教学课程资源库 更多>>

关键词： Computer science  

摘要： In the healthcare sector, there is a strong demand for accurate objective pain assessment as a key for effective pain management. Real-time and accurate objective pain assessment help caregivers and hospital staffs decide the proper dosage of pain medication to be provided to a patient in a timely manner. The state-of-the-art automatic and objective pain assessment techniques in the literature can be classified into two main categories: physiological-based and behavioral-based. The first-class monitors the changes in patients' physiological data such as heart rate (HR), heart rate variability (HRV), Electrocardiography (ECG), Electromyography (EMG), Photoplethysmography (PPG) to identify autonomic nervous system reactions to pain, while the second class utilizes behavioral reactions to pain such as techniques using computer vision-based techniques by extracting features from patients' head poses and facial expressions. Recent pain monitoring systems have recently gained attention on multi-modality meaning that they deploy a combination of both approaches to improve pain monitoring accuracy. Although such complex models are highly accurate in pain monitoring, they are more computationally intensive imposing feasibility limitations to implement them on wearable devices in terms of energy efficiency (battery life) as well as computation latency. A smart and self-aware system capable of adaptively making a decision at run-time in response to the changes in pain level and context can minimize energy consumption by dynamically offloading tasks to the gateway devices at the edge layer. For this reason, in this work, a self-aware system is proposed for the continuous assessment of pain intensity at the edge layer. Using the BioVid heat pain dataset, this approach demonstrates a promising reduction in terms of energy consumption with a negligible accuracy loss compared with its non-adaptive counterpart.

关键词： Robotics   Artificial intelligence   Computer science  

摘要： The basis of this work is a control architecture for collaborative multi-robot systems focusing on the problem of task allocation under hierarchical constraints imposed upon a joint task. For these types of tasks, multiple robots need to dynamically coordinate their execution during the task execution. However, due to the different abilities between different robots, a single task definition is not sufficient to ensure the task can be completely without faults. In order to alleviate these concerns, we have developed a generalized task structure which is able to transfer skills of a learned task to teams of heterogeneous robots. This system uses a small number of human demonstrations to learn a hierarchical task structure on a single robot. This structure acts as a skeleton for the task which has been adapted to work on teams of heterogeneous robots through the development of a continuous-valued metric which is able to account for the robots' variable skills during the task execution. Additionally, to further emphasize the collaboration of the multi-robot team, our previous architecture is extended to include an interdependence constraint which requires explicit cooperation between agents to complete the task. Furthermore, to allow the task execution as well as the learning of the task structure to be as generalizable as possible, several efforts were made to extend the previous architecture to work with human-robot teams. First, a system was created to learn the hierarchical tasks through verbal instruction. Second, a dialogue-based fault recovery system was developed to allow for a more robust task execution. Lastly, an intent recognition system was incorporated into the architecture to allow for human-robot teams to work collaboratively on a task. Each of these extensions were validated separately through several experiments utilizing either multi-robot or human-robot teams for pick and place tasks with hierarchical constraints. The experiments included different

关键词： Communication   Information technology   Computer science  

摘要： Because of the evolution of technology and the increase in threats to online data, security professionals must develop and update robust cybersecurity strategies to protect national data. The purpose of this qualitative, phenomenological study was for the researcher to explore the experiences of 10 security experts supporting federal networks and critical infrastructure systems. The study explored the strategies security experts use to overcome challenges security experts face when supporting federal networks and critical infrastructure systems. Open-ended questions supported the semi-structured interviews, which allowed the researcher to collect data and explore the participants’ experiences regarding the research problem. The data analysis resulted in the development of multiple emerging major themes. The major themes developed were (a) processes and methods, (b) frameworks (standards), (c) resources (knowledge), (d) resources (people), (e) budget, (f) advanced technology, (g) IT governance and leadership. There were 59 references, which reflected that processes and methods are critical to a successful cybersecurity risk mitigation and prevention plan. There were 47 references, which identified that the use of well-known frameworks directly supports the development and implementation of adequate processes and methods. There were 26 reference points identified in two categories: knowledge and people. Seventeen reference points were indicating that a strong budget is mandatory to implement the processes and tools required for risk management. There were 16 reference points regarding the criticality to deploy advanced technologies to secure data. Additionally, 13 participant data points identified that strong IT governance and leadership are required for any cybersecurity strategy to be successful. The results of the study contribute to the business technical problem by providing insight to stakeholders (i.e., leaders, executives, information officers, financial offi

