Identification of Multiple Copy-move Attacks in Digital Images using FFT and CNN

Copy-move forgery is an image manipulation technique wherein significant elements are added or removed from the image to spread misinformation. These forged images have flooded the internet owing to easily accessible image editing software. Copy-move attacks distort edges around the manipulated elements and thus can be detected by analyzing these edges. Traditional detection methods utilize algorithms like Cosine Transform, Scale-invariant Feature Transform, and Convolution Neural Networks (CNN), which have varying levels of accuracy and fail when there are multiple copy-move attacks in the image. Since edge detection and analysis are critical steps, performing Preliminary Edge Detection on the images proves effective. This study proposes Fast Fourier Transform (FFT) to perform preliminary edge detection on the images used to train the CNN model. The proposed algorithm that combines FFT and CNNs achieves enhanced testing accuracy of 95.91%, low validation loss of 0.252, and performs exceptionally well with multiple copy-move attacks.