Chaofan Dai, Qideng Tang, Wubin Ma, Yahui Wu, Haohao Zhou, and Huahua Ding, Science and Technology on Information Systems Engineering Laboratory, National University of Defense Technology, Changsha, China
Entity resolution (ER), which aims to identify whether data records from various sources refer to the same real-world entity, is a crucial part of data integration systems. Traditional ER solutions assumes that data records are stored in relational tables with an aligned schema. However, in practical applications, it is common that data records to be matched may have different formats (e.g., relational, semi-structured, or textual types). In order to support ER for data records with varying formats, Generalized Entity Resolution has been proposed and has recently gained much attention. In this paper, we propose PromptER, a model based on pre-trained language models that offers an efficient and effective approach to accomplish Generalized Entity Resolution tasks. PromptER starts with a supervised contrastive learning process to train a Transformer encoder, which is afterward used for blocking and fine-tuned for matching. Specially, in the record embedding process, PromptER uses the proposed prompt embedding technique to better utilized the pre-trained language model layers and avoid embedding bias. Morever, we design a novel data augmentation method and an evaluation method to enhance the performance of the proposed model. We conduct experiments on the Generalized Entity Resolution dataset Machamp and the results show that PromptER significantly outperforms other state-of-art methods in the blocking and matching tasks.
Entity resolution, data integration, deep learning, contrastive learning, prompt learning.
Lara Alotaibi1, Sumayyah Seher2, and Nazeeruddin Mohammad3, 1Department of Computer Engineering and Science, Prince Mohammad Bin Fahd University, Alkhobar, Saudi Arabia, 2Department of Computer Engineering and Science, Prince Mohammad Bin Fahd University, Alkhobar, Saudi Arabia, 3Cybersecurity Center, Prince Mohammad Bin Fahd University, Alkhobar, Saudi Arabia
The emergence of ChatGPT within the realm of computing has provided considerable advantages to a diverse array of individuals. However, it has also become a tool employed by adversaries to execute cyberattacks. This research paper examines the implementation of prompt engineering as a means to coerce ChatGPT into generating malicious content that deviates from its ethical boundaries. By leveraging these techniques, cybercriminals can effortlessly create a range of attacks, including phishing attempts, creating and propagating malware, backdoor attacks, and impersonation schemes, often in conjunction with deep fakes. To substantiate these cases, we successfully present concrete evidence by prompt engineering, which enabled the production of convincing phishing emails and code snippets for malware generation such as keyloggers. Additionally, we address the pressing concern of defending against these malicious activities, exploring effective approaches such as AI-generated text detection and system vulnerability detection.
AI-Powered Attacks, Backdoor Attacks, ChatGPT, Cyberattacks, Cybercrime, Malware, Prompt Engineering, Phishing Attacks.
Maria Alabdulrahman1, Renad Khayyat2, Kawthar Almowallad3, Zahra Alharz4, and Mohammad Abugurain55, 1Prince Mohammad Bin Fahd University, Alkhobar, Saudi Arabia, 2King Abdulaziz University, Jeddah, Saudi Arabia, 3King Abdulaziz University, Jeddah, Saudi Arabia, 4Imam Abdulrahman Bin Faisal University, Alkhobar, Saudi Arabia, 5King Abdullah University of Science and Technology, Jeddah, Saudi Arabia
We propose a novel approach to Arabic story generation by fine-tuning a pretrained Large Language Model (LLM). Our pipeline includes two stages: text generation and image generation. By fine-tuning the davinci-003 LLM on a dataset of 527 Arabic stories, we tailor the generated stories based on user preferences. For image generation, we utilize the Midjourney model. The results demonstrate the efficacy of fine-tuning a pre-trained image generation model on a limited dataset, as measured by the ROUGE score. Sarid’s contributions include addressing the lack of Arabic story generation models, providing a comprehensive dataset of Arabic stories, and integrating text and image generation for a cohesive story generation pipeline.
Artificial Intelligence, Fine-tuning, Generative Models, Image, Generation, Text Generation.
