In recent years, gradient boosted decision tree learning has proven to be an effective method of training robust models. Moreover, collaborative learning among multiple parties has the potential to greatly benefit all parties involved, but organizations have also encountered obstacles in sharing sensitive data due to business, regulatory, and liability concerns. We propose Secure XGBoost, a privacy-preserving system that enables multiparty training and inference of XGBoost models. Secure XGBoost protects the privacy of each party's data as well as the integrity of the computation with the help of hardware enclaves. Crucially, Secure XGBoost augments the security of the enclaves using novel data-oblivious algorithms that prevent access side-channel attacks on enclaves induced via access pattern leakage.
Thermal videos provide a privacy-preserving yet information-rich data source for remote health monitoring, especially for respiration rate (RR) estimation. This paper introduces an end-to-end deep learning approach to RR measurement using thermal video data. A detection transformer (DeTr) first finds the subject’s facial region of interest in each thermal frame. A respiratory signal is estimated from a dynamically cropped thermal video using 3D convolutional neural networks and bi-directional long short-term memory stages. To account for the expected phase shift between the respiration measured using a respiratory effort belt vs. a facial video, a novel loss function based on negative maximum cross-correlation and absolute frequency peak difference was introduced. Thermal recordings from 22 subjects, with simultaneous gold standard respiratory effort measurements, were studied while sitting or standing, both with and without a face mask. The RR estimation results showed that our proposed method outperformed existing models, achieving an error of only 1.6 breaths per minute across the four conditions. The proposed method sets a new State-of-the-Art for RR estimation accuracy, while still permitting real-time RR estimation.
27 July 2024 in Journal of Computer Assisted Learning
Background: Flying accuracy is influenced by pilots' affective reactions to task demands. A better understanding of task‐related emotions and flying performance is needed to enhance pilot training. Objective: Understand pilot trainees' performance and emotional dynamics (intensity, frequency and variability) based on training phase and difficulty level in a flight simulator. Methods: Twenty‐three volunteers performed basic flight manoeuvres. Trials were divided into three phases: Introduction (trials 1–7), session A (trials 8–15) and session B (trials 16–22). Three task difficulty levels were implemented (low, medium and high). Flying performance was evaluated using root mean square error (RMSE) and expert ratings. Emotional intensity was inferred from physiological (electrodermal activity) and behavioural (facial expressions) emotional responses. Emotional variability was calculated to understand fluctuations among multiple emotions. Emotional responses were mapped into task‐relevant emotions, like sadness with boredom, and fear with anxiety. Results and Conclusions: The most frequent facial expressions neutral, anger and surprise. Neutral and anger were interpreted as deep focus states. Surprise was likely a response to unexpected events. Flying performance and emotional dynamics varied across training phases and difficulty levels. During introduction, performance was less accurate, and emotions were less frequent. During session A, performance improved while participants experienced more physiological arousal and emotional variability. During session B, performance was the most accurate. In high‐difficulty tasks, performance was the least accurate, participants expressed emotions with more frequency, more variability and higher physiological arousal. Future studies can use simulated flying tasks for trainees to familiarize with their emotional reactions to task demands expecting to improve training outcomes.