Exposing inpainting image forgery under combination attacks with hybrid large feature mining
Methods and systems of detecting tampering in a digital image includes using hybrid large feature mining to identify one or more regions of an image in which tampering has occurred. Detecting tampering in a digital image with hybrid large feature mining may include spatial derivative large feature mining and transform-domain large feature mining. In some embodiments, known ensemble learning techniques are employed to address high feature dimensionality. detecting inpainting forgery includes mining features of a digital image under scrutiny based on a spatial derivative, mining one or more features of the digital image in a transform-domain; and detecting inpainting forgery in the digital image under scrutiny at least in part by the features mined based on the spatial derivative and at least in part by the features mined in the transform-domain.
1. A method of detecting tampering in a digital image, comprising:
mining one or more features of a digital image under scrutiny based on a spatial derivative, wherein mining the one or more features based on the spatial derivative comprises:
determining a spatial derivative associated with at least a portion of a digital image under scrutiny; and
extracting, based on the spatial derivative, one or more neighboring joint density features and marginal features from the digital image;
mining one or more features of the digital image using at least one DCT transform, wherein mining one or more features of the digital image by the at least one DCT transform comprises:
extracting one or more neighboring joint density features and marginal joint density features from the digital image; and
determining one or more calibration features in a DCT domain based on the neighboring joint density features and marginal joint density features; and
detecting tampering in the digital image under scrutiny at least in part by the features mined based on the spatial derivative and at least in part by the features mined using the at least one DCT transform;
wherein detecting tampering in the digital image under scrutiny comprises detecting inpainting under combined attacks.
2. The method of claim 1 , wherein detecting tampering in the digital image under scrutiny comprises detecting in painting in at least a portion of the digital image.
3. The method of claim 1 , wherein detecting tampering in the digital image under scrutiny comprises detecting seam carving in at least a portion of the digital image.
4. The method of claim 1 , wherein detecting tampering in the digital image under scrutiny comprises determining that the digital image is a forgery from at least one post-combination attack.
5. The method of claim 1 , wherein detecting tampering in the digital image under scrutiny comprises detecting seam carving under combined attacks.
6. The method of claim 1 , wherein the digital image comprises inpainting removing an object from the digital image.
7. The method of claim 1 , wherein the digital image comprises inpainting adding an object to the digital image.
8. The method of claim 1 , wherein mining one or more features of the digital image using at least one DCT transform comprises merging marginal joint density and neighboring joint density.
9. The method of claim 1 , wherein mining one or more features of a digital image under scrutiny based on a spatial derivative comprises merging marginal joint density and neighboring joint density.
10. The method of claim 1 , wherein the spatial derivative is based at least in part on image intensity changes.
11. The method of claim 1 , wherein detecting tampering in the digital image under scrutiny comprises detecting down-recompression of at least a portion of the digital image.
12. The method of claim 1 , wherein detecting tampering in the digital image under scrutiny comprises detecting inpainting forgery in the same quantization.
13. The method of claim 1 , wherein detecting tampering in the digital image comprises rich feature mining.
14. The method of claim 1 , wherein detecting tampering in the digital image comprises applying an ensemble classifier.
15. The method of claim 1 , wherein detecting tampering in the digital image comprises applying a Fisher linear discriminant.
16. The method of claim 1 , wherein the at least a portion of the features are determined according to different frequency coordinates in a DCT domain and under a shift recompression version.
17. The method of claim 1 , wherein the digital image is a JPEG image.
18. The method of claim 1 , wherein detecting tampering comprises distinguishing between single compression and down compression.
19. The method of claim 1 , further comprising:
generating a digital image comprising with distinguishing features in one or more regions that have been tampered with from an original image; and
displaying the image with the indications of at least one of the regions that have been tampered with to a user.
20. A system, comprising:
one or more computing devices, each of the one or more computing devices comprising a processor and a memory for storing program instructions wherein the program instructions, when executed on one or more computers, cause the one or more computers to implement a digital image tampering detection method comprising;
mining one or more features of a digital image under scrutiny based on a spatial derivative, wherein mining the one or more features based on the spatial derivative comprises:
determining a spatial derivative associated with at least a portion of a digital image under scrutiny; and
extracting, based on the spatial derivative, one or more neighboring joint density features and marginal features from the digital image;
mining one or more features of the digital image using at least one DCT transform, wherein mining one or more features of the digital image by the at least one DCT transform comprises:
extracting one or more neighboring joint density features and marginal joint density features from the digital image; and
determining one or more calibration features in a DCT domain based on the neighboring joint density features and marginal joint density features; and
detecting tampering in the digital image under scrutiny at least in part by the features mined based on the spatial derivative and at least in part by the features mined using the at least one DCT transform;
wherein detecting tampering in the digital image under scrutiny comprises detecting inpainting under combined attacks.
21. A non-transitory, computer-readable storage medium comprising program instructions stored thereon, wherein the program instructions, when executed on one or more computers, cause the one or more computers to implement a digital image tampering detection system configured to:
mining one or more features of a digital image under scrutiny based on a spatial derivative, wherein mining the one or more features based on the spatial derivative comprises:
determining a spatial derivative associated with at least a portion of a digital image under scrutiny; and
extracting, based on the spatial derivative, one or more neighboring joint density features and marginal features from the digital image;
mining one or more features of the digital image using at least one DCT transform, wherein mining one or more features of the digital image by the at least one DCT transform comprises:
extracting one or more neighboring joint density features and marginal joint density features from the digital image; and
determining one or more calibration features in a DCT domain based on the neighboring joint density features and marginal joint density features; and
detecting tampering in the digital image under scrutiny at least in part by the features mined based on the spatial derivative and at least in part by the features mined using the at least one DCT transform;
wherein detecting tampering in the digital image under scrutiny comprises detecting inpainting under combined attacks.