关键词： Computer science  

摘要： Independently developed codebases typically contain many segments of code that perform the same or closely related operations (semantic clones). Finding functionally equivalent segments enables applications like replacing a segment by a more efficient or more secure alternative. Such related segments often have different interfaces, so some glue code (an adapter) is needed to replace one with the other. We present an algorithm that searches for replaceable code segments by attempting to synthesize an adapter between them from some finite family of adapters; it terminates if it finds no possible adapter. The use of symbolic execution during adapter search implicitly relies on being able to summarize the entire adapter into a single formula. Merging related execution paths is a powerful technique for reducing path explosion in symbolic execution. One approach, introduced and dubbed “veritesting” by Avgerinos et al., works by statically translating a bounded control flow region into a single formula. Java’s typed memory structure is very different from a binary, but we present an extension of previous path-merging approaches for symbolic execution of Java. Bugs in commercial software and third-party components are an undesirable and expensive phenomenon. Such software is usually released to users only in binary form. The lack of source code renders users of such software dependent on their software vendors for repairs of bugs. Such dependence is even more harmful if the bugs introduce new vulnerabilities in the software. Being able to automatically repair security bugs breaks this dependence and increases software robustness. In the third part of this dissertation, we present an automated program repair approach for binary code.

关键词： Computer science  

摘要： Malicious hackers utilize the World Wide Web to share knowledge. Previous work has demonstrated that information mined from online hacking communities can be used as precursors to cyber-attacks. In a threatening scenario, where security alert systems are facing high false positive rates, understanding the people behind cyber incidents can help reduce the risk of attacks. However, the rapidly evolving nature of those communities leads to limitations still largely unexplored, such as: who are the skilled and influential individuals forming those groups, how they self-organize along the lines of technical expertise, how ideas propagate within them, and which internal patterns can signal imminent cyber offensives? In this dissertation, I have studied four key parts of this complex problem set. Initially, I leverage content, social network, and seniority analysis to mine key-hackers on darkweb forums, identifying skilled and influential individuals who are likely to succeed in their cybercriminal goals. Next, as hackers often use Web platforms to advertise and recruit collaborators, I analyze how social influence contributes to user engagement online. On social media, two time constraints are proposed to extend standard influence measures, which increases their correlation with adoption probability and consequently improves hashtag adoption prediction. On darkweb forums, the prediction of where and when hackers will post a message in the near future is accomplished by analyzing their recurrent interactions with other hackers. After that, I demonstrate how vendors of malware and malicious exploits organically form hidden organizations on darkweb marketplaces, obtaining significant consistency across the vendors’ communities extracted using the similarity of their products in different networks. Finally, I predict imminent cyber-attacks correlating malicious hacking activity on darkweb forums with real-world cyber incidents, evidencing how social indicators are crucial for

关键词： Linguistics   Psychology   Computer science  

摘要： Theoretical linguists have long argued that humans’ knowledge of language is internalized in the form of rich grammatical representations. Formalizing the connection between grammatical operations and cognitive processes would then make it possible for experimental data to inform syntactic theories of language knowledge and use. This dissertation follows a line of research addressing these issues from a computational perspective. It does so by providing a transparent, interpretable link between structural representations and off-line processing behavior — the empirical observation that some sentences are overall harder to process than others. In particular, I expand on past literature arguing that a top-down parser for Minimalist grammars (Stabler, 1996, MGs) can be used to relate parsing behavior and grammatical structure to memory usage — thus asking to which degree the representations hypothesized by linguists are relevant to processing (Kobele et al., 2013; Gerth, 2015; Graf et al., 2017). First, I explore the performance of the MG model on a variety of word order and relative clause processing asymmetries in Italian, thus demonstrating the sensitivity of the linking theory to detailed grammatical information. Then, I propose the MG parser as a good, non-probabilistic formal model of how gradient acceptability can be derived from categorical grammars. In doing so, I show how psycholinguistic data can address fundamental questions about the nature of grammatical knowledge. Finally, I evaluate the model’s predictions for a variety of psycholinguistic phenomena known as syntactic priming effects, and propose possible extensions to the computational framework that explore the contributions of grammatical features to memory load. By investigating the MG model’s performance across this diverse array of processing phenomena, this dissertation adds further support to the psychological plausibility of fine-grained grammatical knowledge contributing to processing cost. It

