CNN-based Approach for Robust Detection of Copy-Move Forgery in Images

The evolution of image manipulation techniques has presented a paradoxical scenario in contemporary visual culture. This phenomenon operates as a double-edged sword, offering both creative liberation and ethical dilemmas. Consequently, there is a need to develop automated mechanisms capable of discerning such forged data. The proposed methodology leverages transfer learning, utilising pre-trained deep learning models as a foundation and fine-tuning them specifically for the task of copy-move forgery detection. This approach uses the knowledge learned from large datasets, enhancing the network's ability to discern subtle patterns indicative of copy-move manipulations in images. Further, this research introduces a custom-designed CNN architecture tailored to the intricacies of copy-move forgery, optimising feature extraction and classification. Experimental evaluations conducted on diverse datasets, namely MICC-F220, MICC-F600, MICC-F2000, and CoMoFoD demonstrate the effectiveness of the proposed method with a True Positive Rate (TPR) of 100%.