Journal of Visual Communication and Image Representation
Block-based image steganalysis: Algorithm and performance evaluation
Introduction
The goal of image steganography is to embed secret messages in an image so that no one except the intended recipients can detect presence of secret messages. It has many applications such as embedding the copyright information into professional images, personal information into photographs in smart IDs (identity cards), and patient information into medical images [1]. Using image steganalysis, one attempts to detect the presence of secret messages hidden in such images.
With the advance of image steganography, many steganalysis methods have been developed to deal with new breakthroughs in image steganography. In the early stage, it is assumed that some prior information about steganographic algorithms that embeds a secret message into images is available. This is called targeted steganalysis. However, more attention has been paid to a more realistic situation in recent years. That is, no information about steganographic algorithms is available. This is known as blind steganalysis, which attempts to differentiate stego images from cover images without the knowledge of steganographic embedding algorithms [2]. Using features extracted from cover and stego images in a training set, we may design a classifier that separates cover and stego images in the feature space.
Most previous work on image steganalysis focused on extracting features from images and used a binary classifier to differentiate stego images from cover images. The research objective was to find a better feature set to improve the steganalysis performance. Fridrich [3] proposed the use of DCT features for steganalysis since inter-block dependency between neighboring blocks is often affected by steganographic algorithms. Shi et al. [4] proposed to use Markov features since the differences between absolute values of neighboring DCT coefficients can be modeled as a Markov process. This feature set is useful because intra-block correlations among DCT coefficients within the same block can be affected by steganographic embedding. Pevnỳ and Fridrich [5] proposed a set of 274 merged features by combining DCT and Markov features together.
So far, little attention has been paid to the characteristics of cover images to design content-adaptive classifiers in steganalysis. An input image typically consists of heterogeneous regions. We may decompose an image frame into smaller blocks and use each block as a basic unit for steganalysis. The effect of steganographic embedding on similar image blocks is known to have a stronger correlation [6]. As a result, the characteristics of smaller blocks can be used to design content-adaptive classifiers.
The frame-based steganalysis, which extracts a set of features from the whole image, was reported in almost all previous work [3], [4], [5]. In contrast, the block-based steganalysis, which extracts features from each individual block, was proposed by the authors in [7]. Based on the block features, a tree-structured vector quantization (TSVQ) scheme can be adopted to classify blocks into multiple classes. For each class, a specific classifier can be trained using block features, which represent the characteristics of the block class. For a given test image, instead of making a single decision for the entire image, we repeat the block decomposition process and choose a classifier to make a cover/stego decision for each block depending on block features. Finally, a decision fusion technique can be used to fuse steganalysis results of all blocks so that one can decide whether an unknown image is a cover or stego image.
The rest of this paper is organized as follows. Related previous work is reviewed in Section 2. The proposed block-based image steganalysis system is presented in Section 3. Analysis of the performance of block-based image steganalysis by considering the effects of block sizes, block numbers and the block overlapping design is conducted in Section 4. Fusion of multiple block decisions into one final decision for a test image is examined in Sec. 5. Extensive experimental results are shown for thorough performance evaluation in Section 6. Finally, concluding remarks and future research directions are provided in Sec. 7.
Section snippets
Review of previous work
Previous research in blind steganalysis has focused on extracting features from the whole image [3], [4], [5]. The number of features was increased to achieve better steganalysis performance in recent years. Chen et al. [8] proposed a set of updated Markov features (486 features in total) by considering both intra-block and inter-block correlations among DCT coefficients of JPEG images. Kodovskỳ et al. [9] examined a set of updated merged features (548 features in total) using the concept of
System overview
The block-diagram of a block-based image steganalysis system is shown in Fig. 1. It consists of the training process and the testing process, which will be detailed in the following two subsections, respectively.
- •
The training process. The system decomposes an image into smaller blocks and treats each block as a basic unit for steganalysis. A set of features is extracted from each individual image block and a tree-structured hierarchical clustering technique is used to classify blocks into
Analysis of block size, number and overlapping effects
There exists a relationship between the block size and the block number for a given image. If the block size is smaller, there are more blocks. We may ask “what is the best block decomposition strategy?” In the first two subsections, we examine the non-overlapping block case [15]. Then, in the last subsection, we consider the overlapping block case.
Fusion of block decisions
It is often beneficial to combine multiple local decisions to make a single global decision in decision making [16], [17]. The majority voting method was considered in the last two sections. There are more decision fusion methods such as weighted majority voting, Bayesian decision fusion, and the Dempster-Shafer theory of evidence. We will examine them in this section and see how they affect the final decision accuracy in the next section.
For the binary classifier case, there are only two
Performance evaluation
The performance of block-based image steganalysis for a binary classifier (either stego or cover image) will be studied in this section. We will compare the proposed block-based approach with the frame-based approach. We will provide experimental results by varying parameters in block-based image steganalysis so as to understand the effects of block sizes, block numbers, and block overlapping.
The performance of blind steganalysis is measured by the average detection accuracy:
Conclusion and future extension
A block-based image steganalysis system was proposed in this work, and extensive performance evaluation of block-based image steganalysis was conducted. It was shown by experimental results that the proposed method offers a significant improvement in detection accuracy when compared to prior art using an frame-based approach. Besides, block-based image steganalysis offers decision reliability information even with only one test image given, which is not available with the frame-based approach.
References (24)
- et al.
Digital image steganography: survey and analysis of current methods
EURASIP Journal on Signal Processing
(2010) Combining the results of several neural network classifiers
Neural Networks
(1994)Using measures of similarity and inclusion for multiple classifier fusion by decision templates
Fuzzy Sets and Systems
(2001)- et al.
Digital Watermarking and Steganography
(2007) - J. Fridrich, Feature-based steganalysis for JPEG images and its implications for future design of steganographic...
- Y. Shi, C. Chen, W. Chen, A Markov process based approach to effective attacking JPEG steganography, in: Proc. Int....
- T. Pevnỳ, J. Fridrich, Merging Markov and DCT features for multi-class JPEG steganalysis, in: Proc. SPIE Conf....
- B. Rodriguez , G. Peterson, K. Bauer, S. Agaian, Steganalysis embedding percentage determination with learning vector...
- S. Cho, B.-H. Cha, J. Wang, C.-C.J. Kuo, Block-based image steganalysis: algorithm and performance evaluation, in:...
- C. Chen, Y. Shi, JPEG image steganalysis utilizing both intrablock and interblock correlations, in: Proc. IEEE Int....
Cited by (27)
A review of image steganalysis techniques for digital forensics
2018, Journal of Information Security and ApplicationsCitation Excerpt :The reported median (MED) testing error as well as the median absolute deviation (MAD) values showed that the proposed method performs well. Cho et al. [89] discriminated a stego image from its cover image based on steganalysis of decomposed image blocks. They decomposed the image into blocks, classified those blocks into multiple classes and found a classifier for each class.
New framework for unsupervised universal steganalysis via SRISP-aided outlier detection
2016, Signal Processing: Image CommunicationCitation Excerpt :In this approach, the given image is partitioned or segmented into several sub-images before preclassification according to several statistical properties, such as texture complexity. The final result of the entire image is then determined through a weighted fusing process of all categories of the sub-images; this procedure has been demonstrated to mitigate the CSM issue significantly [23–26]. However, it also suffers from expensive training.
Evaluation on algorithm reliability and efficiency for an image flame detection technology
2023, Journal of Thermal Analysis and CalorimetrySteganography and steganalysis using machine learning
2021, Internet of Things: Energy, Industry, and Healthcare