教学课程资源库 更多>>

关键词： Computer science  

摘要： The design of an interface can have a huge impact on human productivity, creativity, safety, and satisfaction. Therefore, it is crucial that we provide user interface designers with the tools to make them efficient, more creative, and better understand their users. However, designers face key challenges in their tools throughout the design process. Designers explore alternatives of their interface layouts when prototyping. However, they are limited to exploring the layouts they can ideate and sketch, create in their prototyping tools, or find in other examples not containing their interface elements. In usability testing, designers can conduct large-scale studies and deploy their interfaces to gather data from crowdworkers, however, such studies can be expensive, time consuming, and difficult to conduct iteratively throughout the design process. Finally, designers often find that existing interfaces can be a platform for prototyping and enabling new forms of interaction, but existing interfaces are often rigid and difficult to modify at runtime. In this dissertation, I explore how we can use advanced technologies from program analysis and synthesis and machine learning, to enable semantic and data-driven analyses of interfaces. If we augment interface design tools with the capabilities of understanding, transforming, augmenting, and analyzing an interface design, we can advance designers' capabilities. Through semantic analysis of interfaces, we can help designers ideate more rapidly, prototype more efficiently, and more iteratively and cheaply evaluate the usability of their interface designs. I demonstrate this through four systems that (1) let designers rapidly ideate alternative layouts through mixed-initiative interaction with high-level constraints and feedback, (2) help designers adapt examples more efficiently by inferring semantic vector representation from an example screenshot, (3) enable designers to quickly and cheaply analyze a key aspect of the usabil

关键词： Economics   Computer science  

摘要： In the recent past the cryptocurrency ecosystem has seen explosive growth. What was once a small market, made popular with the introduction of Bitcoin in 2009, has now grown to a pseudo-financial system with a market capitalization of over 230 billion USD. Many of the cryptocurrencies available have been built with security in mind, utilizing a public ledger as well as safe-guards against counterfeiting. One might be led to believe that the public nature of cryptocurrencies would make them an infrequent target for criminal activity. However, that is not entirely the case. Many of the services utilized by the cryptocurrency ecosystem are centralized utilities that store transactions in databases that, while partially made public through data endpoints, are not attached to the blockchain. This dissertation leverages public data, typically through off-chain utilities, to study cybercriminal activities. First, the impact that distributed denial-of-services attacks have on the Bitcoin ecosystem was investigated by examining trading activity on the Mt. Gox exchange. It was found that there are fewer extremely large trades on days following a shock. Next the suspicious trading activity of two actors found trading on the Mt. Gox exchange was analyzed. Between February 2013 to November 2013 these two actors were able to fraudulently purchase around 600,000 bitcoin worth approximately 112 million USD. The trading activity associated with one of the actors was found to be highly correlated with the impressive BTC/USD exchange rate increase at the end of 2013. Following the fall of Mt. Gox, finding out if cryptocurrencies fail in ways similar to the services they support was of interest. Examining 1,082 coins over nearly a five year period, it was found that 44% of publicly-traded coins are abandoned at least once with 18% of coin abandonments being permanent. In addition to the coins, 725 tokens were analyzed and it was found that 7% were abandoned at least once and 5% were ab

关键词： Computer science  

关键词： Applied mathematics   Information technology   Computer science  

摘要： The vast amount of personal data being collected and analyzed through internet connected devices is vulnerable to theft and misuse. Modern cryptography presents several powerful techniques that can help to solve the puzzle of how to harness data for use while at the same time protecting it---one such technique is homomorphic encryption that allows computations to be done on data while it is still encrypted. The question of security for homomorphic encryption relates to the broader field of lattice cryptography. Lattice cryptography is one of the main areas of cryptography that promises to be secure even against quantum computing. In this dissertation, we will touch on several aspects of homomorphic encryption and its security based on lattice cryptography. Our main contributions are: 1. proving some heuristics that are used in major results in the literature for controlling the error size in bootstrapping for fully homomorphic encryption, 2. presenting a new fully homomorphic encryption scheme that supports k-bit arbitrary operations and achieves an asymptotic ciphertext expansion of one, 3. thoroughly studying certain attacks against the Ring Learning with Errors problem, 4. precisely characterizing the performance of an algorithm for solving the Approximate Common Divisor problem.

