Whether human or digital, the sacrifice you make for having a personal assistant is privacy. Until now, that is. Researchers from the Department of Computer Science developed a digital assistant app that does not compromise privacy and answers the question, how can we provide personalized news recommendations without sharing sensitive data with the provider?

Digital personal assistants have the ability to gear the content of an app or service to your specific likes and dislikes, such as Google News. However, they need a lot of sensitive data about you to be effective. For the past year, assistant professors Aruna Balasubramanian and Niranjan Balasubramanian, former grad students Shashank Jain (Microsoft) and Vivek Tiwari (LinkedIn) to develop an app that can deliver personalized recommendations for news articles, without collecting and sending sensitive user data to a cloud service.

Dubbed PrIA (Private Intelligent Assistant), maintains your privacy by not using remote servers to organize a personal user profile, but instead does so locally on your smartphone and laptop. This ensures that personal information is kept private.

The research team conducted a ten day study of their application and found that Google News, which uses private user data and is a comparable cloud based service, was only 14% better at suggesting news articles to its users. According to Niranjan Balasubramanian, for privacy-minded users this may be a worthy trade-off. PrIA downloads stories from Google News, but does so without signing into a Google account or sending user’s information elsewhere.

“The important thing is, only your phone and your laptop have this information,” says Aruna Balasubramanian.

Running on the user’s laptop, the implementation of PrIA consisted of these applications:

1. Logger – This collected web browsing history and the user’s Twitter and Facebook feeds.
2. Profile Builder – A user profile graph is built on the user’s laptop. Using the Stanford Named Entity tagger, 20 topics were chosen and added to the profile graph.
3. PPR Recommendation - A recommendation algorithm downloads all headline articles from Google News. PrIA reviews the candidate articles and uses the recommendation article to push articles to the user’s smartphone.

An article about the application was recently published in the MIT Technology Review, one of the premier and most prestigious publications for tech research to be announced. The research paper, which was co-authored by Supriyo Chakaraborthy from IBM Research, was also presented at the 18th Annual International Workshop on Mobile Computing Systems and Applications in Sonoma, California.

About the CS Researchers

Aruna and Niranjan Balasubramanian are faculty in the Department of Computer Science within the College of Engineering and Applied Sciences at Stony Brook University. Aruna’s research takes place in her NetSYS Lab which focuses on networking and mobile computing systems. Niranjan, who is also affiliated with the Department of Biomedical Informatics and the Center of Excellence in Wireless & Information Technology (CEWIT), explores natural language processing and information retrieval. Both researchers earned their PhD from the University of Massachusetts, Amherst and joined Stony Brook in 2015.