Research seeks to enhance technology’s benefits, lower risks

Researchers strive to enhance technology’s benefits, lower risks
Thursday, May 23, 2024
Experts use AI technologies, varied techniques and methods

Technology can be useful, but it can also open the door to problems. UNC Charlotte researchers Dongsong Zhang and Lina Zhou are leaders in research that advances technology’s positive aspects and reduces its risks for individuals and organizations.

"The research we do is motivated by critical real-world problems with significant societal implications," said Zhang, Belk Distinguished Professor in Business Analytics. "We hope that our research can benefit people in their daily lives and stimulate new ideas from other people."

Zhou and Zhang are faculty in Belk College of Business’s Department of Business Information Systems and Operations Management, where their KAIM lab is housed. They have worked collaboratively for over two decades, conducting their research using machine learning, natural language processing, text mining, social network analysis, personalization techniques and behavioral research methods. In mid-May, Zhang was named interim executive director for UNC Charlotte's School of Data Science.

They have published hundreds of research articles, won multiple honors for their scholarship and engaged students in their research and through real-world illustrations and activities in their classrooms.

UNC Charlotte researchers Dongsong Zhang and Lina Zhou

When working together, Zhang and Zhou draw upon their varied perspectives and expertise. They also turn to the world around them for research inspiration and continue to evolve their ideas over time.

"Often I come up with a research idea through something I experience myself, or I put myself in other people’s shoes," Zhang said. "I was getting interested in mobile human-computer interaction back in 2003, for example, when we still use the old-fashioned mobile phones. So even at that time I started thinking about how to make user interaction with mobile phones easier."

The interest persisted, evolving into a project to develop a virtual keyboard for mobile devices that would allow users to enter text sight-free with one hand, aiding people with disabilities. Zhou thought about that particular effort from a different perspective, bringing in her experiences as a research scientist at the University of Arizona.

"I have this mindset of applications to improve cybersecurity," she said. "I saw the potential of the tool Dr. Zhang and his former Ph.D. student had developed, and I started thinking the idea could be used to improve the security and accessibility of mobile authentication. We were able to work on a grant proposal for submission to the National Science Foundation, and fortunately the project was funded."

Their lab seeks to design and develop systems to solve problems while also evaluating the broader impact of information technology on users, organizations, communities and society at large. They also want to understand and model user behaviors. Their current research is focused on three main topics:

  • Online deception and misinformation detection
  • Mobile device security
  • Health information technology

Online deception and misinformation detection

This particular research, funded in part by the NSF, has focused on understanding the deceptive behavior and strategies that bad actors use in online communication, including email, chats, discussion forums, social media and the like. Deception is at the heart of most attacks in online communication, including misinformation, and people are poor at detecting deception. Over the past 20 years, the researchers have used text analysis and machine learning techniques to develop new and effective ways to better detect deceptive communication automatically and predictively.

Not all the behaviors that online attackers use are known, but Zhang and Zhou believe that some behaviors are more difficult for attackers to control than others. By identifying such behaviors and their relations in online communication, the project lays the groundwork for developing resilient and predictive approaches to detect the attacks, and advances what is known about online deception behavior and its identification. In recent years, they have extended their research on online deception and misinformation detection to different contexts. This includes building multimodal models for detecting fake online reviews and fake news.

Currently, they are part of an ongoing project housed at N.C. State University and funded in part by the Department of Defense to develop a gamified training system to improve employees’ ability to detect spear phishing emails, which are designed to look legitimate to entice employees to reveal personal or sensitive information.

Mobile security

Mobile devices, particularly smartphones that function as small computers, are increasingly used in work and other aspects of our daily lives, and people increasingly store personal and sensitive data on these devices. Meanwhile, traditional login password protection can be insufficient and vulnerable to security attacks, including surveillance video and shoulder surfing, when people spy on users over their shoulders. This makes improving the security of smartphones critical.

Zhou and Zhang’s work, funded in part by the NSF, focuses on designing and developing a one-handed, keypress-less user-authentication approach that can implicitly and continuously verify if the current user of a smartphone is indeed the owner of the device based on how the user interacts with the touchscreen of the device, such as characteristics of finger swiping and typing.

Health Information Technologies

Another line of their research interest is leveraging information technologies in support of public health. In one recent project, they developed a knowledge-enhanced deep learning model to automatically detect suicidal ideation or depression expressed in the content of social media posts. This model would enable the social media platforms and mental health professionals to take a proactive role by reaching out to the individuals and offering timely intervention, instead of passively waiting for those individuals to seek help. Many people may never seek help due to social bias, lack of self-awareness and cost, the researchers said.

Zhou and Zhang, in collaboration with colleagues Rajib Paul and Ahmed Arif with UNC Charlotte’s College of Health and Human Services and Atrium Health, recently started work on a new project funded by the Centers for Disease Control and Prevention and American Public Health Association. The effort is focused on classifying and nowcasting — or very short range forecasting — of homicide incidents using natural language processing, deep learning and spatiotemporal techniques.