Qim-Integrity-App

A library and an Android application for content integrity watermarking.

This repository represents my final-year major project in Telecom SudParis 'High Tech Imaging'.

The project is pubished under The MIT License (MIT)

Copyright (c) 2014 Cédric Golmard, Institut Mines – Telecom ; Telecom SudParis as part of HighTech Imaging major

The subject was proposed by Mihai Mitrea and Marwen Hasnaoui, researchers at Institut Mines – Telecom ; ARTEMIS.

Folder structure

QimWatermarking: Watermarking library (C/C++) and command-line tool implementing the watermarking (insertion/extraction) algorithms
TestsOnImageFolder: Applescript for automated tests on images: watermaking transparency and robustness
QimNavigationApp: Android application project for image watermarking using the native library
Images: Screenshots

The main objective is to implement a technique asserting that the content of an image has not been modified (Integrity), with constraints:

Watermarking: insert a message, detect and decide if the content has been modified.
Transparency: the insertion must be transparent (according to human eyes)
Robustness: the message must be recoverable if the image has been requantized
Weakness: the decision must detect modification on the content (objects, shapes, ...) of the image

The watermarking is based on QIM and ST-DM techniques to insert signature in the image based on its content.

The Watermarking library (C/C++) uses and requires OpenCV library (under BSD License) for image processing. It provides functions for:

The signature insertion generate a "key file", which is needed for the detection.

The Android application uses the QimWatermaking native library (through JNI) to enable marking and integrity checking of images.

Technical challenges (as part of Télécom SudParis HighTech Imaging major):

use a native library in Android (Write once...)
demonstrate and evaluate performances of the library on lightweight devices
give access to the library features via an easy and confortable user experience

Features:

mark an image loaded from the Image Library
mark an image taken with the camera
share the marked image and the key file
check the integrity given an image loaded from the Image Library and a key file loaded from the device's storage
- it gives an "integrity score" between 0.0 and 1.0
- a score over 0.9 assert that the content has not been modifies
- a score under 0.7 asser that the integrity has been broken
- intermediate scores depend on image and attacks on content / format.

Screenshots: