Nia Poor, Jaime Campanelli, and Daryl Johnson, Department of Cybersecurity, Rochester Institute of Technology, Rochester, United States
This paper proposes a novel covert communication channel within the popular social media network Reddit. The usage of this covert channel is based on the prerequisite of using the older (but still popular) version of Reddit, ”Old Reddit”, to enable sharing of a user’s upvotes publicly. The channel upvotes Reddit posts in subreddits that begin with characters corresponding to a secret message. By upvoting ”hot” posts within top subreddit communities, and incorporating random timing between upvotes, we demonstrate the feasibility of transmitting secret information without arousing suspicion.
Reddit, Old Reddit, covert communications, encoding, decoding.
Eduardo F. Damasceno, Ricardo L. Pires and Armando P. da Silva, Federal University of Technology of Parana, Brazil
Handwriting development is a multifaceted process involving fine motor control, perception, and visualmotor integration skills, crucially developed during the literacy phase. This text emphasizes the role of psychopedagogical professionals in monitoring the psychogenesis of written language. Early childhood educators play a pivotal role in initiating literacy acquisition, necessitating continuous training to enhance their pedagogical practices. While the literacy environment significantly influences knowledge acquisition by fostering student interaction and curiosity, challenges such as school organization and imposed goals can divert educators focus from identifying learning evolution or failure due to cognitive disorders. Recognizing the im-portance of literacy, educators must undergo ongoing training to create conducive literacy environments for comprehensive student development. The daunting task of teaching many students to read and write requires educators to consider variations in language writing, incorporating psychological elements and cultural backgrounds. This work introduces a tool leveraging artificial intelligence and machine learning algorithms to assist educators in monitoring and quickly identifying the stage of language writing formation in 6 to 7-year-old beginners. The prototype aids early childhood educators in categorizing and identifying each childs stage while detecting potential motor or cognitive deficits.
Handwriting Development, Language Writing Formation, Machine Learning.
Prof. Mohamed M. Ghoneim Sywelem, Department of Educational Foundations – Suez University – Suez City – Egypt
The rapid advancement of artificial intelligence (AI) technologies has led to their increasing integration in educational settings. This paper offers an overview of the ethical considerations inherent in this integration. Drawing upon existing literature and ethical frameworks, the paper examines key ethical issues arising from AI adoption in education, including AI-driven decision-making, student data privacy, and algorithmic bias. It also explores the impact of AI on teaching and learning processes. Furthermore, the paper discusses various ethical frameworks and guidelines aimed at addressing these concerns and promoting responsible AI use in education. Through the use of case studies and examples, the paper illustrates real-world ethical challenges faced by educators, institutions, and policymakers. Finally, it provides recommendations for fostering ethical AI integration in education, emphasizing the importance of ethical awareness and proactive measures to ensure the ethical use of AI technologies in educational settings. This overview serves as a foundational resource for educators, policymakers, researchers, and stakeholders involved in navigating the complex intersection of AI and ethics in education.
Ethical Considerations - Artificial Intelligence - Education.
Feng-Hsu Wang, Department of Computer Science and Information Engineering, Ming Chuan University, Taoyuan, Taiwan
AI-based learning technologies, especially deep learning, hold significant promise for enhancing students’ learning experiences in educational systems. However, the application of deep learning in education faces challenges related to transparency and accountability of AI systems, necessitating the development of Explainable Artificial Intelligence (XAI). This study aims to investigate the efficacy of an eXplainable Artificial Intelligence (XAI) technique within the realm of programming education. Specifically, it examines the application of Gradient Integration, a method utilized to generate code segments associated with errors in object-oriented programming. This is achieved through the analysis of a Performer-based deep learning classification model. Thirty-six participants took part in the experiment assessing students’ trust in the system, their cognitive load, and learning performance. The controlled experiments revealed a significant difference in reducing learners cognitive load and improved exam scores, indicating the positive impact of XAI. This study contributes to the ongoing exploration of the efficacy of XAI in educational settings.
Explainable Artificial Intelligence, Deep Learning Technology, Human-Computer Collaborative Learning, Programming Education .
