
The availability of medical devices such as glucose level and blood pressure measuring devices is continuously increasing. These devices have gained user interest as they are easy to use. However, medical devices introduce extra complexity to the network by being numerous, heterogeneous, and more vulnerable to cyber-attacks. For better network management and overall network security, it is important to understand the network traffic characteristics of the devices. Thus, in this paper, we analyze in detail the traffic characteristics of 8 medical devices both at the device level and at the level of individual functionality of each device. We collect and share both network and bluetooth traffic from a total of 51 functionalities of the devices. Our analysis includes different metrics such as protocols, amount of incoming/outgoing traffic, DNS queries, and analysis of traffic destinations. We observed that devices have unique network and bluetooth traffic characteristics, that might be useful in developing networking tools for medical devices.
Alzheimer's Disease and Related Dementias (ADRD) patients and older adults face challenges related to memory loss, navigation difficulties, and social isolation, impacting their daily tasks, appointments, and social connections. To address these multifaceted challenges, this paper presents the design, development, and preliminary evaluation of “CareCompanion,” a virtual assistant tailored specifically for this population. Leveraging advanced AI technologies such as natural language processing, machine learning, and knowledge graphs, CareCompanion provides personalized reminders, navigation assistance, and social connectivity features. Preliminary evaluation results demonstrate the potential of CareCompanion in improving the quality of life, independence, and social engagement for ADRD patients and older adults. Further research and development can enhance its effectiveness, usability, and customization, catering to the unique needs of this population, while fostering connections and mitigating the impact of memory loss, navigation difficulties, and social isolation.
Agriculture has a huge benefit on the global landscape due to rapid population growth and the subsequent increase in demand for food. Hence, there is an urgent need to enhance crop productivity. One of the primary factors contributing to low crop productivity is the spread of diseases caused by bacteria, spores, fungi, and viruses. However, this problem can be mitigated by applying plant disease detection methods, Recently, the field of machine learning and deep learning has achieved remarkable results when combined with IoT -based technology, especially in the field of plant disease recognition. This scientific article aims to clarify the prevailing methodologies used in IoT -based smart agriculture, with a clear focus on the use of IoT and machine learning. In this context, the manuscript presents an innovative model for plant disease detection and recognition, drawing on the fields of machine learning and deep learning, ultimately leading to increased accuracy and efficiency. In essence, this article provides a comprehensive review and overview of various machine learning and deep learning techniques used in plant disease detection, using an IoT approach.
Wildfires in California have consistently been a concern that has been getting more acute in the past years. Rampant wildfires have consistently hit the state of California, creating severe economic and environmental loss. In 2023, wildfires cost nearly 1.2 billion U.S. dollars in financial loss between January and September. Between 2021-2022, wildfires accounted for over 11.2 billion in damage across the United States. Over 1.6 million acres of land have burned and caused large sums of environmental damage. The increasing frequency and severity of wildfires in California have led to a growing need for accurate and reliable wildfire risk assessments. In this research, we propose a machine learning approach based on six different classifiers to determine wildfire risk using environmental data from California. We use a dataset of historical wildfire occurrences and various environmental variables such as temperature, humidity, and wind speed to build a recommendation model using a Random Forest classifier. We use the SMOTE technique to handle class imbalance.
Smart home automation, particularly in lighting, holds the potential to significantly improve comfort, energy efficiency, and security by centralizing control over internet-of-things (IoT) devices. This paper introduces a smart lighting automation system that analyzes human movement trajectories using machine learning, deep learning, and reinforcement learning techniques integrated into the Home Assistant platform. In particular, we introduce a transformer-based deep neural network architecture with reward-based tuning as our backbone model. The system predicts the user's next location and adjusts the lighting accordingly based on the anticipated movement trajectories derived from data collected by IoT devices distributed throughout a residential area. This enhances both convenience and energy efficiency. We deployed the system in a residential setting and conducted experiments to validate its accuracy.