Robust Median Filtering Forensics Using Image Deblocking and Filtered Residual Fusion

Abstract

Median filtering (MF) is frequently applied to conceal the traces of forgery and therefore can provide indirect forensic evidence of tampering when investigating composite images. The existing MF forensic methods, however, ignore how JPEG compression affects median filtered images, resulting in heavy performance degradation when detecting filtered images stored in the JPEG format. In this paper, we propose a new robust MF forensic method based on a modified convolutional neural network (CNN). First, relying on the analysis of the influence on median filtered images caused by JPEG compression, we effectively suppress the interference using image deblocking. Second, the fingerprints left by MF are highlighted via filtered residual fusion. These two functions are fulfilled with a deblocking layer and a fused filtered residual (FFR) layer. Finally, the output of the FFR layer becomes input when extracting multiple features for further classification using a tailor-made CNN. The extensive experimental results show that the proposed method outperforms the state-of-the-art methods in both JPEG compressed and small-sized MF image detection. INDEX TERMS Median filtering, convolutional neural networks, robust forensics, JPEG compression, image deblocking.

Technology of Actual Wind Deviation Monitoring for Suspension Insulator Strings Based on Improved Edge Detection

“…In the process of digital image acquisition and transmission, due to the participation of hardware devices and the influence of external factors, mixed noise will be caused, which has a great influence on the image processing process. Therefore, median filtering [21], [22] is used for denoising. The result of the median filtering process is shown in Figure 7.…”

Novel framework for optimized digital forensic for mitigating complex image attacks

“…Hao et al [19] have presented a forensic analysis of the palmprint using a supervised learning algorithm using a dual feature in order to assess the quality of the image. Shan et al [20] have used a deblocking filtering mechanism for the identification of the traces caused by the usage of median filtering. Singh and Singh [21] have carried out the experiment with respect to the statistically-based approach of the second order for resisting JPEG-based attacks over an image using supervised classifiers.…”

Improvement Feature Vector: Autoregressive Model of Median Filter Residual

“…For MFD, the feature vector types are classified into three main types. One has a single property [3], and the other one has a combination of various properties [2], [4] of the characteristics of an image, and the last one has an output value of the fully connected layer in CNN (Convolutional Neural network) on deep learning [5], [6].…”

“…For high-dimensional feature vector cases, Y. Zhang et al [7] reduced the length of the feature vector using kernel principal component analysis (KPCA). However, the length of the feature vector [5], [6] from the fully connected layer is too long against [2]- [4], and some large epochs and the iteration times are needful to get a satisfactory value of the feature vector.…”

“…In the method of the MFD, there are several main ways: the first is a block-by-block detection of median filtering [1], [3], [5], [10], the second is measuring the streaking artifacts [11], [12] and the percentage streak area (PSA) [13], the third is devised as the occurrence probability of zero values on the one-order difference map of textured regions [14], and lastly a difference of the unaltered and the altered image is used as the filter residual in the space domain [3], [5], [6] and the frequency residual in the spectral domain [1], [15], [16].…”