The Forwardt project aims to improve the resilience of audio-video watermarks against so-called collusion attacks, where multiple attackers hold differently watermarked versions of the same content. We focus in particular on the challenges in real-world content delivery systems, e.g. multi-channel attacks and dynamic tracing.
The cat-and-mouse game of attack and defense gives rise to some highly nontrivial mathematics, involving combinatorics, coding theory, information theory and cryptography. The practical side is equally complex, with much work in progress on attack strategies, signal processing attacks, optimal code constructions, efficient decoding algorithms, and embedding methods.
The task of the PhD student will be to:
design dynamic tracing algorithms to counter multi-channel attacks
attack and analyze these algorithms
try out untested approaches such as machine learning and countermeasures at the embedding layer
Opportunity Focus Areas:
Log in or Sign up to see How to apply details and steps.