教学课程资源库 更多>>

关键词： Artificial intelligence   Computer science  

摘要： Training supervised machine learning models requires labeled examples. A judicious choice of examples is helpful when there is a significant cost associated with assigning labels. This dissertation aims to improve upon a promising extant method—Batch-mode Expected Model Change Maximization (B-EMCM) method—for selecting examples to be labeled for regression problems. Specifically, it aims to develop and evaluate alternate strategies for adaptively selecting batch size in B-EMCM, named adaptive B-EMCM (AB-EMCM). By determining the cumulative error that occurs from the estimation of the stochastic gradient descent, a stop criteria for each iteration of the batch can be specified to ensure that selected candidates are the most beneficial to model learning. This new methodology is compared to B-EMCM using mean absolute error and root mean square error over ten iterations using benchmark machine learning data sets. Using multiple data sets and metrics across all methods, one of the variations of ABEMCM, that uses the max bound of the accumulated error (AB-EMCM Max), showed the best results for an adaptive batch approach. It achieved better root mean squared error (RMSE) and mean absolute error (MAE) than the other adaptive and nonadaptive batch methods while reaching the result in nearly the same number of iterations as the non-adaptive batch methods.

关键词： Computer science  

摘要： Passwords and users have a complicated relationship. Users need passwords to authenticate to services. Meanwhile, secure passwords are difficult to remember and users are only creating more passwords for the myriad of online services that they use. Password managers are one solution to the password problem. A novel generative password manager was presented by Schmittle (2018) called RGPM. I created a zero trust, open source, feature complete implementation of RGPM, called RGPMv2, and performed a user study to investigate the usability of the generative password manager by evaluating participants' understanding and use of the developed application.

关键词： Computer engineering   Computer science   Artificial intelligence   Deep learning   Deep neural network   Processing-step efficiency   Robust model efficiency   Versatile model efficiency  

摘要： Breakthrough of deep learning (DL) has greatly promoted development of machine learning in numerous academic disciplines and industries in recent years. A subsequent concern, which is frequently raised by multidisciplinary researchers, software developers, and machine learning end users, is inefficiency of DL methods: intolerable training and inference time, exhausted computing resources, and unsustainable power consumption. To tackle the inefficiency issues, tons of DL efficiency methods have been proposed to improve efficiency without sacrificing prediction accuracy of a specified application such as image classification and visual object detection. However, we suppose that the traditional DL efficiency methods are not sufficiently flexible or adaptive to meet requirement of practical usage scenarios, based on two observations. First, most of the traditional methods adopt an objective "no accuracy loss for a specified application", while the objective cannot cover considerable scenarios. For example, to meet diverse user needs, a public cloud platform should provide an efficient and multipurpose DL method instead of focusing on an application only. Second, most of the traditional methods adopt model compression and quantization as efficiency enhancement strategies, while these two strategies are severely degraded for a certain number of scenarios. For example, for embedded deep neural networks (DNNs), significant architecture change and quantization may severely weaken customized hardware accelerators designed for predefined DNN operators and precision. In this dissertation, we will investigate three popular usage scenarios and correspondingly propose our DL efficiency methods: versatile model efficiency, robust model efficiency, and processing-step efficiency. The first scenario is requiring a DL method to achieve model efficiency and versatility. The model efficiency is to design a compact deep neural network, while the versatility is to get satisfactory predict

关键词： Computer science  

摘要： In the presence of big data analysis, large volume of data needs to be systematically indexed to support analytical tasks, such as feature engineering, pattern recognition, data mining, and query processing. The volume, variety, and velocity of these data necessitate sophisticated systems to help researchers understand, analyze, and dis- cover insights from heterogeneous, multidimensional data sources. Many analytical frameworks have been proposed in the literature in recent years, but challenges to accuracy, speed, and effectiveness remain hence a systematic approach to perform data signature computation and query processing in multi-dimensional space is in people’s interest. In particular, real-time and near real-time queries pose significant challenges when working with large data sets. To address these challenges, I develop an innovative robust multi-variate feature extraction algorithm over multi-dimensional temporal datasets, which is able to help understand and analyze various real-world applications. Furthermore, to answer queries over these features, I develop a novel resource-aware indexing framework to approximately solve top-k queries by leveraging onion-layer indexing in conjunction with locality sensitive hashing. The proposed indexing scheme allows people to answer top-k queries by only accessing a bounded amount of data, which optimizes big data small for queries.

关键词： Computer science  

摘要： The goal of this work is to lay down a theoretical foundation for the use of symbolic execution to verify and refute relational properties. Relational properties are properties about two runs of two programs on related inputs, or about two executions of a single program on related inputs. Relational properties are useful to formalize notions in security and privacy, and to reason about program optimizations. Standard (unary) symbolic execution has already been used to verify and refute relational properties of programs. Usually, this is done by reducing the relational property to a unary one for a modified program. Relational symbolic execution instead relies on a relational semantics. It allows the user to reason about pairs of programs without the need to modify the program or the specification. The first part of this manuscript will show examples of deterministic relational properties, and strategies to reduce the set of possible set of states. The result, Relational Symbolic Execution, or RSE, is a sound and (relatively) complete verification (and refutation) system for relational deterministic properties. The second part of this manuscript will concern a specific relational probabilistic property: Differential privacy. Differential privacy is a gold standard for privacy in statistical analysis that can be satisfied by algorithms and hence programs. The definition provides a quantitative bound on the difference of the information, about any single user in the input dataset, that an algorithm releases in two possible worlds: one where that single, arbitrary, user is part of the input dataset and one where it is not. Its definition implies that randomness must be used by any differentially private algorithm. While its use is now widespread, its correct implementation is still hard to achieve. Indeed, many algorithms, allegedly differentially private, have been released and used, only to find time later that they are, in fact, not differentially private. Probabilis

