PRNU extraction via denoiser based on convolutional neural network (DRUNet)

Image Processing and Security

Authors

• Mirco Ceccarelli
• Francesco Argentieri

Relators

• Prof. Alessandro Piva
• Dott. Dasara Shullani
• Dott. Daniele Baracchi

Introduction

In this project we have dealt with a branch of the Multimedia Forensics: Device Identification (know what device has taken a photo).

In order to do this, a noise that is introduced by the sensors of each digital camera, is exploited: the PRNU (Photo Response Non Uniformity), which is different for each individual device.

Here below a brief scheme of the Device Identification process via PRNU:

Settings

Before running the code, set up the project correctly. Click here: model_zoo

In the prnu/functions.py file is used the Noise Extract method of DRUNet, if you want to use the Polimi Noise Extract method you need to comment some lines of code.

To Run

python3 example.py