As most dedicated observers of voting bodies like the U.S. Supreme Court can attest, it is possible to guess vote outcomes based on statements made during deliberations or questioning by the voting members. In most forms of representative democracy, citizens can actively petition or lobby their representatives, and that often means understanding their intentions to vote for or against an issue of interest. In some U.S. state legislators, professional lobby groups and dedicated press members are highly informed and engaged, but the process is basically closed to ordinary citizens because they do not have enough background and familiarity with the issue, the legislator or the entire process. Our working hypothesis is that verbal utterances made during the legislative process by elected representatives can indicate their intent on a future vote, and therefore can be used to automatically predict said vote to a significant degree. In this research, we examine thousands of hours of legislative deliberations from the California state legislature’s 2015-2016 session to form models of voting behavior for each legislator and use them to train classifiers and predict the votes that occur subsequently. We can achieve legislator vote prediction accuracies as high as 83%. For bill vote prediction, our model can achieve 76% accuracy with an F1 score of 0.83 for balanced bill training data.

Traditional marketing models are swiftly being upended by the advent of online social networks. Yet, practicing firms that are engaging with online social networks neither have a reliable theory nor sufficient practical experience to make sense of the phenomenon. Extant theory in particular is based on observations of the real world, and may thus not apply to online social networks. Practicing firms may consequently be misallocating a large amount of resources, simply because they do not know how the online social networks with which they interact are organized. The purpose of this dissertation is to investigate how online social networks that are in stark contrast to real-world social networks behave and how they get organized. In particular, I explore how network structure and information flow within the network impact each other, and how they affect the phenomenon of influence in online social networks. I have collected retrospective data from Twitter conversations about six YouTube product categories (Music, Entertainment, Comedy, Science, Howto and Sports) in continuous time for a period of three months. Measures of network structure (Scale Free Metric, Assortativity and Small World Metric), network flows (Total Paths, Total Shortest Paths, Graph Diameter, Average Path Length, and Average Geodesic Length) and influence (Eigenvector Centrality/Centralization) were computed from the data. Experimental measures such as power law distributions of paths, shortest paths and nodal eigenvector centrality were introduced to account for node-level structure. Factor analysis and regression analysis were used to analyze the data and generate results. The research conducted in this dissertation has yielded three significant findings. 1. Network structure impacts network information flow, and conversely; network flow and network structure impact the network phenomenon of influence. However, the impact of network structure and network flow on influence could not be identified in all instances, suggesting that it cannot be taken for granted. 2. The nature of influence within a social network cannot be understood just by analyzing undirected or directed networks. The behavioral traits of individuals within the network can be deduced by analyzing how information is propagated throughout the network and how it is consumed. 3. An increase or decrease in the scale of a network leads to the observation of different organizational processes, which are most likely driven by very different social phenomena. Social theories that were developed from observing real-world networks of a relatively small scale (hundreds or thousands of people) consequently do not necessarily apply to online social networks, which can exhibit significantly larger scale (tens of thousands or millions of people). The primary contribution of this dissertation is an enhanced understanding of how online social networks, which exhibit contrasting characteristics to social networks that have been observed in the real world, behave and how they get organized. The empirical findings of this dissertation may allow practicing managers that engage with online social networks to allocate resources more effectively, especially in marketing. The primary limitations of this research are the inability to identify the causes of change within networks, glean demographic information and generalize across contexts. These limitations can all be overcome by follow-on studies of networks that operate in different contexts. In particular, further study of a variety of online social networks that operate on different social networking platforms would determine the extent to which the findings of this dissertation are generalizable to other online social networks. Conclusions drawn from an aggregation of these studies could serve as the foundation of a more broadly-based theory of online social networks.

