32584 research outputs found
Sort by
Malware Generation and Classification using PixelCNN
No full textMalware poses a serious threat to both data privacy and system security. With the wide variety of malware families and the surge in cyber-attacks, the accurate classification of malware is crucial for building effective detection and prevention systems. In recent years, deep learning (DL) methods in computer vision have shown promise in classifying malware by converting malware files into visual representations and applying DL algorithms to classify the resulting images. Among the different approaches to malware family classification, image-based methods have gained significant interest. This research focuses on leveraging DL techniques for image-based classification of malware. The success of identifying malicious files largely depends on the quality and size of the training dataset, as well as its authenticity. However, a major obstacle in utilizing DL for malware detection is the shortage of training data. To address this challenge, this project utilizes the Pixel Convolutional Neural Network (PixelCNN) architecture to generate synthetic, class-specific malware images pixel by pixel, capturing intricate details of the images. Additionally, a CNN-based DL model was developed, demonstrating that training with a mixture of at least 15% real malware data alongside the generated images resulted in notable multi-class classification accuracy
Extraction of a Knowledge Graph of Biomedical Relationships
No full textRapid release in biomedical literature poses a challenge in linking information. This thesis aims to extract data from expanding datasets to identify and form meaningful relationships between biomedical entities. Large language models (LLMs) enable us to learn at a rapid pace. Creation of LLms from scratch are impractical. This thesis aims to collect a small dataset, containing biomedical papers, and use it to train large language models (LLMs) to extract entities from the text and learn the relationships between these entities. The experiment will be divided into two stages and utilize EU-ADR and ChemProt dataset. Starting with named entity recognition (NER), cleaned datasets will be inserted through four LLMs. To determine the best results, data will be inserted through training relation extraction (RE) models, followed by a display of results graphs and visualization
DETECTING AI-GENERATED NEWS ARTICLES USING UNSUPERVISED MACHINE LEARNING ALGORITHMS
No full textThe widespread adoption of Large Language Models (LLMs) has revolutionized text generation and heightened concerns over misinformation and the erosion of journalistic integrity. Detecting AI-generated text is critical to addressing these challenges, yet current detection methods face adaptability, scalability, and accuracy limitations. This research paper uses machine-learning techniques to explore the classification of human and AI-generated articles, including a mix of human and AI-written content. The primary focus is on evaluating the effectiveness of clustering algorithms (K-Means and Agglomerative Clustering), auto-encoders, and Part-Of- Speech Tag Transition Matrix Log-Likelihood for distinguishing between AI-generated and human-written texts. Our findings reveal that while models perform well on fully AI or human-written texts, mixed content introduces significant challenges for all classifiers. The rapid evolution of Large Language Models has produced AI-generated content that increasingly mirrors human writing patterns, linguistic nuances, and stylistic variations. This convergence between human and machineauthored text presents a fundamental challenge for detection mechanisms, particularly in hybrid content where AI contributions are seamlessly integrated with human writing. However, our ensemble approach combining multiple classifier models achieves comparable performance to Ghostbuster (96.5% vs 92.9%) overall accuracy for the Gemini Improved model), demonstrating particular strength in AI text detection, where Ghostbuster shows relative weakness
MyCelia: Cross-chain Data Oracle using FROST signatures
Get PDFThe interoperability of heterogeneous blockchain networks is the basis for the widespread application of blockchains in various fields. Cross-chain data oracles play a significant role in enabling distributed applications to exchange data and assets across different blockchains, thereby greatly enriching and expanding the application scenarios and use of blockchains. With the continuous advancement of blockchain technology, more and more researchers and industry participants have begun to focus on developing cross-chain data oracles. Current cross-chain data oracles face issues with trust, as they rely on centralized intermediaries or limited validator networks, increasing the risk of manipulation or single points of failure. They also struggle with latency