8/21/2021 Aaron Seidlitz
Professors Dakshita Khurana, Hari Sundaram, Darko Marinov, Tianyin Xu, Lingming Zhang, and Tao Xie and their graduate students were the latest CS researchers to earn best or distinguished paper awards in the spring and summer of 2021.
Written by Aaron Seidlitz
CS faculty garnered four best or distinguished paper awards at prominent national and international conferences during the spring and summer of 2021, once again raising the profile of the department’s broad and impactful research.
Dakshita Khurana’s paper “One-Way Functions Imply Secure Computation in a Quantum World” received a Long Plenary talk award at QIP 2021, a premier international conference for theoretical quantum information research. Only two papers out of 462 submissions received this recognition, which is equivalent to a best paper award.
Secure computation is a cryptography technology that allows collaborative data analysis without revealing any private information from individuals in the process. One of the goals of modern cryptography is to realize secure computation under the simplest possible cryptographic assumptions, which is difficult to attain with classical computing resources.
However, researchers have discovered that quantum key distribution has opened up the tantalizing possibility of using quantum resources to enable secure computation assuming the existence of one-way functions.
Khurana and her co-authors—James Bartusek (Berkeley), Andrea Coladangelo (Berkeley), and Fermi Ma (Princeton)—showed in their paper that general-purpose secure computation can be realized assuming the existence of one-way functions and assuming a few simple quantum capabilities. Their protocols use only simple quantum communication as in the quantum key distribution protocols of Bennett and Brassard, making them amenable to deployment in the foreseeable future as quantum communication becomes a reality.
CS professor Hari Sundaram received a Best Article Award from the Journal of Interactive Advertising for the paper “Computationally Analyzing Social Media Text for Topics: A Primer for Advertising Researchers.” The paper was co-authored by colleagues from the U of I and University of Georgia advertising departments.
An essential component of all organizations’ marketing and communication strategy is understanding how social media users talk about their brand. However, using the algorithms to identify topics of social media conversations requires a technical background.
Smaller organizations often lack the technical expertise or large budgets to sift through the enormous data sets generated by social media posts. They find identifying actionable knowledge from the data complex.
CS alumnus Joseph Yun (BS CS '01), a research professor in the Illinois Gies College of Business, was first author of the article, which provides an accessible introduction to computational techniques developed to analyze document collections for topics. Methods include text preprocessed summarization, phrase mining, topic modeling, supervised machine-learned text classification, and semantic topic tagging.
The article presents the advantages and possible challenges of using each technique. By applying the computational techniques on a real-world Twitter dataset of a brand, the authors explain how the techniques perform and how they would compare against the work of a human analyst. The paper also provides a reference matrix that summarizes each analysis method, explaining when a company or organization wants to use each technique and an overview of tools to execute each technique.
Illinois CS researchers in the Programming Languages, Formal Methods, and Software Engineering area recently won two ACM SIGSOFT Distinguished Paper awards at esteemed international conferences. These awards are presented to fewer than 10 percent of all the papers accepted at an ACM SIGSOFT-sponsored conference.
CS graduate student Runxiang (Sam) Cheng received an ACM SIGSOFT Distinguished Paper Award at the International Symposium on Software Testing and Analysis (ISSTA’21).
His “Test-Case Prioritization for Configuration Testing” paper demonstrated an efficient way to test configuration changes for DevOps-based continuous deployment. Modern cloud systems (e.g., Facebook and Google) perform configuration changes hundreds of times a day and therefore need a fast and reliable way to check those changes to prevent production failures.
Cheng, his co-advisor Tianyin Xu, and their collaborators previously introduced using configuration tests (“ctests” for short) to detect misconfigurations, but running ctests takes a long time.
Cheng and his co-authors—CS faculty Lingming Zhang, Darko Marinov, and Xu—applied traditional test-case prioritization (TCP) methods to reorder ctests. After evaluating 84 traditional and novel ctest-specific TCP techniques on five widely used cloud projects, they discovered that TCP can substantially speed up misconfiguration detection.
CS graduate student Wenyu Wang won an ACM SIGSOFT Distinguished Paper Award from the Joint European Software Engineering Conference and Symposium on the Foundations of Software Engineering (ESEC/FSE’21). He co-authored the paper “Vet: Identifying and Avoiding UI Exploration Tarpits,” with his faculty co-advisors Tianyin Xu and Tao Xie and the University of Texas at Dallas faculty member Wei Yang who did his PhD at Illinois.
Vet is an effective technique for enhancing mobile user-interface (UI) testing by automatically addressing UI exploration tarpits. An exploration tarpit refers to a scenario where testing gets stuck with a small fraction of app functionalities for a long period of time (e.g., a testing tool logs out an app at early stages without being able to log back in). Exploration tarpits are major causes of ineffectiveness in existing mobile UI testing.
Wang and his co-authors applied Vet to enhance three state-of-the-art Android UI testing tools. Vet identified exploration tarpits that cost up to 98.6% of testing time budgets. These exploration tarpits reveal not only limitations in UI exploration designs, but also defects in tool implementations. Vet automatically avoided the tarpits during testing, achieving high code coverage and improving crash-triggering capabilities.
PhD student Qingyun Wang led a collaborative team of researchers to a Best Paper Demo Award at the 2021 Annual Conference of the North American Chapter of the Association for Computational Linguistics (NAACL). The paper was titled “COVID-19 Literature Knowledge Graph Construction and Drug Repurposing Report Generation.”
Wang worked with Illinois CS professors Heng Ji, his adviser, and Jiawei Han, who both focus research efforts on Artificial Intelligence and Data and Information Systems. Many other Illinois CS PhD students contributed to the paper, along with external collaborators at Columbia University; Brandeis University; University of Washington; University of California, Los Angles; Colorado University; Army Research Lab; QS2; and the Department of Defense.
The paper they produced responds to a need clinicians and scientists have exhibited since the outset of the COVID-19 pandemic. To help combat the disease, they need to digest vast amounts of relevant biomedical knowledge in scientific literature.
Through a novel and comprehensive knowledge discovery framework, COVID-KG, these researchers extracted fine-grained multimedia knowledge elements from scientific literature. The group then exploited the constructed multimedia knowledge graphs for question answering and report generation, using drug repurposing as a case study. All of the data, KGs, reports, resources and shared services are publicly available.