关键词： Electrical engineering   Computer science  

摘要： Tremendous successes in machine learning have been achieved in a variety of applications such as image classification and language translation via supervised learning frameworks. Recently, with the rapid increase of electronic health records (EHR), machine learning researchers got immense opportunities to adopt the successful supervised learning frameworks to diverse clinical applications. To properly employ machine learning frameworks for medicine, we need to handle the special properties of the EHR and clinical applications: (1) extensive missing data, (2) model interpretation, (3) privacy of the data. This dissertation addresses those specialties to construct end-to-end machine learning frameworks for clinical decision support. We focus on the following three problems: (1) how to deal with incomplete data (data imputation), (2) how to explain the decisions of the trained model (model interpretation), (3) how to generate synthetic data for better sharing private clinical data (synthetic data generation). To appropriately handle those problems, we propose novel machine learning algorithms for both static and longitudinal settings. For data imputation, we propose modified Generative Adversarial Networks and Recurrent Neural Networks to accurately impute the missing values and return the complete data for applying state-of-the-art supervised learning models. For model interpretation, we utilize the actor-critic framework to estimate feature importance of the trained model's decision in an instance level. We expand this algorithm to active sensing framework that recommends which observations should we measure and when. For synthetic data generation, we extend well-known Generative Adversarial Network frameworks from static setting to longitudinal setting, and propose a novel differentially private synthetic data generation framework. To demonstrate the utilities of the proposed models, we evaluate those models on various real-world medical datasets including cohorts i

关键词： Computer science  

摘要： Malware or malicious software poses an increasing threat to sensitive data and network systems. With the advancement of technology and the omnipresence of internet connectivity and online services, this menace is getting more eminent every day. Malware writers are inventing new and well-built malware and are using different techniques to obfuscate their malcode. Therefore, it is imperative to have an efficient and self-sufficient malware detection system. Machine learning models are capable of producing highly accurate malware classifications. However, there is a growing number of unlabeled datasets that the models rely on for classification. In this thesis, we use Semi-Supervised Learning models to detect malicious Android apps from benign ones. We observe if the application of co-training helps improve classifier performance. We focus our analysis on Android malware and try to determine the best set of classifiers that, alone or combined, generate the most accurate predictions. Furthermore, we incorporate malware sub-classes along with co-training to observe its effect on classifier accuracy.

关键词： Botnet   Systematics   Forensics   Internet of Things   Measurement   Computer science   Bibliographies   IoT   botnet   systematic review  

摘要： The adoption of the Internet of Things (IoT) technology is expanding exponentially because of its capability to provide a better service. This technology has been successfully implemented on various devices. The growth of IoT devices is massive at present. However, security is becoming a major challenge with this growth. Attacks, such as IoT-based botnet attacks, are becoming frequent and have become popular amongst *** has a resource constraint and heterogeneous environments, such as low computational power and memory. Hence, these constraints create problems in implementing a security solution in IoT devices. Therefore, various kind of attacks are possible due to this vulnerability, with IoT-based botnet attack being one of the most *** this study, we conducted a comprehensive systematic literature review on IoT-based botnet attacks. Existing state of the art in the area of study was presented and discussed in detail. A systematic methodology was adopted to ensure the coverage of all important studies. This methodology was detailed and repeatable. The review outlined the existing proposed contributions, datasets utilised, network forensic methods utilised and research focus of the primary selected studies. The demographic characteristics of primary studies were also *** result of this review revealed that research in this domain is gaining momentum, particularly in the last 3 years (2018-2020). Nine key contributions were also identified, with Evaluation, System, and Model being the most conducted.

关键词： Computer science  

关键词： Mathematics   Computer science  

摘要： We investigate the computational complexity for determining various properties of a finite semigroup given by generators. We begin with transformation semigroups and introduce a simple framework to describe properties that are decidable in AC^0, the class of all sets that are decidable by unbounded fan-in Boolean circuits of constant depth. We then use this framework to show that the problems of deciding whether a transformation semigroup is commutative, a semilattice, or a group are in AC^0. We prove that deciding whether a transformation semigroup has a left (resp. right) zero is NL-complete -- that is, as difficult as any problem that can be solved in nondeterministic logarithmic space. We also prove the following problems are NL-complete: testing whether a transformation semigroup is nilpotent, is R-trivial, or has central idempotents. We also give NL algorithms for testing whether a transformation semigroup is idempotent, orthodox, completely regular, Clifford or has commuting idempotents. Some of these algorithms are direct consequences of a more general result that arbitrary fixed semigroup equations can be tested in NL. Moreover, we show how to compute left and right identities of a transformation semigroup in polynomial time. Finally, we show that checking whether an element is regular is PSPACE-complete -- that is, as difficult as any problem that can be solved in polynomial space. We also investigate properties of matrix semigroups. We describe polytime algorithms for checking if the semigroup is commutative, a semilattice, or a group and for enumerating left and right identities. We further prove that checking for nilpotence is in P. Lastly, we investigate properties of partial bijection semigroups. We prove that determining if such a semigroup is a band is in AC^0 and that checking idempotent membership is PSPACE-complete. We conclude with extending our result for semigroup equations to the case of inverse partial bijection semigroups.