This project proposes a real-time DRM and forensic mark technology based on the HEVC/MPEG-H Part.2(H.265) standards. This technology ,on a GPU multi-core basis, utilizes the Entropy Coding technology only to approach the QTC (Quantified Transform Coordinators) area and inserts the mark by applying the XOR operation to the n-bits of the forensic mark bitstream and the key based pseudo random n-bits. Compared to the previous researches, the PSNR was 10dB higher, and the insertion speed increased by 70 times at 8K and 13 times at 4K. And robustness shows an average of 84.37% in 8K video.
url: https://www.riss.kr/link?id=T15917232
out.mp4
It is a VOD (Video On Demand) streaming service with copyright protection technology applied.
First, creators must register as a member to upload their content.
Upon registration, the administrator gives the creators an account with access to the web and an account with access to the forensic mark management database.
Creators can create their own channels by logging in and upload content to the channel.
DRM and forensic marks are inserted in the uploaded content in real time when you play the content for the first time.
In the forensic marking process, the bitstream generated by the forensic mark generator is first stored in the forensic mark management DB and the original QTC key stream is stored in the QTC storage. After that, the insertion process proceeds by reading the mark information from the forensic mark management DB. DRM packaging is processed simultaneously with forensic mark insertion.
Users will watch content embedded with these two technologies.
Content purchase process and verification process.
There are two methods for this process.
The first is to verify the website itself, and the second is to verify it using a verification program without accessing the website.
The purchase process is currently with the billing library function turned off, so members can download it right away. (You can use the billing policy library if necessary) Downloaded content is converted to HEVC for copyright verification later and is downloaded. The verification process includes DRM unpackaging and forensic mark extraction.
It configures a heterogeneous system architecture that uses both CPU and GPU for real-time forensic mark insertion.
HEVC compression technology consists of I-frames, B-frames, and P-frames. In the case of I-frame, multiple CTUs can simultaneously perform CABAC encoding/decoding with WPP and Tile parallelization technology. However, B-frame and P-frame have unidirectional and bidirectional correlation, so parallelization cannot be performed in CABAC codec. In this method, CTUs in the I frame are processed by the CUDA core of the GPU by applying tile parallelization, and while the GPU is processing, the CPU receives the entire image frame and calculates the reference direction of the B frame and P frame. Pass direction information to the GPU's memory. Afterwards, in the section where I-frames are not being processed in the x-axis direction among the GPU's CUDA cores, B-frames and P-frames are processed equally in the tile method.
The DRM packaging server applies PRN-Based SFPT (Pseudo Random Number-Based Selective Frame Packaging Technique) for real-time packaging and parallelizes it based on thread pool.
This is different from applying DRM packaging to all frames in the video, when packaging is performed by selecting a frame to be subjected to DRM packaging, when normal frames to which packaging is not applied are leaked, There is an advantage that the original high-definition immersive content cannot be viewed because it is not possible to watch the low-quality immersive content rather than the existing high-definition immersive content. The proposed technique selects an I frame that does not have a reference frame in HEVC, allocates it to the CUDA (Compute Unified Device Architecture) core in a GOP (Group Of Picture) unit, and performs parallel processing on the GPU.