Jialiang Liu1 and Moddwyn Andaya2, 1La Salle College Prep, Pasadena, CA 91107, 2College of San Mateo, San Mateo, CA 94402
There has been a notable surge in the popularity of 3D donation programs, where numerous participants actively engage in mutual acts of charitable giving. Building upon extensive research conducted within these virtual philanthropic communities, we ve created and implemented a visualization system with the goal of providing users with an immersive and enjoyable experience while exploring and navigating extensive 3D donation networks. Our design harnesses familiar three-dimensional representations to introduce innovative techniques for comprehending the intricate connections within complex donation structures. It supports visual analysis and search functionalities, along with the automatic identification and visualization of philanthropic clusters. Through the deployment of public installations and controlled studies, our system has proven its usability, its capacity to facilitate discovery, and its potential to encourage enjoyable and socially engaging philanthropic endeavors.
3D Modeling, Donation, Computer Science, Unity, Website.
Mohamed Jacem Guezguez1 and Olfa Besbes2, 1Cogicom, Paris, France, 2University of Monastir, Monastir, Tunisia
The fifth generation (5G) network represents the latest evolution in mobile communication technology, offering several significant advancements over its predecessors, including 4G (LTE) and 3G. These advancements include faster speeds, lower latency, and a wealth of new capabilities. In parallel, unmanned aerial vehicles (UAVs), commonly referred to as drones, are gaining increasing popularity and becoming more ubiquitous. Integrating drones with 5G networks unlocks new possibilities and applications that harness the high-speed, low-latency, and extensive connectivity features of 5G technology. However, the misuse of drones can pose various risks and concerns, including issues related to privacy invasion and safety hazards. In response to these challenges, this research paper presents an innovative 5G Open RAN platform, featuring programmable software deployed on 5G gNodeBs, enabling the collection and monitoring of radio-sensitive events in relation to drone intrusion attacks. Additionally, a radio-based detection technique is proposed to identify threats and block unauthorized drones, thus safeguarding private infrastructures. To illustrate the effectiveness of this platform, a case study is included, demonstrating its capabilities in addressing drone intrusion attacks at an airport.
Mobile Network, Drone Attacks, 5G Networks, Beamforming, Network Slicing.
Ertuğrul AKBAŞ, Computer Engineering, Istanbul Esenyurt University, İstanbul, Turkey
This research paper examines the high risks encountered while using a Security Information and Event Management (SIEM) product or acquiring Security Operations Center (SOC) services. The paper focuses on key challenges such as insufficient logging, the importance of live log retentions, scalability concerns, and the critical aspect of correlation within SIEM. It also emphasizes the significance of compliance with various standards and regulations, as well as industry best practices for effective cybersecurity incident detection, response, and management.
SIEM, Security, SOC, Cyber Security, Insufficient logging, Live Log, Hot Log, Log Loss, Correlation.
Nura Muhammad Sani1, Ms. Reena S. Satpute2, 1MCA Sem – II Datta Meghe Institute of Higher Education & Research, Sawangi (M), 2Assistant Professor, School of Allied Sciences, Faculty of Science & Technology, Datta Meghe Institute of Higher Education & Research, Sawangi (M)
Heart disease, also known as cardiovascular disease, encompasses various conditions affecting the heart and blood vessels. It is a significant global health concern and a leading cause of death in many countries. These disorders impact heart function, blood flow, and overall health. In this study, a novel hybrid model is introduced that synergistically leverages the capabilities of Decision Tree (DT), Logistic Regression (LR), and Artificial Neural Network (ANN) algorithms to significantly enhance the accuracy of heart disease prediction. The hybrid model demonstrates remarkable performance, achieving an impressive accuracy of 88%, which outperforms the individual models - DT with 99% accuracy, LR with 80%, and ANN with 86%. Furthermore, the hybrid model showcases competitive precision, recall, and F1-score metrics, underscoring its potential as a robust tool for accurate and reliable heart disease prediction. This research establishes the efficacy of combining multiple algorithms to create a more effective predictive model in the domain of cardiovascular health assessment.
Decision Tree, Logistic Regression, Artificial Neural Network, hybrid model.
