Hyunho Kang

An image steganography using pixel characteristics

This paper presents a steganographic algorithm in digital images to embed a hidden message into cover images. This method is able to provide a high quality stego image in spite of the high capacity of the concealed information. In our method, the number of insertion bits into each pixel is different according to each pixel’s characteristics. That is, the number of insertion bit is dependent on whether the pixel is an edge area or smooth area. We experimented on various images to demonstrate the effectiveness of the proposed method.

Full-Index-Embedding Patchwork Algorithm for Audio Watermarking

For the digital watermarking patchwork algorithm originally given by Bender et al., this paper proposes two improvements applicable to audio watermarking. First, the watermark embedding strength is psychoacoustically adapted, using the Bark frequency scale. Second, whereas previous approaches leave the samples that do not correspond to the data untouched, in this paper, these are modified to reduce the probability of misdetection, a method called full index embedding. In simulations, the proposed combination of these two proposed methods has higher resistance to a variety of attacks than prior algorithms.

A novel steganographic system with information integrity

We propose a novel steganographic system that is able to provide a high quality stego image in spite of the high capacity of the concealed information. The number of insertion bits in a pixel is different according to each pixels characteristics. In addition to that, our system is able to verify whether if the secret information was sent from the sender correctly or if an attacker forged the secret message. We use two frequency coefficients in DCT domain to generate the symbol code to check the integrity of secret information.

Video fingerprinting system using wavelet and error correcting code

In this paper, we present a video fingerprinting system to identify the source of illegal copies. Content is distributed along a specified tree, with the seller as the root of the tree, the legitimate users as the leaves, and the internal nodes as content buyer or seller. Because there is a limited number of user areas available in each tree, we propose to build sub-trees, where each sub-tree has a distinctive logo. In this paper, we will use logos which are bit mapped images of the tree number. The extracted logo shows better performance visually using ECC. The fingerprinting step is achieved by the insertion of a unique information in the video wavelet coefficients by temporal wavelet transform. Our fingerprinting system is able to detect unique fingerprinting information in video content even if it has been distorted. In addition, our method does not need original video frame for extraction step.

Wolf fingerprints against minutiae count matching systems

Wolf attack is a new kind of biometric specific vulnerability and wolf attack probability is defined as a maximum success probability of the wolf attack with one wolf sample. From a theoretical point of view, a possible approach for describing wolf attack probability has been proposed in recent years, yet very little is known about practical approach in using a minutiae pattern. In this paper we propose a making method of wolf fingerprints to show vulnerability against simple count matching system. Even though this approach is valid only in a simple model, it is closely related to wolf attack against the real fingerprint systems. The experimental results show that wolf fingerprint using a minutiae pattern can get very high wolf attack probability.

Tracing illegal users of video: reconsideration of tree-specific and endbuyer-specific methods

In our recent study, we have presented an approach for tracing illegal users in content distribution networks using watermarking and fingerprinting techniques [1][2]. In this paper we generalize our previous work, further the collusion robustness is supplemented by additional security and practical experiment. This includes a more efficient tree decision method, generalization of the fingerprinting system and detailed investigation of the robustness against collusion attacks. Content is distributed along a specified tree, with the seller as the root, and the buyers as the internal nodes or leaves. The fingerprinting step is achieved by the insertion of unique information in the video wavelet coefficients by temporal wavelet transform. Our system is able to detect the fingerprint even if the video content has been distorted by collusion attacks.

Psychoacoustically-Adapted Patchwork Algorithm for Watermarking

For the digital watermarking patchwork algorithm originally given by Bender et al, this paper proposes two improvements applicable to audio watermarking. First, the watermark embedding strength is psychoacoustically adapted, using the Bark frequency scale. Second, whereas previous approaches leave the samples which do not correspond to the data untouched, in this paper, these are modified to reduce the probability of misdetection, a method called full index embedding. In simulations, the proposed combination of these two proposed methods has higher resistance to a variety of attacks that prior algorithms.

A viable system for tracing illegal users of video

Typical uses of watermarks include copyright protection and disabling unauthorized access to content. Especially, copyright protection watermarks embed some information in the data to identify the copyright holder or content provider, while receiver-identifying watermarking, commonly referred to as fingerprinting, embeds information to identify the receiver of that copy of the content. Thus, if an unauthorized copy of the content is recovered, extracting the fingerprint will show who the initial receiver was [1][2]. In this paper we generalize our previous work [3] of a video fingerprinting system to identify the source of illegal copies. This includes a logo embedding technique, generalization of the distribution system and detailed investigation of the robustness against collusion attacks.

Digital Watermarking Scheme Based on Machine Learning for the IHC Evaluation Criteria

Digital watermarking is a technique used for embedding information in digital content and protecting its copyright. The important issues to be considered are robustness, quality and capacity. Our goal is to satisfy these requirements according to the Information Hiding and its Criteria for evaluation (IHC) criteria. In this study, we evaluate our watermarking scheme along the IHC criteria Ver.3 as the primary step. Although image watermarking techniques based on machine learning already exist, their robustness against desynchronization attacks such as cropping, rotation, and scaling is still one of the most challenging issues. We propose a watermarking scheme based on machine learning which also has cropping tolerance. First, the luminance space of the image is decomposed by one level through wavelet transform. Then, a bit of the watermark and the marker for synchronization are embedded or extracted by adjusting or comparing the relation between the embedded coefficients value of the LL space and the output coefficients value of the trained machine learning model. This model can well memorize the relationship between its selected coefficients and the neighboring coefficients. The marker for synchronization is embedded in a latticed format in the LL space. Binarization processing is performed on the watermarked image to find the lattice-shaped marker and synchronize it against cropping. Our experimental results showed that there were no errors in 10HDTV-size areas after the second decompression.