Improved Active Method for Image Forgery Detection and Localization on Mobile Devices

摘要

Nowadays, the most popular device for capturing images is the camera integrated in smartphones. Recent technological advances (glass lenses, dual cameras) made possible taking very good quality photos without the need of a professional bulky camera. The number of photos shared on social media is continuously increasing. The simplicity of use achieved by the newest image editing software makes guaranteeing the image authenticity a key element in multimedia security. The paper presents an improved image authentication application based on digital watermarking which can also detect and highlight the tampered areas in an image. This version has greater flexibility since it allows the user to select the parameters involved in the watermark embedding process, but also increased security since the users can now use a chosen password when securing the photo. Optimization techniques helped shortening the processing time for securing and testing images for authenticity. Tests were done on various images and mobile devices, proving that the proposed method provides very good detection rates and high levels of security, while keeping the PSNR (Peak Signal-to-Noise Ratio) values of secured images high (over 42 dB) and the computation time acceptable.