教学课程资源库 更多>>

关键词： Computer engineering   Computer science  

摘要： The idea of programmable networks has recently regained considerable momentum due to the emergence of the Software-Defined Networking (SDN) paradigm. The architecture associated with SDN allows the decoupling of the control plane from the data plane, hence enabling achievable bi-directional programmability between the two planes. Over the years, academia and industry researchers have worked closely to achieve a set of standards for several use-cases. However, some issues persist in this area. To this end, some pre-existing concepts from traditional networking have been implemented to alleviate this problem, such as the implementation of distributed architecture in the SDN control plane. Although the solution has proven effective in some use cases, some new problem has been inherited from distributed computer architecture. Several distributed SDN controller architectures have been proposed to mitigate the risks of network overload, link and node failures, and other security concerns. However, most of these solutions incur high overhead and low flexibility since they often rely on implementing their solutions on either the already complex control structure or the underlying forwarding agents. Further, due to the statically assigned switches to controller instances, a data store is required to synchronize network states significantly adding to the flow setup latency. Although the need for synchronization can not be over-ruled in such a system, however, the intervals of such inter-controller synchronization can be restricted to a per-need base. Our preliminary shows that by targeting specific events that exert changes to the network topology and how new packets are being handled, we can efficiently utilize the resources available in the network. Furthermore, by delegating such a task to a shim layer logically situated between the data plane and the control plane, we can reduce the communication overhead in the control plane. Since we can safely abstract some functionali

关键词： Medical imaging   Biomechanics   Computer science  

摘要： Joint pain or loss of function manifests in subtle changes in kinematics and the biomechanical behavior of the articulating surfaces, the ligaments, and tendons acting as soft tissue constraints. Understanding the relationships between anatomic morphology and healthy joint function, and how conservative and surgical treatments affect symptomatic behavior in vivo, will inform clinicians’ practice. Clinically motivated research studies benefit greatly from the increased spatial and temporal resolution afforded by biplanar fluoroscopic systems, but the kinematic data generated from these studies come at high computational cost and radiation exposure to test subjects and operators. Additionally, there is no off-the-shelf software or hardware solution available currently for foot and ankle gait biplanar fluoroscopic systems. The imaging chains and software processing pipelines are modified and repurposed versions of stock fluoroscopic equipment. By altering this hardware chain, image quality and automated processing, and thus the accuracy and throughput of the system, are diminished. Therefore, the goal of this thesis is to dissect the custom hardware and software pipelines in our laboratory in order to understand, characterize, and attempt to mitigate sources of error while balancing kinematic accuracy, subject safety, and processing throughput.

关键词： Computer science   data queries   index   summarization   top-k   uncertainty  

摘要： Query processing is an important problem in many research fields including database systems, data mining, and geometric computing. The goal is to preprocess input data into an index to efficiently answer queries over data. In many real applications we need to handle a large amount of data or/and data that are ambiguous because of human errors and data integration. With the data sets becoming increasingly large and complex, queries are also becoming complex, therefore, new challenging problems have emerged in the area of query processing. Data summarization helps to accelerate expensive data queries by running the query procedure in a small data summary rather than the entire large dataset. The first part of this thesis studies the problem of finding small high quality data summaries in the Euclidean space. Data summarization can be applied either in the entire dataset or among points in a query range given by the user. Efficient algorithms are proposed to get a small summarization of the entire dataset so that Top-k or user preference queries are answered efficiently with high quality guarantees by looking only in the data summary. Such a summary can also be used in multi-criteria decision problems. Furthermore, near-linear space indexes are designed so that given a query rectangle, a diverse high-valued summary of k points in the query range is returned efficiently. The second part of the thesis focuses on data queries over uncertain data. Uncertainty is typically captured using stochastic data models, and querying data requires either statistics about the probabilistic behavior of the underlying data, or data cleaning to reduce the uncertainty of the query answer. Small size indexes are built so that, given a query rectangle, statistics of the MAX (or top-k) operator among the points in the query range are computed in sublinear time. In addition, approximation algorithms are proposed for selecting data to clean in order to minimize the variance of the result of a