Karla Karina Ruiz Mendoza and Sooriyan Aliyoglu, University Autonomus of Baca California, MexicoCastle University, New City, Cyprus
This study explores the integration of generative artificial intelligence (AI gen), especially models such as ChatGPT, into teacher training, highlighting both the promises and challenges of adopting advanced technologies in an educational context. Through a course titled "Introduction to Generative Artificial Intelligence for Teachers," 97 educators from various universities in Baja California participated, spanning a wide range of ages. The qualitative methodology adopted allowed an in-depth exploration of teachers perceptions, experiences, and expectations regarding AI in education. Results showed a generally positive evaluation of the course, with significant emphasis on the importance of AI in transforming educational practices. Approximately 41.38% of the comments highlighted the potential of gen AI to support and enhance teaching and learning. Additionally, there was a clear interest in deepening knowledge about AI, as well as a need for ongoing training strategies. However, the study also emphasizes critical reflections on the ethical and practical challenges of integrating AI into education, underscoring the importance of a reflective and ethical approach. The demand for gen AI training by educational institutions indicates a global trend toward the adoption of these technologies. The study concludes with recognition of the potential of AI to enrich pedagogy, provided that the associated risks are considered, and ethical and effective adoption is promoted.
Generative Artificial Intelligence, Education, Teacher Training, Ethical Reflections, Technological Adoption.
Howard Greisdorf, Ph.D., Research Advisor, PsyOptik, LLC
The last two decades of advances in information technologies and systems have also produced ever-increasing attempts to penetrate or corrupt those entities for profitable gain. The criminals, their attacks, and the defenses have subsequently entered the research literature tagged as social engineers, social engineering, and cyber security respectively. This work seeks to clarify and model the process by which perpetrators design their attacks and the victim characteristics that make them vulnerable to those attacks. The supporting studies suggest that any attempt to curtail cyber intrusions should include a training element focused on user personality traits that make them susceptible to criminal intentions and methodologies. The conclusion points to greater user self-awareness as a key component of any cyber security training.
Social engineering, human factors, persuasion, personality traits.
Mohamed Abdelmalek1, Anis Harhoura1, Issam Elaloui1, Mahdi Madani2a andEl-Bay Bourennane2, 1Faculty of Science and Technology, University of Burgundy, Dijon, France, 2ImViA Laboratory (EA 7535), University of Burgundy, 21000 Dijon, France
This paper proposes a model based on Convolutional Neural Network (CNN) for image encryption and decryption. We used the CIFAR-10 dataset containing 60.000 color images of size 32×32 across ten classes to train and test the proposed network. Our focus lies in designing a convolutional autoencoder for image compression and reconstruction, utilizing an encoder-decoder architecture. During training, the autoencoder learns to encode essential image features into a reduced-dimensional latent space and reconstructs the image from this space. The implementation of the proposed encryption model demonstrates efficacy in preserving data integrity while reducing dimensionality. Experimental results show that the used CNN exhibits a proficient encryption process and acceptable decryption process.
Visually image protection, Deep Learning, Encryption, Encoder, Data security, Image Compression.
Oluwabunmi Falebita1 and Oluwafemi Famakinde2, 1Innovation and Technology Policy Department, 2Social Policy Department Nigeria Institute of Social and Economic Research (NISER), Ibadan, Nigeria
This qualitative cross-sectional survey delves into the nuanced perceptions surrounding data privacy practices in the realm of digital credit in Nigeria. Through in-depth interviews (IDI) with Digital Credit Users (DCUs) across various economic hubs in Nigeria, we explore their attitudes and concerns regarding the sensitivity of personal information and their willingness to divulge it to Digital Credit Providers (DCPs). Employing a multi-stage sampling technique, clusters representing Nigerias six zones were purposively selected, with the South-West zone chosen for its economic significance. In this zone, Lagos, Oyo, and Ogun States were further sampled based on economic activity, with 40 DCUs interviewed per state, totalling 120 DCUs. Local Government Areas (LGAs) within these states were selected based on Central Business District (CBD), urban, rural, and peri-urban criteria. Thematic analysis of interview transcripts using NVIVO 14 software unveiled key findings indicating that DCUs’ Bank Verification Number (BVN), National Identification Number (NIN), and debit card details are perceived as the most sensitive information. They expressed a high level of obligation to disclose information to DCPs and identified perceived risks such as financial loss, data breaches, and unwanted contact. Additionally, DCUs exhibited a strong preference for retaining control over their information, with many expressing a reluctance to proceed with digital credit applications if privacy breaches were anticipated. These findings shed light on the complex interplay between data privacy perceptions, risk assessment, and individual autonomy in the digital credit landscape.
