Enhancing Deepfake Image Detection with Deep Convolutional Neural Networks

Siddharth Bhamare, S. S. Bhamare · 2023

The availability of image editing software such as Adobe Photoshop or GIMP has made picture alteration so widespread these days. Finding these phony photos is a must for exposing image-based cyber crimes. Due to its ubiquity, images produced with a digital camera or smartphone are typically saved in the JPEG format. The 8×8 pixel-sized, independently compressed mage grids used by the JPEG technique are used. Images that haven't been altered have a comparable inaccuracy level. Due to a similar number of faults throughout the whole image, each block should degrade at roughly the same rate during resaving operations. Error Level Analysis was used to determine that the compression ratio of the false image was different from the actual images.

Read the paper · More papers on PaperTik