Secrecy of Sequential Decision-Making
The increasing prevalence of large-scale surveillance and data collection infrastructures deployed by government agencies and private companies has brought global attention to the astonishing power enabled by modern cyber technologies. While such information appears to be revealing (e.g., a consumer's past browsing behavior may be indicative of the final purchase decision), we still lack a satisfactory understanding of the true value of the data collected, in terms of the extent to which it allows one to predict an individual's intention or future behavior using his or her past actions. This project aims to create a new mathematical framework to quantify the fundamental degree of information leakage asociated with an individual's sequential decision-making process, as well as design intelligent algorithms and decision-making policies that are capable of obfuscating an individual's future actions, even against a powerful data collector.
Tsitsiklis, John and Xu, Kuang, Delay-Predictability Tradeoffs in Reaching a Secret Goal (August 13, 2017). Stanford University Graduate School of Business Research Paper No. 16-25. Available at SSRN: https://ssrn.com/abstract=2784502