Lalitha Sankar, an associate professor of electrical engineering, has been researching game theoretic models to help retailers and service providers generate accurate purchase recommendations while guaranteeing consumer privacy. Photographer: Erika Gronek/ASU
New research to curb the surge of consumer privacy violations
By Amanda Stoneman
You’ve likely been bombarded with customized coupons and gift recommendations designed to steer you toward products and services you may be inclined to buy. Retailers and free service providers, like Facebook and Google, reap revenue with these targeted advertisements — but at the cost of your private data.
The increase in consumer privacy violations motivated Associate Professor Lalitha Sankar to develop game theoretic models retailers and service providers can use to help them generate accurate recommendations while guaranteeing consumer privacy.
“Recommendation systems are everywhere,” says Sankar. “How can these systems make recommendations without knowing who you are?”
Every time you shop, you reveal data about your purchasing behaviors. Whenever you use a free service, you implicitly consent to having your data collected, stored, sold and shared. The information gathered paints a picture about what you like, dislike and may need.
Understanding retailer and consumer interactions in the context of privacy
When a retailer automatically tracks a consumer’s financial transactions, purchasing behavior and preferences, it’s easier to offer customized incentives. But sometimes these incentives imply the retailer has learned sensitive or private information about the consumer.
Such was the case six years ago when Target figured out a young woman was pregnant and began sending pregnancy-related coupons to her home address, which were intercepted by her father before she ever got a chance to share the news.
“These privacy violations are making consumers very wary,” says Sankar. “So, how can retailers over the long-term maximize their profit without creeping out consumers?”
Sankar and her then-doctoral student Chong Huang, who is now a postdoctoral research associate in the School of Electrical, Computer and Energy Engineering, along with Anand Sarwate, an assistant professor at Rutgers University, proposed a mathematical framework for modeling decision-making so retailers can develop coupon-offering policies that earn revenue while being sensitive to consumer privacy concerns.
The framework builds on a well-known model for nondeterministic systems called partially observed Markov decision processes, which allows modeling of the fact that retailers don’t know how a consumer may react to a single coupon but can improve their understanding of consumers based on numerous interactions.
The work could enable retailers and free service providers to use consumers’ responses — to sponsored ads, for example — to better learn and respect consumer privacy sensitivities.
Privacy consequences of using free online services
Sankar and Huang have also been researching the impact of privacy on free online service markets, such as social media, search engines and mobile applications.
While consumers enjoy the benefits of free services and customized recommendations, privacy violations are occurring more frequently.
“Today, machine learning algorithms and designers have to be cognizant when taking data from a variety of entities about whether they are violating individual privacy or inferring information they shouldn’t be inferring,” says Sankar.
Service providers like Google and Facebook are able to gather such tailored information about users because of their search queries. So, the research team began asking the question: What if there was a competitor to free service providers, such as Google and Facebook?
Would consumers prefer a competitor that offers less targeted advertisements and search results if they could guarantee more privacy? The competitor wouldn’t target ads at users based on their personal information the way data giants currently do.
Sankar and Huang propose a game-theoretic approach to identify if privacy-differentiated free online services can lead to a sustainable marketplace and a meaningful market share for service providers that offer privacy guarantees.
“Maybe no one cares about recommendation systems from Netflix,” says Sankar. “But what if an enterprise has your DNA information? What if it uses it to recommend insurance policies, medical treatment or even leads to a denial of insurance coverage?”
Privacy awareness for middle and high school students
Sankar recognizes that privacy concerns may be less worrisome to the younger generation of online users who have grown up with the internet inextricably woven into their daily lives.
She partnered with a teacher at Xavier College Preparatory in Phoenix to host outreach events focused on the social implications of not having good privacy settings on social media sites.
Using characters from popular movies, young girls can grasp the importance of how privacy settings should be set up to ensure private information doesn’t get shared with unknown recipients. They also learned about location privacy and how geo-tagging can be dangerous.
“It’s important for girls to know how to set up their privacy settings,” says Sankar, “especially now with the onset of cyberbullying. This age group is much more likely to be affected.”