Methodologies for access control and fingerprinting of multimedia.

摘要

This thesis deals with design strategies for access control and traitor tracing. Inspired by the chameleon cipher, a joint fingerprinting and decryption (JFD) architecture is proposed as a computational and bandwidth efficient solution for securing multicast content and tracing pirates. In JFD, slightly different decryption keys are derived from a single encryption key and this structural difference is responsible for casting a fingerprint at the receiver. Based on this methodology a DCT sign bit encryption and fingerprinting algorithm is proposed for protecting and tracking JPEG compressed images in a multicast environment.; One powerful attack within the JFD framework is collusion, in which a group of pirates collude their decryption keys or their fingerprinted copies to remove traces of the fingerprint. Such attacks can be deterred by integrating anti-collusion codes (ACCs) into the decryption key design. The size and effectiveness of an ACC codebook depends on the watermark modulation strategy used and the collusion attack model. In this thesis, we construct an ACC tailored specifically to resist linear collusion attacks, but can also be used against other forms of non-linear collusion such as median, min and max functions. The proposed codebook requires O(n) basis vectors for tracking any K n colluders in a group of n users. All the codewords within this codebook are maximally equidistant with constant weight and can be built using (v, v+1 2, v+1 4 )-BIBDs derived from Hadamard matrices.; Finally we explore the problem of traitor tracing within a n -out-of-n joint access framework, where every subset of n users is treated as a traitor set and must be tracked. As a first step in the investigation we propose a non-perfect secret sharing algorithm called MIX-SPLIT for both access control and tracking. In MIX-SPLIT, every share of the secret has a unique signature which can be used to identify the owner. There is also a unique association between any two shares, which facilitates tracking. One application for MIX-SPLIT is joint access and tracking of highly sensitive data such as medical images. This algorithm can also be engineered to create mixed shares of two different information pieces such as different biometric PINS in applications which require secure PIN storage.