Riding with the Stars: Passenger Privacy in the NYC Taxicab Dataset


Jessica Alba's taxi trip on 7 September, 2013 - now privatized. The blue square shows the actual pickup location. As the privacy requirements are eased, this square gets darker on average, indicating its increased likelihood of being the correct location. The fare and tip amounts below also are more likely to land at their true values.

Note that there are two amounts given below, and two red squares above that appear at high levels of ε. This is because our query, while very specific, actually returned two taxi trips, only one of which belonged to Jessica Alba. I used additional information (namely knowledge of the dropoff location) to identify the correct one above, but this example illustrates the use of differential privacy from the context of this query alone.