A probabilistic image authentication method based on chaos

Abstract

Fragile watermarking methods have been used as active image authentication methods. These methods must have capabilities like high capacity, high visual quality, high image authentication ability and image recovery capability. Many of the fragile watermarking methods suggested in the literature do not provide all of these capabilities. In this article, a new probabilistic image authentication method with high capacity, high visual quality, high image authenticity ability and image recovery capability is proposed. Chaotic pseudo random number generators, a new watermark generation network which similar to neural network, modulo based watermark embedding functions, modulo based watermark extraction functions, tamper detection algorithm and perceptual hash based image recovery algorithm are utilized for creating a new probabilistic image authentication method in this article. The proposed method is a block based authentication method and 3 x 3 size of overlapping neighborhood blocks are used in this method. The proposed watermark generation network uses symmetric pixels as inputs and weights to generate predicted pixel. In this step, modulo operator and random numbers are used for calculating embeddable value. Owing to chaotic random number generator, the proposed method gains probabilistic and nonlinear structure. To generate random number, logistic map is used as random number generator in the proposed image watermarking method. Then, a modulo based watermark embedding function is used for embedding embeddable value (generated watermark). In the watermark extraction step, modulo operator is used. If the predicted value differs from the extracted value, tamper detection is performed. Otherwise, image authentication is success. A new image recovery algorithm is proposed in this method. This algorithm provides image recovery for tampered areas with high visual quality. The proposed image watermarking method can authenticate grayscale and color images. The experimental results demonstrated that, the proposed method has high payload capacity, high visual quality and high image authentication ability.