and scalability, making it challenging for real-time data transfers between blockchains. Additionally, interoperability across different consensus mechanisms and protocols adds complexity, leading to compatibility and integration issues across networks. New technologies have been proposed recently, such as decentralized networks for data verification and FROST decentralized credentials in cryptographic solutions. This paper proposes a novel approach to implement cross-chain data oracles that provide small and inexpensive data proofs using FROST signatures for a secure interoperating blockchain technology ecosystem. This FROST signature enables extremely cheap data proofs that cost only 30k gas to verify and are only 64 bytes in size
On the Adversarial Robustness of Quantized Neural Networks Against Common Adversaries in Time-Series Forecasting
No full textReal-world edge applications now use modern machine learning models which require both resource efficiency and robustness against adversarial threats. Deep neural networks which include time series forecasting models still face risks from adversarial perturbations while quantization techniques used for memory and compute efficiency create unpredictable robustness challenges. This project investigates the adversarial resistance of Long Short-Term Memory (LSTM) models after applying post-training quantization at three different precision levels: 16-bit floating point (FP16), 8-bit integer (INT8) and custom 4-bit quantization. The Jena Climate dataset serves as our main benchmark for training a fullprecision LSTM model followed by multiple quantization strategies which include TensorFlow Lite converters and a custom 4-bit quantization method based on ACIQ techniques that use clipping and scaling with bias correction. The evaluation of all models occurs under clean and adversarial conditions through Fast Gradient Sign Method (FGSM) and Basic Iterative Method (BIM) attacks with various perturbation strengths. The results indicate that quantized models show minor accuracy losses in clean conditions yet 4-bit models demonstrate better resistance to adversarial attacks. We create non-retraining defense strategies to enhance robustness which include input clipping and temporal median filtering and Gaussian smoothing and feature squeezing. These lightweight defenses effectively restore performance after an attack by reducing adversarial degradation without requiring modifications to model parameters. The complete pipeline is validated on the Jena Climate dataset and further tested on two additional multivariate datasets—Beijing PM2.5 and Appliances Energy Prediction—demonstrating consistent robustness patterns across data types. Our research demonstrates essential trade-offs between compression and efficiency and robustness which provides useful guidance for creating secure time-series models suitable for edge AI deploymen
Mastering High-speed and Low Power Hardware System Design
Get PDFThe book “Mastering High-speed and Low-Power Hardware System Design” provides a comprehensive guide to the intricacies of designing embedded systems with a focus on high-speed and low-power constraints. It begins by introducing the fundamental components of embedded systems, including hardware, software, and real-time operating systems, before diving into the essential principles of PCB design. The book emphasizes the importance of understanding the printed circuit board (PCB) as a foundational element in creating electronic devices and explores the various types of PCBs, such as rigid, flexible, and rigid-flex, discussing their unique applications and benefits.
A significant portion of the book is dedicated to understanding the trade-offs between speed, power, and thermal constraints. The impact of these factors on performance and the design process is explored, with a focus on achieving a balance that meets specific system requirements. The author delves into various techniques, such as dynamic voltage and frequency scaling, multi-core processors, and advanced cooling solutions, to address these trade-offs while optimizing system efficiency.
The book also highlights the role of material selection in PCB design, comparing commonly used materials like FR4 and Rogers laminates, each suitable for different high-frequency and high-speed applications. It discusses the importance of thermal management in embedded systems, with specific solutions provided by Rogers for addressing thermal dissipation in power electronics and high-performance devices.
Furthermore, the book covers essential aspects of signal integrity, including transmission line effects, impedance matching, and noise management. Practical design rules are outlined to ensure that signal quality is maintained in high-speed circuits, with a focus on optimizing power distribution, decoupling, and PCB layout for minimal interference and optimal performance.
