Fingerprinting with Minimum Distance Decoding

TL;DR

This paper develops a new fingerprinting scheme using minimum distance decoding, demonstrating significant performance improvements over existing methods through theoretical analysis and practical code constructions.

Contribution

It introduces a novel information theoretic framework for collusion-resistant fingerprinting, characterizes achievable rates under various attacks, and designs effective coding schemes based on belief propagation.

Findings

Achieves rate 1/2 for t=2 under averaging attack.

Achieves rate approximately 0.188 for t=2 under marking assumption.

Constructs practical codes with very low misidentification probability at rates 1/3 and 1/9.

Abstract

This work adopts an information theoretic framework for the design of collusion-resistant coding/decoding schemes for digital fingerprinting. More specifically, the minimum distance decision rule is used to identify 1 out of t pirates. Achievable rates, under this detection rule, are characterized in two distinct scenarios. First, we consider the averaging attack where a random coding argument is used to show that the rate 1/2 is achievable with t=2 pirates. Our study is then extended to the general case of arbitrary $t$ highlighting the underlying complexity-performance tradeoff. Overall, these results establish the significant performance gains offered by minimum distance decoding as compared to other approaches based on orthogonal codes and correlation detectors. In the second scenario, we characterize the achievable rates, with minimum distance decoding, under any collusion attack that satisfies the marking assumption. For t=2 pirates, we show that the rate $1-H(0.25)\approx 0.188$ is achievable using an ensemble of random linear codes. For $t\geq 3$, the existence of a non-resolvable collusion attack, with minimum distance decoding, for any non-zero rate is established. Inspired by our theoretical analysis, we then construct coding/decoding schemes for fingerprinting based on the celebrated Belief-Propagation framework. Using an explicit repeat-accumulate code, we obtain a vanishingly small probability of misidentification at rate 1/3 under averaging attack with t=2. For collusion attacks which satisfy the marking assumption, we use a more sophisticated accumulate repeat accumulate code to obtain a vanishingly small misidentification probability at rate 1/9 with t=2. These results represent a marked improvement over the best available designs in the literature.