Forensic Watermarking Concepts and Specifications

Describes various concepts and specifications related to DoveRunner Forensic Watermarking service.

Forensic Watermarking

This technology is designed to track illegal distribution of multimedia content such as videos, images, and audio. By inserting information that can not be detected by the human eye into the content, the user information can be extracted from the leaked content.

System Architecture for SaaS-based preprocessing integration

graph LR;
    subgraph Content Service Provider
    A[Source
Content] --> B(T&P service)
    B -->|A/B variants
DASH/HLS| C(Origin
storage)
    C -->|A segments| D(CDN Edge
Embedding)
    C -->|B segments| D
    G(Service
Backend)
    end
    subgraph End user
    G --> |Session URL| E(Client
player)
    D -->|Watermarked
stream| E
    end
    subgraph DoveRunner service
    G -->|Session data| F(Watermark
session manager)
    F -->|Session URL| G
    end

System Architecture for 3rd Party Encoder/Transcoder integration

graph LR;
    subgraph Content Service Provider
    A[Source
Content] -->|Frame
image| B(Preprocessor
plugin)
    subgraph Encoder/Transcoder with Preprocessor plugin
    B -->|A/B
frame| C(Encoding)
    C -->|A/B
mp4| D(Packaging)
    end
    D -->|A/B variants
DASH/HLS| E(Origin
Storage)
    E -->|A segments| F(CDN Edge
Embedding)
    E -->|B segments| F
    H(Service Backend
CMS)
    end
    subgraph End User
    H --> |Session URL| G(Client
player)
    F -->|Watermarked
stream| G
    end
    subgraph DoveRunner Service
    H -->|Session data| I(Watermark
session manager)
    I -->|Session URL| H
    end

Forensic Watermarking Tech Demo

In the video below, you can see the process of embedding, mixing and detecting the watermark using forensic watermarking technology. This demo is intended to help you understand DoveRunner Forensic Watermarking solution , and it is different from the scenarios that apply to actual commercial services.

For optimal playback, select ‘1080p’ as the video quality and select subtitle (Korean or English) for playback.

Watermark Preprocessing

In order to embed a watermark in the original video, a preprocessing operation is required in the content encoding process. Different values (0/1 or A/B) are inserted in the original uncompressed video frames and output as two sets (A/B variants) of encoded video.

The inserted watermark value is encrypted using a secret key unique to each content service provider, and can not be extracted from the outside arbitrarily.

The watermark preprocessing can be done by Transcoding & Packaging (T&P) Service or pre-integrated 3rd party encoder/transcoder solutions depending on the customer’s choice.

Go to Watermark Preprocessing Guide

Transcoding & Packaging Services

Content transcoding, watermark preprocessing, and DRM packaging services provided in the form of cloud SaaS. Without the need to build a separate server for watermark preprocessing, you can use DoveRunner cloud service to transcode the original video and apply multi-DRM and forensic watermarking while packaging it.

Preprocessing SDK

It is a watermark preprocessor implemented as a C ++ library so that it can be ported to an encoding solution. By applying the preprocessing directly in the video encoding process, you can omit the re-encoding process required for CLI preprocessor.

By default, porting to FFmpeg-based encoders is supported, and work with encoder solution vendors is required to apply them to other commercial encoders.

Watermark Embedding

The A/B variants of video (marked with ‘0’ or ‘1’) are combined real-time on playback. It converts session information, such as user ID, to binary format, and thus composes the output stream by mixing A/B version segments in the DASH or HLS content.

As a result, unique information about the playback session is inserted invisibly in the video and sent to the client.

Note

The segment list of the DASH manifest (.mpd) or HLS playlist (.m3u8) passed as a Session URL applies the same path and filename to all users, regardless of session information. Therefore, it is safe from segment naming attack which is mixing the segment lists of multiple sessions for the same content.

Go to Watermark Embedding Guide

Session Data

Various client information such as user ID, content ID and time of playback. Depending on the needs of the service site, you can apply any session data as a string up to 254 bytes length.

Watermark Payload

Up to 7 bytes of payload data can be embedded into the final content with a watermark combination of 0 and 1. Actual session data such as user ID is managed in a separate database in the DoveRunner service, and the payload is used as the key of the session data.

Watermark Embedder

This module is applied to the CDN of the service site. When the content is requested for playback on the client, the watermarked video is streamed by mixing A/B version segments. It delivers the segments in a unique order according to the watermark payload value for the session information.

By default, it supports the Lambda@Edge module of the Amazon CloudFront CDN. And Akamai CDN is also pre-integrated so you can configure it to enable the DoveRunner integration.

Session URL

It is a content URL which is used for playback of watermarked content stream. Depending on the type of streaming, a DASH manifest(.mpd) or HLS playlist(.m3u8) URL is used, and an encrypted watermark payload is added as part of the unique session-based URL.

Session Manager

It is a function of DoveRunner cloud server which generates Session URL for watermark embedding. It stores session information received from service site in session database and responds a Session URL which has encrypted payload.

Watermark Detecting

If the watermarked video is leaked and distributed illegally, you can request the DoveRunner service to detect the suspicious content. Watermark detection analyzes each frame of the video to detect the original watermark pattern and decrypts the data with the secret key used at the time of insertion.

If the watermark payload is found through the detection process, the session database finds the session information whose key value is the payload and reports it as the detection result.

Note

Our forensic watermarking solution requires at least 5 minutes of recorded video to detect watermark. For other detailed requirements for detection, please refer to the Support Specifications section below.

Go to Watermark Detecting Guide

Supported Specifications

DoveRunner Forensic Watermarking supports the below specifications.

Item	Specification	Remarks
Content Type	Supports video contents only
Format and codec	MP4 (CBR) VOD video, H.264/H.265 codec supported	Live video will be supported later
Supported Resolution	Up to 4K/UHD resolution
Supported FPS	23.976, 24, 25, 29.97, 30, 47.952, 48, 50, 59.94, 60
Dynamic Range	SDR (Standard Dynamic Range), HDR (High Dynamic Range) supported	Requires H.265 to support HDR
Streaming Protocol	MPEG-DASH, HLS (Http Live Streaming)
CDN Integration	Amazon CloudFront, Akamai, and Fastly CDN supported	Contact us for other CDN integration
Watermark Data Size	Up to 255 bytes string using mapping table
DRM Integration	Multi DRM (PlayReady, Widevine, FairPlay) supported

The detailed requirements for watermark detection are as follows:

Item	Description
Minimum video length	In order to detect watermarks, continuous recorded video without repeating sections with a length of at least 5 minutes is required
Video quality	At least 480p with 1Mbps is required for detection. 720p or higher quality recommended
Video stability	Requires fixed recorded video without shaking. Undetectable if the screen is shaken due to shooting with a handheld camera or smartphone
Buffering or freezing	There must be a section that was played normally without buffering or screen freezing for at least 5 minutes