To assist policymakers in making adequate decisions to stop the spread of the COVID-19 pandemic, accurate forecasting of the disease propagation is of paramount importance. This paper presents a deep learning approach to forecast the cumulative number of COVID-19 cases using bidirectional Long Short-Term Memory (Bi-LSTM) network applied to multivariate time series. Unlike other forecasting techniques, our proposed approach first groups the countries having similar demographic and socioeconomic aspects and health sector indicators using K-means clustering algorithm. The cumulative case data of the clustered countries enriched with data related to the lockdown measures are fed to the bidirectional LSTM to train the forecasting model. We validate the effectiveness of the proposed approach by studying the disease outbreak in Qatar and the proposed model prediction from December 1st until December 31st, 2020. The quantitative evaluation shows that the proposed technique outperforms state-of-art forecasting approaches.

Glucose monitoring is key to the management of diabetes mellitus to maintain optimal glucose control whilst avoiding hypoglycemia. Non-invasive continuous glucose monitoring techniques have evolved considerably to replace finger prick testing, but still require sensor insertion. Physiological variables, such as heart rate and pulse pressure, change with blood glucose, especially during hypoglycemia, and could be used to predict hypoglycemia. To validate this approach, clinical studies that contemporaneously acquire physiological and continuous glucose variables are required. In this work, we provide insights from a clinical study undertaken to study the relationship between physiological variables obtained from a number of wearables and glucose levels. The clinical study included three screening tests to assess neuropathy and acquired data using wearable devices from 60 participants for four days. We highlight the challenges and provide recommendations to mitigate issues that may impact the validity of data capture to enable a valid interpretation of the outcomes.

In this paper, we propose a novel session-based continuous authentication model using photoplethysmography (PPG). Unlike previous PPG-based authentication techniques that generate user signatures only during the initial interaction, our session-based approach tackles inter session PPG drifting by generating a user signature at the start of each session. Our model is composed by two modules: Firstly, heavy deep autoencoders (AE) are utilized for feature extraction and, secondly, a lightweight Local Outlier Factor (LOF) is employed for user authentication. Additionally, we introduce a continuous updating system for the LOF model, which automatically recovers from security breaches and can enhance authentication accuracy by more than 9%. Our experiments show that in a single-session scenario, our model achieves authentication accuracies of 93.5% and 91.8% on the CapnoBase and BIMDC benchmarking datasets, respectively, outperforming the state-of-the-art baseline model by 3.2% and 1.6% on both datasets, respectively. In multiple-session scenarios, our scheme attains an authentication accuracy of 95% when tested on the BioSec2 dataset, effectively mitigating inter-session PPG drifting and achieving an advantage of more than 8.5% in authentication accuracy over the state-of-the-art method. In terms of execution speed, our solution is seven times faster at runtime compared to competing state-of-the-art solutions.

Machine learning and deep learning models play a crucial role in optimizing smart grid operations. While offering valuable utility, concerns arise regarding their exposure of sensitive information. Previous studies have focused on optimizing load forecasting models, such as Long Short Term Memory (LSTM), overlooking their associated risks of revealing private information about the consumer. Existing privacy preservation studies have focused mainly on addressing risks associated with classification models, but similar concerns may exist in their load forecasting counterparts. In this paper, we investigate potential data leakage from deep learning-based load forecasting models. Our study analyzes the ability of personalized forecasting models to leak global properties about the consumer. We present a novel property inference attack tailored for forecasting models that can be executed in a black box setting, with only query access to the target model. We demonstrate that black box access to an LSTM model can reveal sensitive information about the user, with the attack achieving results comparable to having direct access to the data (with the difference being as low as 1% in Area Under the ROC Curve). This highlights the necessity of securing access to forecasting models with the same level of diligence as the data itself.