International Association for Cryptologic Research

International Association
for Cryptologic Research

IACR News item: 10 September 2014

PhD Database PhD Database
Name: Carolyn Whitnall
Topic: Statistical methods for non-profiled differential side-channel analysis: Theory and evaluation
Category: (no category)

Description:

\r\nDifferential side-channel analysis (DSCA) aims at recovering cryptographically-secured secret information by exploiting the relationship between the physically-observable characteristics of a device and the data manipulated inside it. Prior knowledge about this relationship (obtained, perhaps, by detailed examination of an equivalent device) is known to greatly enhance attack success. What may be achieved with little or no prior knowledge at all is less clear. Strategies designed on such a basis have been loosely termed `generic\', but the scenarios in which these are possible without some meaningful knowledge on the leakage appear rare.\r\n

\r\n\r\n

\r\nIn this thesis we formalise the notion of `generic DSCA\' in order to understand it better and to make concrete statements about when and in what sense it is possible. We confirm that the range of scenarios to which it may be applied truly is limited---requiring that the device at some stage implements a predictable function which is non-injective and sufficiently nonlinear (e.g. the DES S-Box transformations).\r\n

\r\n\r\n

\r\nWe explore popular proposals based on mutual information and other non-parametric statistics. To facilitate meaningful comparisons we first introduce a theoretic evaluation framework to enable like-for-like comparisons between different methods and avoid the pit-falls of (necessarily estimator-dependent) empirical comparisons. One of the lessons learned by employing this framework is that mutual information is indeed optimal in some information-theoretic sense (as was initially supposed) and that it is the added burden of estimation which makes it a poor choice in all but the most unusual of leakage scenarios.\r\n

\r\n\r\n

\r\nWe also analyse linear regression-based methods and their use as `generic\' strategies. Applied in this way, they are restricted to the same limited scope as any other such strategy. However, we identify a unique feature of the way they operate whi[...]

Expand

Additional news items may be found on the IACR news page.