With the prevalent of smart devices and home automations, voice command has become a popular User Interface (UI) channel in IoT environment. Although Voice Control System (VCS) advantages great convenience, it is extremely vulnerable to spoofing attack (e.g., replay attack, hidden/inaudible attack) due its broadcast nature. In this study, we present WiVo, device-free liveness detection system based on wireless signals generated by without any additional or sensors carried users. The basic...

Autonomous vehicles are expected to significantly enhance the human mobility. However, recently researchers have discovered and demonstrated some attacks on vehicles, which caused a panic among public. Furthermore, these that security issue is still one of major challenges vehicles. In this paper, we propose novel edge computing based anomaly detection, coined vehicle detection (EVAD), exploits sensor data fusion identify events. The time domain property, i.e., correlation between different...

The availability of a large amount electronic health records (EHR) provides huge opportunities to improve care service by mining these data. One important application is clinical endpoint prediction, which aims predict whether disease, symptom or an abnormal lab test will happen in the future according patients' history records. This paper develops deep learning techniques for are effective many practical applications. However, problem very challenging since contain multiple heterogeneous...

Sepsis is a life-threatening condition that seriously endangers millions of people over the world.Hopefully, with widespread availability electronic health records (EHR), predictive models can effectively deal clinical sequential data increase possibility to predict sepsis and take early preventive treatment.However, prediction challenging because patients' in EHR contains temporal interactions multiple events.And capturing long event sequence hard for traditional LSTM.Rather than directly...

Clinical outcome prediction based on the Electronic Health Record (EHR) plays a crucial role in improving quality of healthcare. Conventional deep sequential models fail to capture rich temporal patterns encoded longand irregular clinical event sequences. We make observation that events at long time scale exhibit strongtemporal patterns, while within short period tend be disordered co-occurrence. thus propose differentiated mechanisms model different scales. Our learns hierarchical...

Autonomous vehicles are expected to be a disruptive technology that has the potential revolutionize human mobility. However, recent research progress on intra-vehicle network (e.g., revealing of series security vulnerabilities CAN design) demonstrated issue still represents one major challenges future self-driving cars. In this study, we propose novel edge based anomaly detection system, coined VeAnDe, which exploits sensor data fusion identify events. VeAnDe analyzes pair-wise correlations...

Sepsis is a life-threatening condition that seriously endangers millions of people over the world. Hopefully, with widespread availability electronic health records (EHR), predictive models can effectively deal clinical sequential data increase possibility to predict sepsis and take early preventive treatment. However, prediction challenging because patients’ in EHR contains temporal interactions multiple events. And capturing long event sequence hard for traditional LSTM. Rather than...

The advent of the Internet era has led to an explosive growth in Electronic Health Records (EHR) past decades. EHR data can be regarded as a collection clinical events, including laboratory results, medication records, physiological indicators, etc, which used for outcome prediction tasks support constructions intelligent health systems. Learning patient representation from these events is important but challenging step. Most related studies transform into sequence temporal order and then...

Mortality prediction of diverse rare diseases using electronic health record (EHR) data is a crucial task for intelligent healthcare. However, insufficiency and the clinical diversity make it hard directly training deep learning models on individual disease or all from different diseases. these patients with can be viewed as multi-task problem insufficient large number. But tasks little also to train task-specific modules in models. To address challenges diversity, we propose an...

Sepsis is a life-threatening condition that seriously endangers millions of people over the world. Hopefully, with widespread availability electronic health records (EHR), predictive models can effectively deal clinical sequential data increase possibility to predict sepsis and take early preventive treatment. However, prediction challenging because patients' in EHR contains temporal interactions multiple events. And capturing long event sequence hard for traditional LSTM. Rather than...

The integration of Artificial Intelligence (AI) into clinical medicine has recently garnered significant attention, particularly in the context digital pathology and precision medicine. Real-time monitoring analysis patients play a pivotal role overall treatment process. This includes basic body detection indices, analyzing patient's condition, predicting future possibilities for their medical status. In addition to fundamental machine learning (ML), both deep (DL) twin technology have...

Sequential Recommendation (SR) is a fundamental problem in recommender systems, where the goal to learn user/item embeddings from historical purchase sequences of users. Most existing methods focus on capturing sequential patterns item transitions but ignore temporal collaborative signals that reflect evolving user-item interactions over time. These are important for modeling dynamics user preferences and popularity. Therefore, we propose novel framework jointly considers SR, which...

The advent of the Internet era has led to an explosive growth in Electronic Health Records (EHR) past decades. EHR data can be regarded as a collection clinical events, including laboratory results, medication records, physiological indicators, etc, which used for outcome prediction tasks support constructions intelligent health systems. Learning patient representation from these events is important but challenging step. Most related studies transform into sequence temporal order and then...