Dear colleagues,

This is a reminder of today's Scientific Meeting:

Speakers: Jeroen Witteveen (Scientific Computing group) and Serge Fehr (Cryptology group)
Time: Thursday, April 4, 13.00 - 14.00
Place: Room Z009 (Euler)

Titles and abstracts are given below.  After the two talks, Tijs van der Storm will speak for a few minutes about the CWI internship program.

Note that the meeting is on a Thursday rather than the usual Friday.
It is a lunchtime meeting, so sandwiches will be provided before the talks.  We hope to see you there!

Best regards,
Willem Hundsdorfer and Ronald de Wolf



13.00-13.30: Jeroen Witteveen
Title: Uncertainty Quantification in Computational Fluid Dynamics
Abstract:
Computational Fluid Dynamics (CFD) is the field that simulates fluid flow around objects by solving partial differential equations on supercomputers. Examples of the objects in this context range from wind turbines to airplanes and internal combustion engines. In the computer implementation of the mathematical description of these problems, a number of parameter values needs to be posed. These parameters include, for instance, model parameters, initial conditions, and boundary conditions for the geometry or the undisturbed flow field. However, in practice these input parameter values are often not exactly known due to fluctuating atmospheric wind conditions, measurement inaccuracies, and production tolerances. If the system is sensitive to these sources of variability, then small parameter changes can already have a large impact on the numerical prediction. This motivates the stochastic description of these parametric uncertainties using probability distributions. Uncertainty quantification algorithms will be discussed for propagating these probability distributions through the computational problem in stochastic flow simulations. These methods are based on a robust interpolation of a small number of deterministic solutions for different parameter values in the range of the input distributions.

13.30-14.00: Serge Fehr
Title: Robust Secret Sharing
Abstract:
The security of cryptographic schemes is typically very centralized: knowledge of a single digital key is necessary and sufficient to, say, decrypt a ciphertext, or digitally sign a document. An important tool to decentralize security is "secret sharing": we want to split up a "secret" (e.g. a cryptographic key) into several pieces, called "shares", so that with sufficiently many of the shares, the secret can be reconstructed, but fewer shares give no information on the secret. For example, a bank director wants to share the vault combination among
his staff so that any 5 staff members can recover the combination and open the vault, but any 4 staff members alone have no information on the combination at all; note that it is not good enough to just distribute the different digits of the combination. If in addition, we want to protect against dishonest participants that lie about their shares in order to sabotage the reconstruction of the secret, we get to the notion of "robust secret sharing". In this talk, I give a brief overview over previous (ordinary and robust) secret sharing schemes, and I discuss a new scheme that is more efficient than previous ones.