Foday Junior Conteh, Marymount University, Arlington, Virginia
Cloud computing (CC) represents a significant technological advancement in the United States (U.S.) healthcare. Despite its advantages like reduced costs, scalability, resource sharing, and high availability, CC raises concerns, especially in privacy and security. This study employs Grounded Theory methodology to delve into these concerns within cloud-based Healthcare Information Management Systems (HIMS) in the U.S., which operates under stringent patient privacy and security laws. The research focuses on healthcare organizations strategies to mitigate these challenges. In-depth interviews and document analysis, conducted using a qualitative research strategy, will gather data from healthcare professionals and Information Technology (I.T.) specialists interacting with cloud-based HIMS. Through thematic analysis and constant comparison, the research will construct a theoretical framework showcasing CCs impact on HIMS privacy and security. This framework will establish a basis for subsequent research to improve U.S. healthcare delivery by directing organizations to adopt and implement cloud-based HIMS compliant with U.S. data privacy regulations.
Cloud Computing, Privacy, Security, Health Information Management Systems (HIMS), Grounded Theory Methodology.
ALMONZER SALAH NOORALDAIM1 and ADIL ALI SAED2, 1Department of Computer Science, Faculty of Electronics & Informatics, Xian Jiatong University, China, 2Department of Computer Science, International University of Africa, Sudan
This review delves into the realm of artificial intelligence (AI) applications within dentistry, with a specific focus on the identification of teeth and caries. A thorough exploration was conducted across PubMed and the Institute of Electrical and Electronics Engineers (IEEE) Xplore databases, yielding 29 pertinent studies (17 focusing on caries detection and 12 on tooth detection). The studies incorporated diverse dental images, including panoramic, bitewing, periapical, intraoral radiographs, radiovisiography, and computed tomography. Panoramic images were the most utilized (n=8), followed by bitewing (n=6), periapical (n=5), and computed tomography. Various neural networks were employed to discern the targeted variables, with outcomes diverging notably based on the quality and characteristics of the input data. To further broaden the scope of AI applications in dental diagnostics, upcoming research endeavors should explore the integration of neural networks in different radiological studies, such as cone beam computed tomography (CBCT) or cephalometry.
Carries Detection, Tooth Detection, Artificial intelligence, Clinical Decision System.
Negin Sokhandan, Amazon, United States of America
In Virtual Product Placement (VPP) applications, the discrete integration of specific brand products into images or videos has emerged as a challenging yet important task. This paper introduces a novel three-stage fully-automated VPP system. In the first stage, a language-guided image segmentation model identifies optimal regions within images for product inpainting. In the second stage, Stable Diffusion (SD), fine-tuned with a few example product images, is used to inpaint the product into the previously identified candidate regions. The final stage introduces an ’Alignment Module’, which is designed to effectively sieve out low-quality images. Comprehensive experiments demonstrate that the Alignment Module ensures the presence of the intended product in every generated image, and enhances the average quality of images by 35%. The results presented in this paper demonstrate the effectiveness of the proposed VPP system, which holds significant potential for transforming the landscape of virtual advertising and marketing strategies.
ALMONZER SALAH NOORALDAIM1 and ADIL ALI SAED2, 1Department of Computer Science, Faculty of Electronics & Informatics, Xian Jiatong University, China, 2Department of Computer Science, International University of Africa, Sudan
This review delves into the realm of artificial intelligence (AI) applications within dentistry, with a specific focus on the identification of teeth and caries. A thorough exploration was conducted across PubMed and the Institute of Electrical and Electronics Engineers (IEEE) Xplore databases, yielding 29 pertinent studies (17 focusing on caries detection and 12 on tooth detection). The studies incorporated diverse dental images, including panoramic, bitewing, periapical, intraoral radiographs, radiovisiography, and computed tomography. Panoramic images were the most utilized (n=8), followed by bitewing (n=6), periapical (n=5), and computed tomography. Various neural networks were employed to discern the targeted variables, with outcomes diverging notably based on the quality and characteristics of the input data. To further broaden the scope of AI applications in dental diagnostics, upcoming research endeavors should explore the integration of neural networks in different radiological studies, such as cone beam computed tomography (CBCT) or cephalometry.
Carries Detection, Tooth Detection, Artificial intelligence, Clinical Decision System.