The survival of products in online services such as Google Play, Yelp, Facebook and Amazon, is contingent on their search rank. This, along with the social impact of such services, has also turned them into a lucrative medium for fraudulently influencing public opinion. Motivated by the need to aggressively promote products, communities that specialize in social network fraud (e.g., fake opinions and reviews, likes, followers, app installs) have emerged, to create a black market for fraudulent search optimization. Fraudulent product developers exploit these communities to hire teams of workers willing and able to commit fraud collectively, emulating realistic, spontaneous activities from unrelated people. We call this behavior “search rank fraud”. In this dissertation, we argue that fraud needs to be proactively discouraged and prevented, instead of only reactively detected and filtered. We introduce two novel approaches to discourage search rank fraud in online systems. First, we detect fraud in real-time, when it is posted, and impose resource consuming penalties on the devices that post activities. We introduce and leverage several novel concepts that include (i) stateless, verifiable computational puzzles that impose minimal performance overhead, but enable the efficient verification of their authenticity, (ii) a real-time, graph based solution to assign fraud scores to user activities, and (iii) mechanisms to dynamically adjust puzzle difficulty levels based on fraud scores and the computational capabilities of devices. In a second approach, we introduce the problem of fraud de-anonymization: reveal the crowdsourcing site accounts of the people who post large amounts of fraud, thus their bank accounts, and provide compelling evidence of fraud to the users of products that they promote. We investigate the ability of our solutions to ensure that fraud does not pay off.

Social network data analysis becomes increasingly important today. In order to improve the integration and reuse of their data, many social networks start to apply RDF to present the data. Accordingly, one common approach for social network data analysis is to employ SPARQL to query RDF data. As the sizes of social networks expand rapidly, queries need to be executed in parallel such as using the MapReduce framework. However, the state-of-the-art translation from SPARQL queries to MapReduce jobs mainly follows a two layer rule, in which SPARQL is first translated to SQL join, is not efficient. In this thesis, we introduce two primitives to enable automatic translation from SPARQL to MapReduce, and to enable efficient execution of the SPARQL queries. We use multiple-join-with-filter to substitute traditional SQL multiple join when feasible, and merge different stages in the MapReduce query workflow. The evaluation on social network benchmarks shows that these two primitives can achieve up to 2x speedup in query running time compared with the original two layer scheme.

An important challenge in mobile sensor networks is to enable energy-efficient communication over a diversity of distances while being robust to wireless effects caused by node mobility. In this thesis, we argue that the pairing of two complementary radios with heterogeneous range characteristics enables greater range and interference diversity at lower energy cost than a single radio. We make three contributions towards the design of such multi-radio mobile sensor systems. First, we present the design of a novel reinforcement learning-based link layer algorithm that continually learns channel characteristics and dynamically decides when to switch between radios. Second, we describe a simple protocol that translates the benefits of the adaptive link layer into practice in an energy-efficient manner. Third, we present the design of Arthropod, a mote-class sensor platform that combines two such heterogeneous radios (XE1205 and CC2420) and our implementation of the Q-learning based switching protocol in TinyOS 2.0. Using experiments conducted in a variety of urban and forested environments, we show that our system achieves up to 52% energy gains over a single radio system while handling node mobility. Our results also show that our system can handle short, medium and long-term wireless interference in such environments.

All engineering careers require some level of programming proficiency. However, beginning programming classes are challenging for many students. Difficulties have been well-documented and contribute to high drop-out rates which prevent students from pursuing engineering. While many approaches have been tried to improve the performance of students and reduce the dropout rate, continued work is needed. This research seeks to re-examine what items are critical for programming education and how those might inform what is taught in introductory programming classes (CS1). Following trends coming from accreditation and academic boards on the importance of professional skills, we desire to rank knowledge and professional skill areas in one list. While programming curricula focus almost exclusively on knowledge areas, integrating critical professional skill areas could provide students with a better high-level understanding of what engineering encompasses. Enhancing the current knowledge centric syllabi with critical professional skills should allow students to have better visibility into what an engineering job might be like at the earliest classes in the engineering degree. To define our list of important professional skills, we use a two-group, three-round Delphi survey to build consensus ranked lists of knowledge and professional skill areas from industry and academic experts. Performing a gap analysis between the expert groups shows that industry experts focus more on professional skills then their academic counterparts. We use this resulting list to recommend ways to further integrate professional skills into engineering programming curriculum.