关键词： Computer science  

摘要： This dissertation is about the development of a parallel software system for representing and solving problems of finite model theory and its application. The theoretical foundation of the system is presented, as well as an in-depth explanation of the implementation in Python. In particular, a parallel method for computing Boolean expressions based on the properties of finite free Boolean algebras is developed. It is also shown how various finite combinatorial objects can be coded in the formalism of Boolean algebras and counted by this procedure. Specifically, using a translation of first order predicate formulas to propositional formulas, we developed a technique for constructing and counting finite models of first order theories. Finally, we have developed some general techniques that enable more effective use of our system. We illustrate these techniques on two examples. The first one deals with partial orders, while the other one is about random graphs.

关键词： Computer science  

摘要： Legacy binaries that do not have source code remain a vital part of our software ecosystem. Lifting and recompilation of legacy binaries allows for a wide range of late program transformations such as security hardening, deobfuscation, and reoptimization even when the source code is unavailable. Existing binary lifting approaches are based on static binary disassembly which has several limitations. Distinguishing code from data statically, for example, is undecidable in the general case. Static disassembly must rely on heuristics and assumptions to disassemble binaries in the absence of dynamic information and high-level language semantics. Consequently, static disassembly cannot reliably handle indirect jumps, inline assembly, and obfuscated code. Lifting approaches that rely on static disassembly, therefore, often produce unsound binaries. Dynamic disassembly of binaries can circumvent the limitations of static disassembly including the ability to handle obfuscated and encrypted binary code. In this dissertation, we present BinRec, a new approach to heuristic-free binary recompilation which lifts dynamic traces of binaries to a compiler-level intermediate representation (IR); the lifted IR is lowered to a "recovered'' binary by taking advantage of the existing compiler toolchain. Our approach allows applying rich program transformations, such as compiler-based hardening and optimization passes, on top of the recovered representation. We identify and address a number of challenges in binary lifting, including unique challenges posed by our dynamic approach. In contrast to existing frameworks, our dynamic front end can accurately disassemble and lift binaries without heuristics, and we can successfully recover all SPEC INT 2006 benchmarks including C++ applications. We evaluate our approach in four application domains: I) binary reoptimization, ii) deobfuscation (by recovering partial program semantics from virtualization-obfuscated code), iii) binary hardening (by

关键词： Information technology   Information science   Computer science  

摘要： Microaggregation is a disclosure control method that uses k-anonymity to protect confidentiality in microdata while seeking minimal information loss. The problem is NP-hard. Iterated local search for microaggregation (ILSM) is an effective metaheuristic algorithm that consistently identifies better quality solutions than extant microaggregation methods. The present work presents improvements to local search, the perturbation operations and acceptance criterion within ILSM. The first, ILSMC, targets changed clusters within local search (LS) to avoid vast numbers of comparison tests, significantly reducing execution times. Second, a new probability distribution yields a better perturbation operator for most cases, significantly reducing the number of iterations needed to find similar quality solutions. A third improves the acceptance criterion by replacing the static balance between intensification and diversification with a dynamic balance. This helps ILSM escape local optima more quickly for some datasets and values of k. Experimental results with benchmark data show that ILSMC consistently reduces execution times significantly. Targeting changed clusters within LS avoids vast numbers of unproductive tests while allowing search to concentrate on more productive ones. Execution times are decreased by more than an order of magnitude for most benchmark test cases. In the worst case it decreased execution times by 75%. Advantageously, the biggest improvements were with the largest datasets. Perturbing clusters with higher information loss tend to reduce information loss more. Biasing the perturbation operations toward clusters with higher information loss increases the rate of improvement by more than 50 percent in the earliest iterations for two of the benchmarks. Occasionally accepting worse solutions provides diversification; however, increasing the probability of accepting worse solutions closer in quality to the current best solution aids in escaping local optima.

