Abstract:Given a JPEG pipeline (compression or decompression), this paper shows how to find the antecedent of a 8 x 8 block. If it exists, the block is compatible with the pipeline. For unaltered images, all blocks are always compatible with the original pipeline; however, for manipulated images, this is not always the case. This article demonstrates the potential of compatibility concepts for JPEG image forensics. It presents a solution to the main challenge of finding a block antecedent in a high-dimensional space. This solution relies on a local search algorithm with restrictions on the search space. We show that inpainting, copy-move, or splicing applied after a JPEG compression can be turned into three different mismatch problems and be detected. In particular, when the image is re-compressed after the modification, we can detect the manipulation if the quality factor of the second compression is higher than the first one. Our method can pinpoint forgeries down to the JPEG block with great detection power and without False Positive. We compare our method with two state-of-the-art models on localizing inpainted forgeries after a simple or a double compression. We show that under our working assumptions, it outperforms those models for most experiments.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: how to detect whether a JPEG image has been tampered with and be able to accurately locate the tampered 8×8 blocks. Specifically, the paper proposes a method based on JPEG compatibility for detecting and locating the forged parts in a modified JPEG image. Through this method, it can be identified which 8×8 blocks in the image are not compatible with the original compression pipeline, thereby determining whether these blocks have been modified. ### Main Problem Description 1. **JPEG Image Tampering Detection**: - The paper aims to develop a reliable and interpretable tool for detecting whether a JPEG image has been tampered with. - For an unmodified image, all 8×8 blocks should be compatible with the original compression pipeline; while for a tampered image, some blocks may no longer be compatible. 2. **Predecessor Search in High - Dimensional Space**: - The paper solves the main challenge of finding the predecessor (i.e., the original block) of an 8×8 block in high - dimensional space. - By using a local search algorithm and imposing restrictions on the search space, the paper provides a solution. 3. **Detection of Different Types of Tampering**: - The paper shows how to transform image editing operations (such as patching, copy - move or splicing) into three different mismatch problems and detect these tamperings. - In particular, after the image is recompressed, if the quality factor of the second compression is higher than that of the first compression, tampering can be detected. 4. **Accuracy and No False Positives**: - This method can accurately locate the forged content at the JPEG block level and will not generate false positives (False Positive). ### Method Overview - **Compatibility Concept**: The paper introduces the concept of "compatibility", that is, given an 8×8 block B, whether there exists a block A such that A can be obtained by processing through the JPEG pipeline to get B. If such a block A exists, then block B is compatible with the pipeline; otherwise, it is incompatible. - **Local Search Algorithm**: In order to find the predecessor block, the paper proposes a local search algorithm, which imposes theoretical constraints on the search space to reduce the search range and improve efficiency. - **Theoretical Constraints**: By introducing the theoretical constraints proposed by Fridrich et al., the search space is further narrowed, especially when the quality factor is less than 95. ### Conclusion The method proposed in the paper can not only detect the forged parts in JPEG images but also accurately locate specific 8×8 blocks, with high detection ability and interpretability. The experimental results show that this method is superior to the existing state - of - the - art models in most cases. ### Related Formulas 1. **DCT Transformation**: \[ \tilde{c}=DCT(x)\times Q \] \[ c = [\tilde{c}] \] 2. **IDCT Transformation**: \[ \tilde{y}=IDCT(Q\times c) \] \[ y = [\tilde{y}][0;255] \] 3. **Upper Bound of Quantization Error**: \[ \|y - \tilde{y}\|_2\leq4 \] 4. **Upper Bound of Compatibility Distance**: \[ \|x - x_b\|_2\leq M_f=\left\|\frac{Q}{2}\right\|_2 \] Through these formulas and methods, the paper successfully solves the key problems in JPEG image tampering detection and provides effective solutions.

Dual JPEG Compatibility: a Reliable and Explainable Tool for Image Forensics

Image Tampering Localization Via Estimating the Non-Aligned Double JPEG Compression

A Robust JPEG Image Tampering Detection Method Using GLCM Features

Multiple Heterogeneous JPEG Image Hierarchical Forensic

JPEG compression history detection based on detail deviation.

Double-Compression Detection in JPEG Steganalysis

Source Camera Identification Using Support Vector Machines

Forgery Detection Using Polynomial Fitting in Recompressed JPEG Images

Finding Incompatible Blocks for Reliable JPEG Steganalysis

JPEG Compression Forensics Against Resizing.

Revealing the Traces of Nonaligned Double JPEG Compression in Digital Images

Detection of Composite Images Based on JPEG Compression Properties

Block-level Double JPEG Compression Detection for Image Forgery Localization

Detecting Shifted Double JPEG Compression Tampering Utilizing Both Intra-Block and Inter-Block Correlations

Compatibility and Timing Attacks for JPEG Steganalysis

An Effective Method for Detecting Double JPEG Compression with the Same Quantization Matrix.

JPEG Compression-aware Image Forgery Localization

An image forensic technique based on JPEG ghosts

Robustness of copy-move forgery detection under high JPEG compression artifacts

Deep Multi-scale Discriminative Networks for Double JPEG Compression Forensics