Strategies for software development often slight security-related considerations, due to the difficulty of developing realizable requirements, identifying and applying appropriate techniques, and teaching secure design. This work describes a three-part strategy for addressing these concerns. Part 1 provides detailed questions, derived from a two-level characterization of system security based on work by Chung et. al., to elicit precise requirements. Part 2 uses a novel framework for relating this characterization to previously published strategies, or patterns, for secure software development. Included case studies suggest the framework's effectiveness, involving the application of three patterns for secure design (Limited View, Role-Based Access Control, Secure State Machine) to a production system for document management. Part 3 presents teaching modules to introduce patterns into lower-division computer science courses. Five modules, integer over ow, input validation, HTTPS, les access, and SQL injection, are proposed for conveying an aware of security patterns and their value in software development.

Unmanned Aerial System (UAS) is a versatile and essential tool for law enforcement, first responders, utility providers, and the general public. Integrating the UAS into the National Airspace System (NAS) poses a significant challenge to policymakers and manufacturers. A UAS Integration Safety and Security Technology Ontology (ISSTO) has been developed in the Web Ontology Language (OWL) to aid in this integration. ISSTO is a domain ontology covering aviation topics corresponding to flights, aircraft types, manufacturers, temporal/spatial, waivers and authorizations, track data, NAS facilities, air traffic control advisories, weather phenomena, surveillance and security equipment, and events, sensor types, radio frequency ranges, actions, and outcomes. As ISSTO is a domain ontology, it models the current state of UAS integration into the NAS and provides a comprehensive view of every aspect of UAS.

In most video games today that focus on a single story, scripting languages are used for controlling the artificial intelligence of the virtual actors. While scripting is a great tool for reliably performing a story, it has many disadvantages; mainly, it is limited by only being able to respond to those situations that were explicitly declared, causing unreliable responses to unknown situations, and the believability of the virtual actor is hindered by possible conflicts between scripted actions and appropriate responses as perceived by the viewer. This paper presents a novel method of storytelling by manipulating the environment, whether physically or the agent's perception of it, around the goals and behaviors of the virtual actor in order to advance the story rather than controlling the virtual actor explicitly. The virtual actor in this method is completely autonomous and the environment is manipulated by a story manager so that the virtual actor chooses to satisfy its goals in accordance with the direction of the story. Comparisons are made between scripting, traditional autonomy, Lionhead Studio's Black & White, Mateas and Stern's Façade, and autonomy with environmental manipulation in terms of design, performance, believability, and reusability. It was concluded that molding an environment around a virtual actor with the help of a story manager gives the actor the ability to reliably perform both event-based stories while preserving the believability and reusability of the actor and environment. While autonomous actors have traditionally been used solely for emergent storytelling, this new storytelling method enables them to be used reliably and efficiently to tell event-based stories as well while reaping the benefits of their autonomous nature. In addition, the separation of the virtual actors from the environment and story manager in terms of design promotes a cleaner, reusable architecture that also allows for independent development and improvement. By modeling artificial intelligence design after Herbert Simon's “artifact,” emphasizing the encapsulation of the inner mechanisms of virtual actors, the next era of digital storytelling can be driven by the design and development of reusable storytelling components and the interaction between the virtual actor and its environment.

Authentication is the vital link between your real self and your digital self. As our digital selves become ever more powerful, the price of failing authentication grows. The most common authentication protocols are static data and employed only once at login. This allows for authentication to be spoofed just once to gain access to an entire user session. Behaviometric protocols continuously consume a user’s behavior as a token of authentication and can be applied throughout a session, thereby eliminating a fixed token to spoof. Research into these protocols as viable forms of authentication is relatively recent and is being conducted on a variety of data sources, features and classification schemes. This work proposes an extensible research framework to aid the systemization and preservation of research in this field by standardizing the interface for raw data collection, processing and interpretation. Specifically, this framework contributes transparent management of data collection and persistence, the presentation of past research in a highly configurable and extensible form, and the standardization of data forms to enhance innovative reuse and comparative analysis of prior research.