Poorvi Vora - Research: Applications of Information Theory to Problems in Security

This research was funded in part by HP Research Gift (2004) and NSF Awards 0505510, 0830576 and 0831149.
All of these publications are also listed as part of my research elsewhere on my website: voting [1,4,5], cryptanalysis [3,8] and privacy [2,6,7,9,10,16-20] research. Similarly with the invited presentations.
My formal training in communication theory motivates me to use information theory to understand security challenges.

My first attempt to do so was in research on inference attacks in database privacy, where I observed that data perturbation had the effect of a noisy channel, and hence that attacks would behave as error-correction codes [7]. This observation enabled the application of classical information-theoretic results to upper bound the complexity of an attack in number of queries per bit of entropy [1,6]. While this work also shows that it is possible to determine protected data values to any desired accuracy with a constant cost per bit, and is hence an impossibility result wrt data protection, I prefer to view it as a calibration of the cost of an attack, and hence a measure of the privacy of the data perturbation scheme. This view of privacy led to a proposal on "variable privacy" [10,19,20].

The above approach can also be used to understand all cryptanalysis as channel communication, and attacks as efficient channel codes [3,8,11-15]. The cryptanalysis work is joint with thesis masters' student Darakhshan Mir (now a doctoral student at Rutgers).

With Coney, Hall and Wagner, I proposed a privacy measure for information leakage in voting systems [5]. My doctoral student Ben Hosp has extended the use of entropy-related measures to measure the verifiability of voting systems, and applied his model to several voting system proposals [1,4,16].

Email: my first name at gwu dot edu
Last modified: 20:15:57, Tuesday, 26 March, 2013 local time