State of Art-Existing Techniques

STATE OF ART- EXISTING TECHNIQUES FOR IDENTIFICATION AND PREVENTION OF FORGERY IN IMAGES

In our project we are concentrating on the anti forgery techniques in image forgery. We have two main results one is image processing whose output will be given as input to the other ML,DL and CNN part of the project.
Digital image forgery can be categorised into
1. Active approach
2. Passive approach
based on the process involved to create the fake images. New techniques for countering image forgery attacks

Active Approach: Refers to the kind of pre processing of watermark embedded, digital signatures, message authentication code and image hashing that act as active protection techniques as information such as location and time the image was captured is embedded with the image when it is obtained. If such vital information cannot be extracted from the obtained image it may be tampered and this method is used to evaluate the integrity of the digital image. Inverse analysis is done to locate the tampered region of the image.

Passive Approach: Can be challenging as there is not a particular method that can treat all the cases. Scrutinizing the image based on image feature statistics and security construct used, also called raw image analysis.

Popular Image Forgery Techniques:

Image Retouching:
A method which is not harmful but enhances or reduces certain features of the original image. Used by magazine photo editors.

Copy-Move Forgery Attack:

Difficult but most commonly used forgery technique to cover a part or hide a part of the image by copying from a part of the the image and pasting it in another part of the same image.
Many copy-move methods:
1.Plain copy-move
2.Multiple copy-move
3.Copy-move with reflection attack
4.Copy-move with image inpainting

Image Splicing:
Process of merging two or more images, also called photomontage or image composting. Region is copied from source image and the spliced portion is pasted in the target image creating a fake image.

Resampling and Resizing:
Process of transforming an image into another coordinate system. Resizing refers to the change in the document size of the image without changing the number of pixels. Resampling refers to the change in the number of pixels representing the image and can be done in many ways
1.Up-sampling
2.Down-sampling
3.Mirroring
4.Skewing

Anti-Forensic Forgeries:
Such forgeries are targeted at escaping forgery detection methods. Mostly applied by forensic teams to find loopholes in their detection methods and can develop counter anti-forensic methods. They also have the capability of effectively eliminating the imprints left by a particular image forgery.

Image Forgery Detection and Prevention Techniques:

SVM Classifier:

The accuracy of detecting forgery is enhanced by using SVM classifier.It consists of 2 phases training and testing phase Actual working of SVM Classifier determines the decision boundaries in the training step and their methods can provide good generalization in high dimensional input spaces

Copy-move Forgery Detection using Pixel based approach:

This Algorithm is based on Pixel Based approach. The dyadic wavelet transform (DWT) is applied to the input image and the transform yield of original Image is a reduced dimension representation. The Copy-Move regions can be located by pixel matching, that shifts the input image according to the offset and calculates the difference between its shifted version and the original image.

Copy-move Forgery Detection using Partition based approach:
1. Block-Based Approach: Color conversion is done on the taken images. The source and the target regions are both located in the same image hence, the forged image must exhibit at least two similar regions. helps to avoid the high computational cost of the exhaustive search by dividing into blocks and the comparison is done at a block level. After extracting the features using transformations like DCT, DWT, copy-move pairs are identified by matching those features, which can be done easily by searching the blocks with similar feature vectors.
2.Nonblock-Based Approach:
Segment-Based Approach:
Partitioning image into relatively homogeneous segments and a single segment fully contains an object. Corresponding noise segment is extracted from the input image. Finally, histograms of the extracted noise segments are compared for similarity check, which leads to the detection of the forgery image.
Sub image-Based Approach:
Partition the image into sub-images of same size and divide into four non-overlapping sub-images. To evaluate the spatial offset between the copied region and the pasted one, the phase correlation between every pair of sub-images is calculated. The location of the forgery is got by shifting the input image in-line with the obtained offset and subtracting this shifted image from the original input image.

Fusion-based method: fusion methods are used to get good localization for detection results of different approaches. Two forensic approaches are used for forgery localization, one is a statistical feature-based approach, the other is a copy-move forgery detection technique. Another approach, a PatchMatch algorithm is used to find out the similar patches in the different areas of images and obtain offset fields for such similar patches. Six different fusion methods are implemented to fuse the two techniques. In some cases, it cannot be distinguished between a source region and a tampered region, so copy move detection fails.