Digital Credits, Digital Data Privacy, Digital Credit Providers & Digital Credit Users, Nigeria.
AtheerAlmoamary,Raghad Alsuwailem, Ruba Altawil, Samiah Alanazi,and Ala Alluhaidan, Department of Information Systems, College of Computer and Information Sciences,Princess Nourahbint Abdulrahman University, Riyadh, Saudi Arabia
Objective: to create an environment accessible through any device, that assesses in the ease of the journey of living with cancer disease. Methods: For the development of the project, we followed the iterative waterfall method. A simple SDLC method that allows for continuous feedback and adjustments. Results: This idea was designed to take advantage of technology and help people with cancer or those recovering from it, their symptoms need to be tracked by a doctor and reminded of appointments, medications, or chemotherapy sessions, It is expected to advance patient adherence to their overall wellbeing, and motivate them during the treatment journey. Conclusion: “Cancer Ease” has been designed to empower individuals dealing with cancer by providing a user-friendly interface for information, personalized health tracking, easily accessible, and ultimately striving to enhance the quality of life throughout their journey with the disease.
PLWC, Cancer, Mental Symptoms, Physical Symptoms.
Maya Trutschl1, Urska Cvek2, Andrea L. Brumley3, Steven A. Conrad3, and Marjan Trutschl2, 1Caddo Parish Magnet High School, Shreveport, USA 2LSU Shreveport, Department of Computer Science, Shreveport, USA 3LSU Health Sciences Center, Department of Emergency Medicine and Department of General Surgery, Shreveport, USA
Pressure ulcers, or pressure injuries, are localized areas of skin and/or underlying tissue necrosis that typically occur over bony prominences due to a prolonged pressure or friction. They can lead to serious morbidity and mortality, emphasizing the need for prevention. This project utilizes a database of demographic and clinical features of a large patient data set and applies machine learning to determine the higher risk patients, coupled with a complimentary device to assist in the prevention of pressure injuries. A complementary monitoring system is built based on an inertial wearable utilizing an inexpensive microcontroller and a gyroscope. The combined approach is evaluated, fine-tuned, and assessed based on different performance metrics.
Pressure ulcers, Machine Learning,Inertial Wearable
Indra Kumari and Minho Lee, Department of Machine Learning Data Research, Applied AI, Korea Institute of Science and Technology Information (KISTI), Korea National University of Science and Technology (UST), Daejeon, 34113, South Korea
Advanced Persistent Threats (APTs) pose a significant challenge for gov ernments, businesses, and organizations due to their dynamic nature. Identifying the exact type of attack is difficult, making traditional detection methods ineffective. While the MITRE ATT&CK framework provides valuable information on APT groups, it lacks real-world network traffic and system logs. This work addresses this limitation by leveraging DAPT 2020, a dataset mimicking real-world APT attacks. The aim is to detect APT attack phases through a deep learning-based framework, ultimately map ping them to the MITRE ATT&CK framework for better prevention. Traditional ma chine learning techniques are insufficient for this complex task, prompting the use of intelligent optimization and deep learning-based classification methods. The proposed framework consists of four modules: preprocessing, feature extraction, optimization, and classification. Preprocessing involves data cleaning and normalization to improve quality. Relevant features are then extracted using the Pearson Correlation Coefficient. Feature selection with Principal Component Analysis optimizes the data, reducing pro cessing time and boosting classification accuracy. Finally, a combined Recurrent Neu ral Network (RNN) with Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) units classifies the attack phases. This framework outperforms both deep learn ing and machine learning baselines, offering a promising solution for early APT detec tion and prevention, ultimately safeguarding organizations.
Advanced Persistent Threat (APT); MITRE ATT&CK; DAPT2020 dataset; Pearson Correlation Coefficient (PCC); Principal Components Analysis (PCA); and RNN-LSTM-GRU.
