Abstract:

We present a novel design for large-scale multimedia content protection systems that harness the power of cloud infrastructure to ensure cost efficiency, rapid deployment, scalability, and adaptability to varying workloads. The proposed system is capable of safeguarding diverse multimedia content types, including 2D and 3D videos, images, audio clips, and music. It can be deployed on both private and public cloud environments. Our approach introduces two key innovations: (i) a signature generation method for 3D videos and (ii) a distributed matching engine for multimedia content. The signature method produces robust and representative signatures that effectively capture depth signals in 3D videos while maintaining computational efficiency and minimal storage requirements. The distributed matching engine is designed for high scalability and supports various multimedia formats. We implemented the proposed system and deployed it on two cloud platforms—Amazon Cloud and a private cloud. Experimental evaluations conducted on a dataset of over 11,000 3D videos and 1 million images demonstrate the system’s high accuracy and scalability. Furthermore, a comparative analysis with YouTube’s protection system reveals that while YouTube’s system fails to detect most copies of 3D videos, our system achieves a detection rate exceeding 98%. This highlights the necessity of our 3D signature method, as existing commercial solutions struggle to handle 3D video content effectively.

Description:

.