IFIP Open Access Digital Library
Not a member yet
22614 research outputs found
Sort by
Revelatory Insights into Parkinson’s: Hand Gestures Deciphering with Mobilenet SSD
No full textPart 1: SDG 3 Good Health and Well-BeingInternational audienceHand gestures have evolved into a natural and intuitive means of engaging with technology. Parkinson’s disease is a neurological ailment that worsens with time. Sophisticated deep learning-based system for the precise identification of Parkinson’s Disorder is developed through the interpretation and analysing of hand gestures exhibited by affected individuals. Relevant features are extracted from the images, employing computer vision techniques, enabling the analysis of hand movements. A deep learning model is trained to recognize distinctive motor symptoms associated with the disorder. Object detection algorithms are employed to pinpoint specific deviations in finger and hand positions, thereby quantifying tremors, bradykinesia, and other relevant motor symptoms. The approach lies in the utilisation of a dataset comprising 40 images for each Parkinson’s hand gesture, enabling robust model training. The Mobilenet SSD algorithm for object detection to find the specific deviations in finger and hand positions. The model exhibits a high level of accuracy in distinguishing between individuals with the possibility of Parkinson’s disorder and those without, based on the subtle motor symptoms exhibited in their hand gestures
Smart Armband with Fall Detection in Elderly
No full textPart 4: SDG 11 Sustainable Cities and CommunitiesInternational audienceThe paper aims to create an affordable wearable fall detection system aimed at accurately identifying falls. The system comprises a smart armband prototype equipped with embedded sensors to capture motion data from elderly individuals during their daily activities. The primary objectives include designing a compact hardware prototype incorporating an accelerometer, and gyroscope, with a microcontroller unity, implementing fall detection algorithms utilizing threshold-based techniques and machine learning models for real-time classification of falls versus other activities, and optimizing the algorithms for on-device execution with limited computing resources. The ultimate goal is to develop a non-invasive, cost-effective solution for elderly fall monitoring that is both accurate and robust. The prototype’s development will encompass hardware design and optimization of the fall detection algorithm for on-device inference, with future research potentially integrating this system into assistive technologies to enhance safety for independently living elderly individuals
The Perceived Barriers to the Adoption of Blockchain in Addressing Cybersecurity Concerns in the Financial Services Industry
No full textPart 3: Technical Attacks and DefensesInternational audienceThe financial services industry has witnessed a notable surge in cyber-related crimes, necessitating the adoption of more robust cybersecurity measures. Current literature has identified that the adoption of blockchain can enhance the efficiency and integrity of cybersecurity mechanisms. Despite blockchain’s acknowledged cybersecurity capabilities in existing literature, the adoption rate of blockchain in the South African financial services industry has been relatively slow due to barriers associated with adopting a novel technology. This study interviewed 11 financial technology experts to gain insights into the perceived barriers to the adoption of blockchain technology to address cybersecurity concerns in the South African financial services industry. This research followed a deductive approach by combining constructs from the Innovation Resistance Theory and Technology-Organisation Framework to investigate the perceived barriers to the adoption of blockchain in addressing cybersecurity concerns in the financial services industry. The findings revealed that the adoption of blockchain in addressing cybersecurity in the South African financial services industry is influenced by the four main barriers: functional organisational-level, functional industry-level, psychological organisation-level, and psychological industry-level barriers. The functional barriers were influenced by the perceived value barrier, risk barrier, usage barrier, technology barrier, and regulatory environment barrier. The psychological barriers affecting blockchain adoption were influenced by the perceived image barrier and tradition barrier. Two additional barriers were found to inhibit the adoption of blockchain in addressing cybersecurity, namely, the use case barrier as a functional industry-level barrier and the knowledge barrier as a psychological organisation barrier
Defining Measures of Effect for Disinformation Attacks