By blending theory with real-world applications, the book serves as an invaluable resource for engineers and designers looking to master the complexities of high-speed and low-power hardware system design
Using Video Education to Improve Patient Knowledge of Cellular Therapy Side Effects: Cytokine Release Syndrome, Neurological Toxicities, & Prolonged Pancytopenia
Get PDFThis project evaluated the effectiveness of short educational videos on CAR T-cell therapy, cytokine release syndrome, neurological toxicities, and prolonged pancytopenia for blood cancer patients. At Stanford Health Care in California, in the ambulatory clinic of Blood and Marrow Transplantation and Cellular Therapy (BMT-CT), we obtained a convenience sample of individuals with multiple myeloma or lymphoma preparing for CAR T-cell therapy. A pre-and post-questionnaire assessed the patients\u27 knowledge of CAR T-cell therapy and its side effects. Wilcoxon signed-rank tests were used to determine changes in knowledge. Thirteen patients enrolled, eight completed post-surveys, and four watched the videos. The median knowledge score did not change significantly after viewing the videos (10.09 vs. 10.32 after, z = -0.37, p = .715). Patients who watched the videos reported that they were helpful. However, most patients did not view them due to time constraints and feeling overwhelmed
Implementation of an Evidence-Based Hearing Education Project for Patients and Their Care Partners
Get PDFHearing loss is a significant communication issue that impacts relationships, safety, and mental and physical health. There is an increased risk of dementia as well as increased mortality associated with untreated hearing loss. Hearing aids are one of the main treatments for hearing loss, but they are frequently not used. Cognitive impairment, osteoarthritis of the hands, cost and complexity are some of the reasons that hearing aids do not get utilized. But research has shown that involving family members and care partners in trainings on how to use hearing aids may increase use of them. The goal of this project was to determine if an educational intervention, using multiple learning modalities, could increase knowledge about hearing loss and hearing aids and augment hearing aid use, specifically by offering training to those with hearing loss and their care partners. A single-group pretest posttest quasi-experimental project was done to measure knowledge gain, and a follow-up survey was conducted to determine if hearing aid use increased. The results showed that both knowledge and reported hearing aid use did increase after the educational intervention. This demonstrates that educating caregivers and family members could be a useful option to prevent the health and safety issues associated with untreated hearing loss
Enhancing Hospital Throughput with the Power of Reminders: Implementing a Discharge Before Noon Strategy at An Urban Teaching Hospital
Get PDFThe doctoral project deploys a Discharge Before Noon strategy at an urban teaching hospital to resolve discharge delays and improve patient throughput efficiency. Hospitals experiencing overcrowding must prioritize timely patient discharges to enhance clinical results and operational performance. The project adopted the PARIHS framework to steer implementation through its emphasis on evidence-based practices, contextual factors, and facilitation strategies. The research gathered discharge data and team coordination results before and after the intervention period. The findings revealed elevated discharge numbers before noon and substantial progress in interdisciplinary team collaboration. The initiative indicated an optimizing trend for discharge procedures, even though statistical significance was not achieved in discharge rates. The essential components of the program featured proactive discharge communication with patients alongside interdisciplinary discharge planning meetings and real-time discharge goal tracking. The study faced limitations due to the brief intervention length and inconsistent team dynamics that affected implementation processes. Organizations should shift project tasks to case management and resource nurses to sustain improvements and embed discharge efficiency practices into regular procedures for ongoing enhancement. Structured discharge methods and effective teamwork are crucial elements that improve patient-centered care delivery while boosting hospital throughput. The Discharge Before Noon initiative uses strategic interventions to reduce discharge delays while enhancing health outcomes in complex healthcare settings
Is the Mahāyāna Buddhist Doctrine of Mutual Dependence a Case of Begging the Question?
Get PDFThis paper employs a number of argumentative strategies that the Buddhist philosopher Nāgārjuna employs in his Yukti-corpus to demonstrate that the Mahāyāna Buddhist doctrine of mutual dependence is not a case of begging the question. More precisely, this paper demonstrates that to say that A and B are mutually dependent is not to say that what grounds B is grounded in what it itself grounds; it is to say that both A and B lack ontological independence, and that they are a part of a changing, interdependent whole. Mutual dependence thus is a case of dynamic holism, and not of begging the question