Identification of Illegal Users from Pirated Copies

March.19.2018RESEARCH HIGHLIGHTS

In order to identify illegal users from pirated copies, each user’s identity information called fingerprint should be inserted into multimedia content before selling. The most challenging issue is the robustness against collusion attack such that a coalition of users remove/modify the fingerprints by comparing their copies. A fingerprinting code is one of the promising approaches for the collusion-secure tracing system.

At the identification of illegal users, each user’s codeword is subjected to check a level of suspicion by calculating the similarity score with the codeword extracted from a pirated copy. Although an optimal scoring function has been presented in a conventional study, the realization is difficult because the number of illegal users and their attack strategy are inevitable.

Now, Minoru Kuribayashi at Okayama University have developed a new scoring function and proposed a semi-optimal method which can well classify the illegal users and innocents with simple operation.

The proposed method estimates the attack strategy in order to select its corresponding weights for calculating similarity scores. The advantage is its simplicity required for the estimation because it merely observes the bias of symbols “0” and “1” and then roughly classifies the strategy into three classes.

The performance is evaluated in the presence of additive white Gaussian noise, and it is compared with some state-of-the-art methods. As the result, it is confirmed that the best performance is obtained by the proposed method and it is very close to the optimal one.

At the distribution of multimedia content, a server inserts unique ID information called fingerprint into the content. Once a copy is found, the illegal users can be identified if the fingerprint is correctly extracted.

If a coalition of users compares their codewords, they can find the positions where some of their symbols are same and they cannot modify the symbols at such positions. Otherwise, they can select the symbols with an arbitrary strategy such as majority voting, minority voting, and so on.

The proposed method first estimates the attack strategy into three classes, and then customizes the scoring function according to the estimated strategy. Both the estimator and scoring function exploit the bias of symbols in a pirated codeword y which can be observed directly from the codeword.

Reference:
Authors
Minoru Kuribayashi and Nobuo Funabiki.

Title of original paper
Universal Scoring Function Based on Bias Equalizer for Bias-Based Fingerprinting Codes.

Journal, volume, pages and year
IEICE Trans. E101-D, No.1, 119-128, (2018).

Digital Object Identifier (DOI)
10.1587/transfun.E101.A.119

Journal website
http://search.ieice.org/bin/summary.php?id=e101-a_1_119

Affiliations
Graduate School of Natural Science and Technology, Okayama University.

Department website
http://www.ec.okayama-u.ac.jp/~dist/kuribayashi/index_e.php


Fig.1: Framework of the fingerprinting system.


Fig.2: Collusion attack


Fig.3: Flowchart of the proposed method

ACADEMIC YEAR