No full textPart 2: Social EngineeringInternational audienceThis study proposes three measures for assessing the survival of beliefs in a population subjected to a disinformation attack. The intent of these three measures was to simplify the task of assessing damage effects arising from disinformation attacks, and provide a means for comparing the relative effectiveness of alternate defensive or damage mitigation strategies. To define these measures and to bound the measures problem, disinformation attacks are characterised, disinformation effects and propagation behaviours are surveyed and summarised. Nine attributes are identified spanning scalability relative to a population and disinformation attack, propagation media independence, target attributes, media propagation attributes, effects of uncertainty, use of established models, probabilistic measures, and measurement methods. The three proposed measures were critically assessed against these nine desirable attributes. The three proposed measures are capable of capturing the aggregated effects of a disinformation attack, exposure effects produced by propagation through channels such as digital media, and the direct effects against the individuals or population being subjected to an attack. The separation of exposure and cognitive effects makes these measures suitable for use in defensive or damage mitigation strategies that include measures against disinformation propagation, and measures to increase individual or population resistance to disinformation
Assessing the Cybersecurity Needs and Experiences of Disabled Users
No full textPart 4: Usable SecurityInternational audienceDigital technology is incredibly crucial in today’s world. The use of technology is considered a right for both able and disabled users. Accessibility and security are two important concepts in the technology context. Accessibility refers to the level to which a product or service is designed to be utilized by people with disabilities. While security focuses on protecting a product or service from threats and harm. Accessible security refers to the practice of ensuring that digital products and services are not only secure but also accessible to everyone, including people with disabilities. Numerous studies have been conducted on the usage of technologies among people with disabilities. However, little research has been undertaken on accessible cybersecurity. Understanding encounters of disabled individuals with cybersecurity challenges can help develop more accessible and secure technologies and improve user experience. The first step to improving the accessibility of cybersecurity safeguards for users with disabilities is assessing their attitudes and needs. The aim of the study is to explore the cybersecurity attitude, behavior and awareness of people with various types of disability. The survey used to determine the most significant gap for people with disabilities in the accessible cybersecurity context to help them better handle and understand cyber threats in their everyday lives. The survey findings point out that having cybersecurity awareness does not always result in preventing security breaches. There is a gap between theoretical knowledge and practical application. There is a notable concern regarding insufficient technological safeguards. Recommendations are included for software developers to create a more accessible and secure digital environment
Hybrid Efficient IDS Against Adversarial Attacks in IoT Networks
No full textPart 1: Applications of AI/ML in KDM, Cloud Computing & SecurityInternational audienceIoT services become more dominant as time passes and the growing security concerns become less relevant. Due to proliferating heterogeneity, emerging technologies, and resource-constrained IoT systems, intelligent systems are becoming more vulnerable to cyberattacks. As a result, practical solutions to security issues like privacy, scalability, authenticity, trust, and centralization are required. Since malicious actors frequently use obfuscation tactics to elude detection, traditional approaches to intrusion detection systems are no longer relevant. Furthermore, these methods fail to detect zero-day Attacks. Applying an intelligent mechanism based on machine learning or deep learning at a different stage is necessary to detect attacks. The purpose of selecting multiple algorithms is to satisfy the specific requirements of various users or groups. In other words, it is crucial to identify the most effective model for each sort of user. The suggested method consists of two stages: a signature-based IDS with a malicious signature already developed and an ML/DL algorithm trained using the IoT-23 public dataset. Results show that the proposed engine works better than existing cutting-edge methods, with an average accuracy of 95.3%
SC-EcapaTdnn: ECAPA-TDNN with Separable Convolutional for Speaker Recognition