关键词： African American studies   Higher education   Computer science  

摘要： Voices of African American Women in Computer Science is a qualitative dissertation about African American women who have successfully completed a computer science degree. This research explored how computer science education departments in predominantly White institutions (PWIs) contribute to the oppression of African American females. Critical race feminism (CRF) provides the platform that places the voices of women of color who have successfully obtained a computer science degree despite difficulties that have been imposed by the reality of racial biases that are present throughout the American education system and American culture at large. This research was guided by the following primary question: What are the unique experiences that have influenced the success of African American females, who have obtained degrees in computer science from PWIs? In order to establish fulfilling narratives of the participants' experiences the following secondary questions are also relevant: What factors have influenced participants' decision to major in computer science? What were the experiences of the participants during their time in their computer science program? How do these women make meaning of their experiences? Both narrative analysis and analysis of narrative was used to interpret data gathered using detailed interviews. Findings were presented using both methods in an attempt to show a detailed use of data, as well as to also push the boundaries of CRF with the creation of guidelines for CRF methodology. The results show that African American women in computer science have unique but powerful stories. This study highlights both positive, and negative factors that influenced these women during and after their academic experiences and fills gaps left by previous research in this area. Overall, this study holds implications for K-12 and higher education institutions as well as policymakers, and those who would simply pick up the torch of mentorship.

关键词： Logic   Statistics   Artificial intelligence   Computer science  

摘要： As the twin movements of open science and open source bring an ever greater share of the scientific process into the digital realm, new opportunities arise for the meta-scientific study of science itself, including of data science and statistics. Future science will likely see machines play an active role in processing, organizing, and perhaps even creating scientific knowledge. To make this possible, large engineering efforts must be undertaken to transform scientific artifacts into useful computational resources, and conceptual advances must be made in the organization of scientific theories, models, experiments, and data. This dissertation takes steps toward digitizing and systematizing two major artifacts of data science, statistical models and data analyses. Using tools from algebra, particularly categorical logic, a precise analogy is drawn between models in statistics and logic, enabling statistical models to be seen as models of theories, in the logical sense. Statistical theories, being algebraic structures, are amenable to machine representation and are equipped with morphisms that formalize the relations between different statistical methods. Turning from mathematics to engineering, a software system for creating machine representations of data analyses, in the form of Python or R programs, is designed and implemented. The representations aim to capture the semantics of data analyses, independent of the programming language and libraries in which they are implemented.

关键词： Computer engineering   Artificial intelligence   Computer science  

摘要： Applications of machine learning have become increasingly common in recent years. For instance, navigation systems like Google Maps use machine learning to better predict traffic patterns; Facebook, LinkedIn, and other social media platforms use machine learning to customize user's news feeds. Central to all these systems is user data. However, the sensitive nature of the collected data has also led to a number of privacy concerns. Privacy-preserving machine learning enables systems that can perform such computation over sensitive data while protecting its privacy. In this dissertation, we focus on developing efficient protocols for machine learning as a target analytics application. To incorporate privacy, we use a multi-party computation-based approach. In multi-party computation, a number of non-colluding entities jointly perform computation over the data and privacy stems from no party having any information about the data being computed on. At the heart of this dissertation are three frameworks—SecureNN, FALCON, and Ponytail—each pushing the frontiers of privacy-preserving machine learning and propose novel approaches to protocol design. SecureNN and FALCON introduce, for the first time, highly efficient protocols for computation of non-linear functions (such as rectified linear unit, maxpool, batch-normalization) using purely modular arithmetic. Ponytail demonstrates the use of homomorphic encryption to significantly improve over prior art in private matrix multiplication. Each framework provides both significant asymptotic as well as concrete efficiency gains over prior work by improving computation as well as communication performance by an order of magnitude. These building blocks—matrix multiplication, rectified linear unit, maxpool, batch-normalization—are central to machine learning and improvements to these significantly improve upon prior art in private machine learning. Furthermore, each of these systems is implemented and benchmarked to reduce the ba