Cédric Maron1, 2, Virginie Fresse1, Karynn Morand2 and Freddy Havart2, 1Laboratoire Hubert Curien, 18 rue Professeur Benoît Lauras Bâtiment F, 42000 SaintEtienne, France, 2SEGULA Technologie, 1 Rue des Combats du 24 Août 1944, 69200 Vénissieux, France
With the growing interest in neural network compression, several methods aiming to improve the networks accuracy have emerged. One of them, data augmentation aims to enhance model robustness and generalization by increasing the diversity of the training dataset. Another one, knowledge distillation, aims to transfer knowledge from a network (teacher) to a network (student) during its training phase. Knowledge distillation is generally carried out using high-end GPUs because teacher network architectures are often heavy and not adapted to be implemented on the small resources present in the Edge. This makes the distillation process impossible to implement a pure Edge infrastructure. However, this paper proposes a new distillation method adapted to an edge computing infrastructure. By employing multiple monoclasse teachers of small sizes, the proposed distillation method becomes applicable even within the constrained computing resources of the edge. The method proposed is evaluated with classical knowledge distillation based on bigger teacher network, using different data augmentation methods and using different amount of training data.
Neural network compression, knowledge distillation, edge computing, data augmentation.
Rohit Khankhoje, Avon, Indiana, USA
In the ever-changing landscape of software testing, the incorporation of Artificial Intelligence (AI) represents a significant shift in thinking, especially in the context of chatbot systems. This paper examines the transformative influence of AI-Based Test Automation on ensuring the dependability, effectiveness, and intelligence of chatbots. The research investigates the synergy between AI techniques and automation strategies, highlighting their crucial role in improving the testing lifecycle of smart chatbot systems. By thoroughly examining AI-driven test case generation, scenario exploration, and adaptive testing, this study clarifies the significant advancements that AI contributes to the field of chatbot quality assurance. The paper offers valuable insights into the evolving sphere of conversational AI, providing a roadmap for professionals and researchers to leverage the potential of AI in enhancing the standards of chatbot testing methodologies.
Artificial Intelligent, Test Automation,Chatbot,AI-driven,Quality Assurance.
René Peinl and Johannes Wirth, Institute for Information Systems, Hof University of Applied Sciences, Hof, Germany
Large language models (LLMs) have garnered significant attention, but the definition of "large" lacks clarity. This paper focuses on medium-sized language models (MLMs), defined as having at least six billion parameters but less than 100 billion. The study evaluates MLMs regarding zero-shot generative question answering, which requires models to provide elaborate answers without external document retrieval. The paper introduces an own test dataset and presents results from human evaluation. Results show that combining the best answers from different MLMs yielded an overall correct answer rate of 82.7% which is better than the 60.9% of ChatGPT. The best MLM achieved 71.8% and has 33B parameters, which highlights the importance of using appropriate training data for fine-tuning rather than solely relying on the number of parameters. More fine-grained feedback should be used to further improve the quality of answers. The open source community is quickly closing the gap to the best commercial models.
question answering, language model, survey, benchmark.
Najah Alsubaie1, Tahani Alqahtani1, and Syarifah Bahiyah Rahayu2, 3, 1Department of Computer Sciences, College of Computer and Information Sciences, Princess Nourah bint Abdulrahman University (PNU), Saudi Arabia, 2Department of Defence Science, Faculty of Defence Science and Technology, National Defense, 50603 Kuala Lumpur, Malaysia, 3Cyber Security and Digital Industrial Revolution Centre, National Defense, University of Malaysia, 50603 Kuala Lumpur, Malaysia
Pneumonia, a lung inflammation and consolidation disorder, poses diagnostic challenges necessitating accurate detection. This paper introduces an innovative automated approach using segmented lung morphology and texture attributes from Chest X-ray (CXR) images. Unlike conventional methods analyzing the entire CXR, our focus narrows to segmented lung regions. Discriminative ranking of extracted features enhances the categorization of CXR images into pneumonia and normal cases. Diverse machine learning classifiers are evaluated, yielding a compelling 86% accuracy—validating our method’s efficacy in distinguishing pneumonia from normal cases. This study offers a robust and efficient diagnostic avenue for improved pneumonia differentiation.
lung, chest X-ray, machine learning, pneumonia, morphology, texture.