No full textPart 5: Perceptual IntelligenceInternational audienceThe efficacy of time-delay neural networks (TDNN) in speaker recognition has been demonstrated. ECAPA-TDNN builds on TDNN, improving performance levels at the cost of increased computational complexity and slower inference speed. However, the effectiveness of ECAPA-TDNN does not meet the expected standards for speaker recognition in complex scenarios. This motivates us to seek an architecture superior to ECAPA-TDNN. In this paper, we propose an efficient network called SC-EcapaTdnn, which is a fusion of separable convolutions and ECAPA-TDNN. This innovative design uses ECAPA-TDNN as the backbone, uses depth-separable convolution blocks to encode acoustic features, and generates high-resolution frequency feature maps, allowing the backbone model to obtain more refined and effective speaker features. At the same time, we replace the squeeze excitation (SE) module in ECAPA-TDNN with adaptive one-dimensional convolution to generate channel attention weights to extract inter-channel dependencies. Ultimately, we trade a small increase in model parameters for a significant increase in performance. Training on the AISHELL and CN-Celeb datasets shows that our proposed architecture outperforms other mainstream speaker recognition systems
Cascaded Sliding-Window-Based Relativistic GAN Fusion for Perceptual and Consistent Video Super-Resolution
No full textPart 5: Perceptual IntelligenceInternational audiencePerceptual video super-resolution aims at converting low-resolution videos to visually appealing high-resolution ones. It may lead to temporal inconsistency due to the drastically changing outputs. In this paper, we propose cascaded sliding-window-based relativistic GAN (Generative Adversarial Network) fusion for perceptual and consistent video super-resolution (PC-VSR). Firstly, cascaded sliding-window-based relativistic GAN is designed to extract more useful information. It enlarges the temporal receptive field of sliding-window-based model in each step. It is able to enhance perceptual quality and compensate temporal consistency progressively and sufficiently. The trained separate refinement generator networks are fused into a final refinement generator. The final refinement generator can be calculated recursively at the testing stage. With our generator fusion, the parameter number is reduced and good quality is maintained. Extensive experimental results demonstrate that our approach outperforms state-of-the-art super-resolution methods in terms of perceptual quality. Our method also achieves good temporal consistency and per-pixel accuracy, compared with other perceptual approaches
Chat Bot in Banking Sector Using Machine Learning and Natural Language Processing
No full textPart 1: Applications of AI/ML in Natural Language ProcessingInternational audienceThe demand for the effective customer services in banking industry has increased day by day and requirement of perfect solutions for the customer query also increased in these modern days. This research paper says about the design, development, various kind of chat bots present in now a days and how these chat bots are implemented and in which way it works according to customer queries. Chatbots are actually created to give accurate answers for banking inquires. The chat bot provides a user friendly interface for customers, they can ask information about their account balance, transaction history, fund transfer, loan application and other common banking tasks. The research methodology includes methods for how to access a normal chat bot in easy steps and various chatbots present now a days in banking industry and training of chatbot to understand and respond to diverse customer queries. The chat bot has advance machine learning algorithms and natural processing language to adapt for different queries and improve its performance in future and ensuring the perfect accuracy level of the answers. The system is combined with bank database so customers can see their bank details, ensuring real time access for their account information, secure transaction. Our research findings reveals that after coming of chat bot in banking sector the human customer services has been decreased and customer satisfaction has increased. After many testing of this model and according to user feedback the chat bot provides accurate solutions for their quires except in some situations. This research contains about growing field of artificial intelligence customer service solutions and potential benefits of implementing a banking chat bot to address common customer services ultimately leading to improved customer engagement and loyalty
Evaluating the Language Translation Accuracy of GPT-3.5 Using Prompt Engineering
No full textPart 1: Applications of AI/ML in Natural Language ProcessingInternational audienceInvestigation and optimization of prompt design strategies can unlock the full potential of GPT-3.5 and other LLMs for high-fidelity language translation, furthering the boundaries of machine intelligence and its impact on global communication. This study investigates the effectiveness of prompt engineering in improving the language translation accuracy of GPT-3.5. We propose various prompts designed to guide GPT-3.5 towards generating more accurate translations. We consider the translations from English to Tamil, Tamil to English, English to Hindi, and Hindi to English. The performance is evaluated using the Bilingual Evaluation Understudy (BLEU) and Metric for Evaluation of Translation with Explicit ORdering (METEOR) scores. Our analysis reveals that carefully crafted prompts can significantly enhance the quality of GPT-3.5 translations, demonstrating the promising potential of this approach for boosting machine translation performance