Many images are spread in the virtual world of social media. With the many editing software that allows so there is no doubt that many forgery images. By forensic the image using Error Level Analysis to find out the compression ratio between the original image and the fake image, because the original image compression and fake images are different. In addition to knowing whether the image is genuine or fake can analyze the metadata of the image, but the possibility of metadata can be changed.
image forensic is a field of the study identifying the origin and verifying the authenticity of the image.
Many methods are used to determine the level of authenticity of the picture, one with
determining the quality of the image compression level results. He, the methods used to measure the level of compression is using Error Level Analysis (ELA).
Error level analysis is one technique for knowing images that have been manipulated by
storing images at a certain quality level and then calculating the difference from the
compression level.
In this we have applied Deep Learning to recognize images of manipulations through the dataset of a fake image and original images via Error Level Analysis on each image and supporting parameters for error rate analysis.
The result of my experiment is given below:
| Architecture | Accuracy |
|---|---|
| CNN | 88.8 |
| Pretrained Resnet | 96.2 |
Classification Report
ELA Images batch
Decrease in Loss on a subset of dataset
Increase in F1 Score over on a subset of dataset
Training
python predict.py --image_path sample\Au_txt_30028.jpg --model_path resnet.pth
python predict.py --image_path sample\Au_txt_30028.jpg --model_path cnn.pth