Organizations have to plan on migrating to quantum-resilient cryptographic measures, also known as PQC. However, this is a difficult task, and to the best of our knowledge, there is no generalized approach to manage such a complex migration for cryptography used in IT systems that explicitly integrates into organizations’ steering mechanisms and control systems. We present PMMP, a risk-based process for managing the migration of organizations from classic cryptography to PQC and establishing crypto agility. Having completed the initial design phase, as well as a theoretical evaluation, we now intend to promote PMMP. Practitioners are encouraged to join the effort in order to enable a comprehensive practical evaluation and further development.
Post-Quantum Cryptography (PQC), PQC Migration Management Process (PMMP), Crypto Agility.
David Noever and Samantha Elizabeth Miller Noever, PeopleTec, 4901-D Corporate Drive, Huntsville, AL, USA, 35805
This study explores the capabilities of multimodal large language models (LLMs) in handling challenging multistep tasks that integrate language and vision, focusing on model steerability, composability, and the application of long-term memory and context understanding. The problem addressed is the LLMs ability (Nov 2023 GPT-4 Vision Preview) to manage tasks that require synthesizing visual and textual information, especially where stepwise instructions and sequential logic are paramount. The research presents a series of 14 creatively and constructively diverse tasks, ranging from AI Lego Designing to AI Satellite Image Analysis, designed to test the limits of current LLMs in contexts that previously proved difficult without extensive memory and contextual understanding. Key findings from evaluating 800 guided dialogs include notable disparities in task completion difficulty. For instance, Image to Ingredient AI Bartender (Low difficulty) contrasted sharply with AI Game Self-Player (High difficulty), highlighting the LLMs varying proficiency in processing complex visual data and generating coherent instructions. Tasks such as AI Genetic Programmer and AI Negotiator showed high completion difficulty, emphasizing challenges in maintaining context over multiple steps. The results underscore the importance of developing LLMs that combine long-term memory and contextual awareness to mimic human-like thought processes in complex problem-solving scenarios.
Large language models, creativity, steerability, composability, dataset.
Saud Althabiti1,2,3, Mohammad Ammar Alsalka1, and Eric Atwell1, 1School of Computing, University of Leeds, Leeds, United Kingdom, 3Faculty of Computing and Information Technology, King Abdulaziz University, Jeddah, Saudi Arabia
This paper introduces TaKeed, an explainable Arabic automatic fact-checking system with justifications. While existing research often focuses on classifying claims as "True" or "False," there is limited exploration of generating explanations for claim credibility, particularly in Arabic. TaKeed addresses this gap by assessing claim truthfulness based on the retrieved information, utilizing two main components: information retrieval and LLM-based claim verification. We collected a testing gold-labelled dataset with manually justified references to evaluate the system. The initial model achieved a promising F1 score of 0.72 in the classification task. The systems generated explanations are compared with gold-standard explanations syntactically and semantically, we recommend evaluating using semantic similarities, resulting in an average cosine similarity score of 0.76. Additionally, we explored the impact of varying snippet quantities on tweet classification accuracy, revealing a potential correlation, with the model using the top seven hits outperforming others with an F1 score of 0.78. .
TaKeed, Generative Fact-checker, Arabic Fact-Checking, Claims Verification, LLMs-based FactChecking
Cléber Viana
This paper presents a novel approach to school safety, integrating the Internet of Things (IoT), artificial intelligence (AI), and cloud computing to enhance detection and response systems in educational settings subtly. Our system uses IoT and AI to detect unusual activities and environmental changes, focusing on non-intrusive monitoring to maintain a supportive atmosphere. The cloud computing component ensures efficient data processing and real-time response coordination. We prioritize ethical technology use, upholding data privacy and personal integrity. The preventive approach fosters a safe and supportive environment rather than enforcing control. The paper discusses the technical framework, implementation challenges, and case studies demonstrating effectiveness in real-world scenarios. Our model offers a balanced solution for enhancing school safety while maintaining a positive educational atmosphere.
Educational Safety, IoT in Schools, AI-Powered Detection, Cloud Computing in Education, Connected Technology Implementation.
Copyright © SIGML 2024