coordinating bypass valve settings with storage tank cycles and auxiliary boiler transitions.
PEM fuel cells are shown to excel in high-efficiency cogeneration due to their direct conversion of chemical to electrical energy at low operational temperatures, minimizing heat loss and optimizing hydrogen utilization. MGT systems, on the other hand, benefit from hydrogen combustion’s higher flame temperatures, boosting power generation. Parametric analysis reveals that increasing rotational speed, pressure ratios, and working parameters in MGTs enhances power output, while higher cell counts and ambient temperatures improve PEM fuel cell efficiency and hydrogen consumption.
To reduce emissions from MGTs, a dual axial swirler combustor is proposed, ensuring uniform temperature distribution, minimizing hot spots, and enhancing fuel-air mixing. These features improve combustion efficiency and stability under partial loads, effectively lowering NOX and CO emissions. The emission characteristics are assessed using CFD simulations, an Equivalent Chemical Reactor Network (ECRN) model, and a custom mathematical model. Hydrogen combustion is associated with high NOX emissions due to its flame temperature, while methane and biogas show lower NOX concentrations. However, the inert CO2 in biogas presents challenges for efficiency.
In summary, this research provides a robust framework for evaluating renewable-fueled cogeneration systems, offering strategies to enhance efficiency and reduce emissions, supporting urban energy sustainability