[crypto] [secdam at tue.nl: minicourse by Gabor Tardos October 4-6, 2009]

R. Hirschfeld ray at unipay.nl
Thu Jun 25 23:22:29 CEST 2009


------- Start of forwarded message -------
From: Secretariaat DAM <secdam at tue.nl>
Date: Mon, 22 Jun 2009 15:18:14 +0200
Subject: minicourse by Gabor Tardos October 4-6, 2009

Dear all,

Gabor Tardos will give a mini course on October 4-6, 2009 at the TU in Eindhoven.

Registration will be possible in a few weeks on the Eidma website. Please find below an abstract:


Traitor tracing



When fighting intellectual piracy one can go after the illegitimate user of the information or alternately the legitimate user that leaked it. Here we deal with this latter approach by hiding a different serial number (or "fingerprint") in each copy of the document. Hiding is crucial to prevent the user from finding and arbitrarily modifying the fingerprint. But even if we can hide the code perfectly the scheme might be vulnerable to collusion attacks: if several malicious users (the "pirates") compare their copies any difference they find indicates the hidden code and can later be used to erase or alter the code. The goal of fingerprinting is to design codes that are secure against the collusion attack of a limited number of pirates, that is the code distributor should be able to identify at least one of the colluding pirates with high probability from observing the copy of the document that was leaked.



Several mathematical models were suggested to capture this scenario.

In a continuous model we assume the information consists of real number parameters that can be modified a little without altering its content. We will mostly concentrate however to the digital model, where the information is a long string of digits and the fingerprint is hidden as some of these digits. We assume that the pirates will only find those digits of the fingerprint that are revealed by comparing their copies of the document. In other words they can alter the fingerprint at positions where not all of their codewords agree but cannot elsewhere. This assumption is called the "marking assumption". Further distinction in the model is whether in positions where the codewords they receive differ they can modify the codeword arbitrarily or can only use one of the several digits their codewords have at that position. This last distinction turns out to be important because in the model more restrictive to the pirates a deterministic solution is possible over larger alphabets!
 , while in the other model increasing the alphabet size does not help and only probabilistic solutions exist.



In the course we will survey several code constructions from simplest to most efficient (low error, short code). We will also survey techniques upper bounding the rates achievable in the marking assumption model showing that our codes are close to optimal and are asymptotically optimal.



A broad array of mathematical techniques will be used from information theory to game theory and combinatorics but the course should be self contained and accessible for anyone interested and having sufficient mathematical maturity.


In case you have any questions or remarks, please let me know.

Kind regards,
Rianne van Lieshout

Secretariaat Eidma Technische Universiteit Eindhoven
HG 9.50
Postbus 513
5600 MB Eindhoven
Tel  +31 40 247 2254
Fax +31 40 247 5366
E-mail: eidma at tue.nl<mailto:eidma at tue.nl>
------- End of forwarded message -------


More information about the crypto mailing list