Abstract
With the growing popularity of social networking services and the rapid development of smart devices, an increasing number of people are now uploading media such as images and videos to social networks for sharing with acquaintances, and making that content available for public use. Given the flexibility of uploading and sharing media, a common question arises: can we trust all these images and videos? To address this issue, we propose a mobile cloud-based media uploading framework that checks images for authenticity (i.e., by detecting known forgery techniques). For images that are found to be genuine, the framework will allow public sharing. The authenticity check will be carried out on the private or local cloud to which the image is uploaded. For the check itself, we propose a curvelet transform and Weber local descriptor-based system to extract features from the image. Statistical features obtained from the system are then fed into a support vector machine-based classifier. Offline experiments show that the proposed system can achieve very high detection accuracy for image forgery. Online experiments show that the proposed framework also works in real time. To validate the suitability of this media uploading framework, workloads were measured in the Amazon Elastic Compute Cloud environment.