In this paper, the authors present Heimdall, a tool for preference collection in a privacy-respecting manner. This tool can be implemented in IoT devices as well. Many applications, be it mobile or web applications rely on the users’ feedback and reviews for suggestions of online recommendation systems. These systems access to certain implicit information like location, purchase history and browsing history and also explicit information like reviews and ratings from the users.
Due to privacy concerns, many users are uncomfortable with the implicit data collection so the applications mainly have to rely on explicit feedback from the users. This over-dependence on explicit feedback from the users creates problems for the service providers in many ways. The amount of feedback received from the users is usually very small as compared to the amount of feedback required for quality recommendations. Since each explicit feedback requires an effort from the user ranging from short interaction like check-in, like, etc to long interaction like writing a review. This makes it difficult for current recommendation systems to accurately answer simple questions such as “What is the most popular theater in the area?” or “What time is the gym least crowded?”
One plausible solution for this problem is to enable recommendation systems to run a collector on a user’s device and precisely control the information a collector transmits to the recommendation system backend. Heimdall introduces immutable blobs as a mechanism to guarantee this property. We implemented Heimdall on the Android platform and wrote three example collectors to enhance recommendation systems with implicit feedback.
The authors evaluated the performance in three parameters: performance, can we develop useful collectors and does Heimdall decreases users’ concerns successfully. The performance results suggest that the overhead of immutable blobs is minimal, and a user study of 166 participants indicates that privacy concerns are significantly less when collectors record only specific information, which Heimdall does successfully.
Link to the paper: https://nsr.cse.buffalo.edu/mobisys_2017/papers/pdfs/mobisys17-paper31.pdf
Rishabh Jain's Corner 