关键词： Medical imaging   Artificial intelligence   Computer science  

摘要： Cardiac magnetic resonance (CMR) image segmentation has been a crucial tool for medical professionals to diagnose cardiovascular diseases (CVDs), which are the leading causes of death throughout the world. Segmenting CMR images is very time consuming and increases the cost of CVD diagnoses and treatment, making them inaccessible to many. Automated CMR image segmentation models strive to lower the cost of CVD diagnosis, but such models must be efficient and accurate in such failure-sensitive domains as human medicine. This thesis proposes to apply γ-Net, a recurrent extension of the popular U-Net, to automatically perform high-quality CMR image segmentation. γ-Net is a recent development by Linsley et al. of Brown University, and has exhibited the ability to outperform U-Net on very small datasets, which is beneficial given the very limited amount of patient CMR data available to the scientific community. γ-Net leverages biological principles backed by anatomical evidence as well as attention mechanisms in order to achieve its high efficiency. In this thesis, we examine the following topics: (a) γ-Net’s resilience to smaller training set sizes, which is cruicial when little patient data is available; (b) resilience to variation in training and validation data, which is shown to significantly degrade performance in state-of-the- art models; and (c) the ability to transfer to new datasets with minimal fine tuning, which saves training cost for practical applications. We have found that (a) γ-Net significantly outperforms an equivalent U-Net in validation performance when trained using a reduced training set; (b) γ-Net is much more resilient to input variations than U-Net; and (c) γ-Net generalizes to new datasets better than comparable U-Nets.

关键词： Computer science  

摘要： Failures in cyber physical systems can be costly in terms of money and lives. The mars climate orbiter alone had a mission cost of 327.6 million USD which was almost completely wasted do to an uncaught design flaw. This shows the importance of being able to define formal requirements as well as being able to test the design against these requirements. One way to define requirements is in Metric Temporal Logic (MTL), which allows for constraints that also have a time component. MTL can also have a distance metric defined that allows for the calculation of how close the MTL constraint is to being falsified. This is termed robustness. Being able to calculate MTL robustness quickly can help reduce development time and costs for a cyber physical system. In this thesis, improvements to the current method of computing MTL robustness are proposed. These improvements lower the time complexity, allows parallel processing to be used, and lowers the memory foot print for MTL robustness calculation. These improvements will hopefully increase the likelihood of MTL robustness being used in systems that were previously inaccessible do to time constraints, data resolution or real time systems that need results quickly. These improvements will also open the possibility of using MTL in systems that operate for a large amount of time and produce a large amount of signal data

关键词： Computer engineering   Computer science  

摘要： Query-based summarization problem is an interesting problem in the text summarization field. On the other hand, the reinforcement learning technique is popular for robotics and becoming accessible for the text summarization problem in the last couple of years (Narayan et al., 2018). The lack of significant works using reinforcement learning to solve the query-based summarization problem inspired us to use this technique. While doing so, We also introduce a different approach for sentence ranking and clustering to avoid redundancy in summaries. We propose an unsupervised extractive summarization method, which provides state-of-the-art results on some metrics. We develop two abstractive multi-document summarization models using the reinforcement learning technique and the transformer model (Vaswani et al., 2017). We consider the importance of information coverage and diversity under a fixed sentence limit for our summarization models. We have done several experiments for our proposed models, which bring significant results across different evaluation metrics.

关键词： Computer science  

摘要： The cyberspace is keeping evolving in many aspects. In the last decade by 2020, we have witnessed the pervasiveness of mobile computing, the increasing adoption of IoT platforms, and the emerging application of artificial intelligence in various scenarios. However, along with those exciting advancements are emerging and critical cyberspace threats, as demonstrated in this thesis. Particularly, this thesis reveals, for the first time, the ecosystem of residential proxy (RESIP) as a service, in which a provider utilizes the hosts within residential networks (in contrast to those running in a datacenter) to relay their customers’ traffic, in an attempt to avoid server-side blocking and detection. This ecosystem was found to have a large scale with more than 6M RESIP peers captured, a suspicious recruitment process involving IoT devices and malware, and a concerning usage in online abuse activities such as advertisement fraud. More importantly, it reveals an emerging security threat to common users by abusing the network bandwidth of their electric devices. Following this work, we further studied how mobile devices have been turned into proxy peers through proxy SDKs. Results show that mobile users are not well informed of traffic relaying through their mobile devices. Also, this mobile proxy ecosystem has become a monetization channel for mobile malware. In the third work, the spotlight is moved to another emerging area: virtual personal assistants (VPA). VPA platforms including Google Home and Amazon Alexa allow third parties to publish diverse functions (skills). In this study, two new attacks (Voice Squatting and Voice Masquerading) have been discovered where VPAs can misunderstand a user’s intention and mistakenly invoke a malicious skill. This study demonstrates how error-prone AI can incur new security threats to common users. Lastly, this thesis studied the interaction of IoT devices in the smart home scenario. Specifically, a comprehensive study was carried out