Traditional email encryption methods are too difficult to setup, requiring senders to obtain message recipient's public keys before secure communication can be sent. Easy Accessible Secure Email (EASEmail) seeks to solve the key exchange issues of encrypted email by using a lightweight symmetric key server for key exchange. Users can send mail to arbitrary recipients without directly establishing or exchanging keys with the recipient. With usability as its primary goal, EASEmail strives to bring usable secure email communication to the masses.

Libraries, private or public, offer valuable resources to library patrons. As of today the only way to locate information archived exclusively in libraries is by searching their catalogs. Library patrons, however, often find it difficult to formulate a proper query, which requires using rigid subject terms chosen by the Library of Congress, to obtain relevant results. These improperly formulated queries often yield irrelevant results or no results at all. This negative experience in dealing with existing library systems turn library patrons away from library catalogs; instead, they rely on Web search engines to perform their searches first and upon obtaining the initial information (such as book/manuscript titles, subject areas, authors, etc.) on the desired library materials, they consult the library catalog. This searching phenomenon is an evidence of failure of today’s library systems. To solve this problem, we propose an enhanced library system, which relies on the Fuzzy Set Information Retrieval model which allows partial, similarity matching of (i) descriptive tags that describe the content of books recommended by ordinary users at a folksonomy site and (ii) keywords in a user’s query to improve the searches performed on existing library catalogs. The proposed library system allows patrons to post a query Q using commonly-used words and ranks the retrieved results according to their degrees of resemblance with Q while maintaining the query processing time comparable with existing Web search engines.

The ability to delegate privileges to others is so important to users of online identity systems that users create ad hoc delegation systems by sharing authentication credentials if no other easy delegation mechanism is available. With the rise of internet-scale relationship-based single sign-on protocols like OpenID, the security risks of password sharing are unacceptable. We therefore propose SimpleAuth, a simple modification to relationship-based authentication protocols that gives users a secure way to selectively delegate subsets of their privileges, making identity systems more flexible and increasing user security. We also present a proof-of-concept implementation of the SimpleAuth pattern using the sSRP authentication protocol to demonstrate the generality of our technique.

Abstract for Return on Investment for Active Learning:

Active Learning (AL) has been defined as a learning protocol intended to present those data to an oracle for labeling which will be most enlightening for machine learning. While AL traditionally accounts for the value of the information obtained, it often ignores the cost of obtaining
the information thus causing it to perform sub-optimally with respect to total cost. We present a framework to enhance the notion of AL to account for the expected cost of the samples to be obtained. We discuss optimality and tractability in this framework. We then propose new selection algorithms for AL built upon exist-ing AL algorithms by accounting for cost using a metric that captures return on in-vestment. We demonstrate the power of our algorithms in the context of the crea-tion of labeled corpora.

Learning platforms are a great way to give students a laboratory-like experience, allowing students to experiment with different ideas and see the results. They are used widely in education to help students grasp difficult concepts related to the design of intelligent decision algorithms. Traditionally, platforms in this area have been physics oriented, such as inverted pendula, ball and beam systems, or robot soccer. We propose a novel decomposition of decision algorithms into four fundamental problems: decision making, learning, model approximation, and verification. We then introduce three non-traditional learning platforms in portfolio optimization, business intelligence, and automated water management, and show how each one illustrates these four fundamental issues for understanding algorithmic decision processes.

Due to the inherent complexity of software, researchers need tools to help them understand software systems and development processes. Software development artifacts (such as source code, emails, and bug reports) are excellent data sources for these tools. Therefore, the SEQuOIA lab at BYU is developing an artifact-based analysis and visualization tool designed for researchers and practitioners. One set of particularly important artifacts, revision control systems, contain the history of file modifications for a project. Unfortunately, current tools to extract data from these artifacts do not capture the full range of data available. We propose the creation of extraction tools and relational database schemas to store data from the CVS, SVN, and Bazaar revision control systems. Our proposed tool will be specifically designed as a part of the SEQuOIA architecture, and will capture core revision control data and textual details not present in existing solutions. Extraction tools will be written in Python and leverage a database-independent object relation manager. To verify our approach, we will 1) populate repositories with test data and conduct extensive unit testing against those repositories and 2) use our tools to extract data from a number of open source projects.