关键词： Computer science  

摘要： The rise of deepfake and disinformation is growing exponentially since its inception three years ago. A combination of deep learning and fake, deepfakes use machine learning and artificial intelligence technologies to manipulate audio and video media to create fake media content with the high potential to deceive. Cybercriminals are using deepfakes to connect with our emotional biases to alter our way of thinking. The innovation of deepfake technology has been grossly misused to defame an individual’s character, create political havoc, disseminate fake news via social media platforms, and threaten stability and democracy globally. There is wide concern that foreign adversaries may seek to threaten and influence the U.S. 2020 electoral process with deepfake disinformation. Disinformation is being used as a weaponization to cause harm and compromise a corporation’s reputation. The digital age of deepfakes is growing in sophistication each day and the potential of spreading misinformation via social media platforms is far and wide, eroding the trust citizens have in digital online media content. While there is a great concern about the problematics of deepfakes, the technology has the potential to be used for the good of humanity. Ultimately, as new technologies are being developed to detect and suppress deepfakes, cybercriminals will continue to develop more advances and new forms of threats to maintain a step ahead. To ensure development to protect the nefarious use of deepfake technology, governments, businesses, and academia must work together to create the proper infrastructure for our information ecosystems to grow. Keywords: Cybersecurity, Professor Duane Corbo, national security, authenticity, cyber, warfare.

关键词： Computer science  

摘要： Deep learning methods have shown unparalleled performance when trained on vast amounts of diverse labeled training data, often collected at great cost. In many contexts, we have lots of labeled examples but only a few individuals, can be thought of as “little big data,” where we would like to take advantage of the large number of samples while still being cognizant of the fact that the number of observed groups is small. This problem is often known as domain adaptation or transfer learning. In this dissertation, I will cover four major topics. Electroencephalography (EEG) and Local Field Potential (LFP) signals are “big” in terms of the size of recorded data but rarely have sufficient labels required to train complex models. Furthermore, they are collected from limited number of individuals. The first topic I will introduce an interpretable neuro model for electrophysiological signals and explain why transfer learning helps in real situations. In the following two topics, I will expand the discussion of transfer learning problem with two real and challenging setups. Since data outliers will always exist in practice, many of the sources may be irrelevant to the target task, so ignoring the structure of the dataset is detrimental. Learning domain relationships are often insightful in their own right, and they allow domains to share strength without interference from irrelevant data. On top of the problems of outliers, label shift, where the percentage of data in each class is different between domains, is also essential for transfer learning. Transfer learning needs target sample during the training stage, while this requirement may not be satisfied in practice. The last topic discusses the situation on generalizing a trained model on unseen testing samples, where each training domain has a unique classifier and each test data point is predicted by an admixture over the different domain classifiers.

关键词： Information technology   Computer science  

摘要： The term Advanced Persistent Threats (APTs) refers both to highly-sophisticated, often nation-state attackers with tremendous resources and to the malware they employ to compromise their target – to which no organization has proven immune. Dynamic and deception-based defense techniques offer a possible solution. Such techniques, including Moving Target Defense (MTD) and Adaptive Cyber Defense(ACD) techniques, prevent or delay attacks against computer networks by dynamically altering characteristics of the systems or network in a manner to present attackers with a variable, possibly deceptive attack surface and disrupt the planning or execution of cyber-attacks. To better leverage these techniques, this work proposes a novel model to capture how advanced, stealthy adversaries, including APT actors, acquire knowledge about the target network and establish and expand their foothold within the system. This model quantifies the cost and reward, from the adversary's perspective, of compromising and maintaining control over targets within the network. With this foundational understanding of attacker incentives and deterrents, as well as their predicted position in the network, existing defenses can be refined and innovative defenses can be built specifically to counteract the threat posed by APTs.

关键词： Computer science  

摘要： The brain is a complex, interconnected network par excellence. It is made up of intricate connections of neurons, responsible for transmission of signals throughout the body. Over time, experimental studies have illustrated a variety of behavioral features of network dynamics, ranging from stochastic spiking to synchronized bursting observed in the living preparations of neuronal cultures. A fundamental feature of developing neural circuits is the presence of spontaneous network activity. Such spontaneous activity plays putative roles ranging from synaptic development and maintenance to anticipatory states which assist animals in reaching rapid decisions with limited sensory input. Understanding the mechanisms of spontaneously generated activity and interaction patterns between neurons are, therefore, issues of substantial importance. But while changes in the dynamics of coordinated, spontaneous spiking activity have been investigated in different in vivo and in vitro neuronal systems, the mechanisms underlying the generation of these spontaneous spiking activities still remain unclear. Computational simulations help researchers to gain a more detailed understanding of activity patterns in large networks. The BrainGrid simulator is a neural simulator, based on a leaky integrate-and-fire computational model and developed in UW Bothell, that allows us to perform detailed analysis of the effects of model parameters on burst shape and timing, their changes, their patterns and the inter-relationship among these behaviors, gross network structure, and model parameters. Researchers draw inferences about dynamics of these networks from simulation results. These simulations’ high spatiotemporal resolution and long duration produce data that, in terms of both quantity and complexity, challenge our interpretative abilities. Therefore, this thesis focuses on uncovering underlying patterns during spontaneous activities using various machine learning and deep learning techniques

