Balancing Value and Risk in Information Sharing Through Obfuscation
Abstract
Fast-paced data-to-decision systems are heavily dependent on the reliable sharing of sensor-derived information. At the same time a diverse collection of sensory information providers would want to exercise control over the information shared based on their perception of the risk of possible misuse due to sharing, and also depending on the consumer's requirements. To attain this utility vs. risk trade-off, information is subjected to varying but deliberate quality modifying transformations which we term as obfuscation. In this paper, treating privacy as the primary motivation for information control, we highlight initial considerations of using feature sharing as an obfuscation mechanism to control the inferences possible from shared sensory data. We provide results from an activity tracking scenario to illustrate the use of feature selection in identifying the various trade-off points.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 20, 2012
- Accession Number
- ADA565915
Entities
People
- Chatschik Bisdikian
- Kasturi R. Raghavan
- Lance Kaplan
- Supriyo Chakraborty
Organizations
- IBM Thomas J. Watson Research Center