Secured Video Stream Face Portion Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies are ineffective in reliably detecting and preventing the creation of synthetic media, such as deepfake videos, which can be misleading and difficult to recognize, as they fail to secure existing videos of subjects that could be used to generate realistic synthetic content.
Innovation Solution
A media hosting system that automatically identifies and extracts face portions from videos, stores them securely, and dynamically combines protected face portions with video streams to prevent unauthorized access and creation of synthetic media, ensuring only approved media players can access the protected content.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If existing videos of subjects are made publicly accessible for media streaming, then media availability and accessibility are improved, but the risk of synthetic media creation increases
Solution Approach 1:
The video is segmented into two separate components: face portions and non-face portions. These segments are stored and transmitted separately, with face portions encrypted and requiring authentication. This segmentation allows the system to provide broad media accessibility while protecting vulnerable face regions from unauthorized extraction and synthetic media creation.
Solution Approach 2:
The face portions are extracted and removed from the main video stream. Instead of being included in the publicly accessible media stream, they are stored separately in an encrypted form and only provided to authenticated users. This extraction eliminates the vulnerability of face portions in the public domain while maintaining overall media accessibility.
2Reliability
If face portions are extracted and stored securely with encryption, then protection against synthetic media creation is improved, but system complexity increases
Solution Approach 1:
Face portions are extracted, encrypted, and stored in a secure database before being needed. Authentication credentials are prepared in advance. When a user requests media, the system already has the encrypted face portions ready to be selectively provided based on authentication results. This preliminary action simplifies the real-time processing while maintaining strong security.
Solution Approach 2:
An authentication credential system acts as an intermediary between the encrypted face portions and users. The credentials verify user identity and authorize access to specific face portions. This intermediary layer manages the complexity of secure access control while keeping the underlying encryption and extraction mechanisms relatively simple.
3Object-affected harmful factors
If face portions are removed from public media streams, then prevention of deepfake creation is improved, but video completeness for public viewing deteriorates
Solution Approach 1:
The system dynamically adjusts what content is provided based on user authentication status. Unauthenticated users receive media streams with face portions removed or replaced with placeholders. Authenticated users receive the complete video with all face portions restored. This dynamic approach allows the system to maintain video completeness for authorized users while preventing deepfake creation by keeping face portions inaccessible to unauthorized users.
Data Source
AI summary
A system receives an audio-video recording that includes image frames. For each image frame, a portion of the image frame is identified that corresponds to a first face of a first subject. An image region is determined for the identified portion of the image frame. A first face portion of the audio-video recording is stored that corresponds to the determined image region for each of the image frames along with a private key for the audio-video recording and a first mask identifier for the first subject. A public key is generated for the private key. A remaining portion of the audio-video recording is associated with the generated public key and the first mask identifier for the first subject. The remaining portion, the public key, and the first mask identifier are provided to a media server that communicates with a media player.