Code implemented in Keras.

cnn.py and cnn_256.py are files implemented for camera source identification.

CNN model using 3 convolution layer.

Image patches from dataset are collected in .csv file.

In cnn.py, preprocessing, i.e., highpass filtering has been performed before feeding data to CNN module.

Input : Image patches from 10 camera models from Dresden Image Dataset.

Output: camera source of test image.

Model gives good classification accuracy, i.e., ~96 % for 10 camera models.