关键词： Computer science  

摘要： The application-layer distributed denial of service (App-DDoS) attack is one of the most menacing types of cyberattacks that circumvent web servers. As attackers have developed different techniques and methods, preventing App-DDoS attacks has become more difficult. A commonly targeted protocol in the application layer is the HTTP-GET flooding attack, where the attacker sends a large number of HTTP-GET requests from multiple infected devices, forcing the server to devote all available resources responding to all the requests. This attack exhausts the server’s resources and denies service to legitimate users. App-DDoS attacks are extremely costly in terms of resource exhaustion, affecting intended clients’ quality of service (QoS). The current range of defense mechanisms against App-DDoS attacks have several limitations, which include slow and delayed attack detection, increased computation load, and reduced hardware computational capacity. The aim of this research is to design and develop an App-DDoS attack detection and mitigation approach to defend web servers against such attacks. We develop a holistic DDoS mitigation framework to detect and mitigate all types of DDoS attacks. Our general defense model has four main components: a screener, policy control, a resource monitoring protocol, and a reporting module. These components interact during screening and security service stages to achieve robust mitigation of various types of DDoS attacks. Based on this general defense model, we derive a new specific scheme to detect and mitigate App-DDoS attacks at an early stage, ensuring the App-DDoS attacks will not degrade the QoS for legitimate users. Our defense system employs three principle modes: normal, screening, and suspicious. The defense scheme transits between these modes, based on the server load. The detection method employs machine learning (ML) techniques during the screening mode, improving detection of App-DDoS attacks. Our defense system is designed to aut

关键词： Fault-tolerant computing   Computer Science  

摘要： In-situ scientific workflows, i.e., executing the entire application workflows on the HPC system, have emerged as an attractive approach to address data-related challenges by moving computations closer to the data, and staging-based frameworks have been effectively used to support in-situ workflows at scale. However, running in-situ scientific workflows on extreme-scale computing systems presents fault tolerance challenges which significantly affect the correctness and performance of workflows. First, scientific in-situ workflow requires sharing and moving data between coupled applications through data staging. As the data volumes and generate rates keep growing, the traditional data resilience approaches such as n-way replication and erasure codes become cost prohibitive, and data staging requires more scalable and efficient approach to support the data resilience. Second, Increasing scale is also expected to result in an increase in the rate of silent data corruption errors, which will impact both the correctness and performance of applications. Moreover, this impact is amplified in the case of in-situ workflows due to the dataflow between the component applications of the workflow. Third, since coupled applications in workflows frequently interact and exchange the large amount of data, simply applying the state of the art fault tolerance techniques such as checkpoint/restart to individual application component can not guarantee data consistency of workflows after failure recovery. Furthermore, naive use of these fault tolerance techniques to the entire workflows will limit the diversity of resilience approaches of application components, and finally incur a significant latency, storage overheads, and performance degradation. This thesis addresses these challenges related to data resilience and fault tolerance for in-situ scientific workflows, and makes the following contributions. This thesis first presents CoREC, a scalable resilient in-memory data staging runti

关键词： Computer science  

摘要： The purpose of this qualitative grounded theory study was to understand the problem behind the lack of in-house, small business Cybersecurity technical experts to respond to the government Cybersecurity Controlled Unclassified Information (CUI) mandate. After notable government breaches in 2015, most memorably, the Office of Personnel Management and the Internal Revenue Service breaches affecting close to five million government employees records, the government issued mandates requiring defense contractors to protect CUI residing in non-Federal information systems and organizations. The deadline passed with a large proportion of small businesses still non-compliant with the mandate. This research examined why small business contractors are so unprepared, especially working for the government. The study’s findings provided valuable information for the government and contractors moving forward to the pursuant of new mandates. Survey responses of 16 questions by 21 small business contractors from state and federal government revealed three emerging themes: lack of resources, government standards and regulation, and lack of understanding. The findings of the study has proven to be consequential and relevant. In January 2020 the government cited the implementation of a new program which they declared should help small business contractors and other government agencies with Cybersecurity compliance and mandates. This study gives the government additional arguments and materials to generate the most effective tool to assist small businesses in protecting government information residing in information systems and organizations outside the governments control.