Engineering News Fall 2016

Each day, millions of consumers slog through seemingly endless Amazon customer reviews, trusting online communities guide them in making the very best purchases, and Yuhong Liu, assistant professor of computer engineering, helps ensure that trust is well placed.

"Overall, my research is about trust," she said. "I borrow the sociological concept of how humans build trust among each other and apply it to the computer world—human to machine, and machine to machine. As social media has grown, people have gotten used to generating online content and trusting online sources for news and information. But online product rating is not just social; today, individual users have more power and influence—they write every word, every line. People enjoy sharing their knowledge and direct experience to help others make their buying decisions." But how can you tell which reviewers to trust?

Unfortunately, a whole cottage industry has popped up to take advantage of the trust being built within the online shopping community. Companies or individuals providing "reputation services" are paid to create multiple user reviews to boost a client's rating scores or denigrate those of a competitor. "Smart people are using multiple strategies to manipulate the ratings," Liu said.

So Liu is putting machines to use to identify fake online product reviews, and she has enlisted the help of Yu Wang, a former accountant and senior auditor with Price Waterhouse Cooper in Beijing who is now pursuing a master's degree in computer engineering at SCU.

An accountant might seem an offbeat choice to add to Liu's research team, but no. "I like collaborating with people from different disciplines," Liu said. "Looking at things from a different angle enhances my research." And it didn't hurt that Wang was the top student in Liu's Intro to Algorithms class.

Together, the two have developed an algorithm that will help ferret out suspicious activity by focusing solely on numerical ratings. They start with the basic assumption that a product or company's quality doesn't change overnight. "From a statistical point of view, ratings from honest reviewers should reflect reality without showing great fluctuations. A change in rating distributions may indicate anomaly. We have developed a change detector," said Liu.

Detecting dubious ratings is just one of Liu's research projects. Her team of four graduate students and two undergraduates are also examining the hot topic of privacy issues in social media and applying trust theory in cloud computing. "Social media plays a very important role in daily life, and cloud computing has tremendous potential but its rapid adoption heavily relies on its trustworthiness. Our research in these areas is interesting and important," Liu said.