Week 4 — Copy Move Forgery Detection

Hi again, in this week we studied the papers about copy move forgery detection to be aware of the common solutions, pitfalls and how to make a successful CMFD (Copy Move Forgery Detection) solution. This helped us contextualize a better look for our solution. Based on these information, we selected how we are going to continue our project and what we are trying to achieve.

An In-Depth Look at Copy Move Forgery Detection

As the technological progress goes on, the forgeries and forgery detection methods both increase like an arms-race. There are multiple techniques to confuse forgery detection methods. These techniques include, rotation of the copied image (as we displayed in our 2nd week report), scaling the image, adding noise to image, compression of image, etc. A proper CMFD solution should handle all of them as correctly as possible.

All current solutions to Copy Move Forgery Detection have a similar structure like Figure 1 with some steps added or removed.

Figure 1: Copy Move Forger Detection General Pipeline

While there are multiple methods to do each step, the most caveats occur at keypoint detection step as this is the step that differentiates between a good CMFD solution and a bad CMFD solution. The most important aspects of a good keypoint detection method is uniqueness and robustness. If the keypoints are not unique enough there would be many false positive results and a keypoint detection should be able to robust enough to detect changes and transformations like scaling, rotation or added noise etc.

The accuracy of a Copy Move Forgery Detection can not be measured like an identified or unidentified situation in an unlabeled test data set. But as we are using an established labeled data set (which can be accessed here), our accuracy criteria will be false positives to true positives ratio of detection as with the paper attached to this study. Since we are using their data set we decided to use their study and add our contribution to it by improving it’s accuracy.

Project Participants

Cemre Şenokur - Medium

Eflatun Doğa AYDIN - Medium

Mustafa Kucuk - Medium

References

References

https://www.researchgate.net/publication/224225329_A_SIFT-Based_Forensic_Method_for_Copy-Move_Attack_Detection_and_Transformation_Recovery

https://www.researchgate.net/publication/317495890_Review_Analysis_and_Parameterization_of_Techniques_for_Copy-Move_Forgery_Detection_in_Digital_Images

https://www.researchgate.net/publication/270157882_STATE_OF_THE_ART_OF_COPY-MOVE_FORGERY_DETECTION_TECHNIQUES_A_REVIEW

https://ieeexplore.ieee.org/document/8673945