AI Video Call Screening for Sensitive Data Obfuscation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Video calls and conferences pose risks of unintended sensitive data disclosure through backgrounds or unintended participants, including text, images, or audio, which existing technologies fail to adequately protect.
Innovation Solution
Implementing Artificial Intelligence (AI) with computer vision, optical character recognition (OCR), facial recognition, and voice recognition, along with noise reduction techniques, to detect and prevent the viewing or hearing of sensitive data in video and audio feeds during conferences.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If AI monitoring is implemented to detect sensitive data in video and audio feeds, then security against sensitive data leakage is improved, but device complexity increases
Solution Approach 1:
An AI-based monitoring system is introduced as an intermediary component that analyzes video and audio feeds to detect sensitive data. This mediator processes the feeds independently and triggers notifications or actions only when sensitive information is detected, rather than requiring complex real-time processing of all participant interactions.
Solution Approach 2:
Traditional manual monitoring of video conferences for sensitive data is replaced with automated AI-based detection systems. The patent employs machine learning models and computer vision algorithms to automatically identify sensitive information in video feeds and audio signals, substituting human review with automated technological detection.
2Reliability
If real-time detection and obfuscation of sensitive data is performed, then sensitivity of sensitive data protection is improved, but processing time increases
Solution Approach 1:
The system performs preliminary analysis of video and audio feeds by continuously monitoring for patterns indicative of sensitive data. AI models are pre-trained to recognize sensitive information types, allowing the system to quickly identify and respond to potential leaks without requiring extensive real-time processing when no sensitive data is present.
Solution Approach 2:
The monitoring system selectively processes only those portions of video and audio feeds that contain potential sensitive data. When sensitive information is detected, the system rapidly applies obfuscation or notification mechanisms, skipping unnecessary processing of normal, non-sensitive content to minimize processing time overhead.
3Measurement precision
If AI techniques including computer vision and voice recognition are implemented to detect sensitive data, then detection precision is improved, but device complexity increases
Solution Approach 1:
The detection system is divided into separate specialized modules: computer vision components for analyzing video feeds, voice recognition components for processing audio, and natural language processing components for identifying sensitive information patterns. Each module focuses on a specific detection task, improving overall precision while allowing independent optimization and maintenance of each component.
Solution Approach 2:
The AI monitoring system is designed to detect multiple types of sensitive data across different modalities (video, audio, text) using a unified platform. The system can identify various forms of sensitive information including personal identifiable information, financial data, and confidential business information through multiple detection techniques, providing versatile protection without requiring separate systems for each data type.
Data Source
AI summary
Protection of sensitive data, such as Non-Public Information (NPI) in a video call/conference environment. In response to initiating a video call/conference amongst multiple call participants, Artificial Intelligence is implemented to monitor for detection of sensitive data in the video feed and/or audio feed being transmitted by a call participant. In response to detecting sensitive data in the video feed or the audio feed, the invention is configured to perform one or more actions that prevent one or more other call participants participating in the video call from viewing and/or hearing the sensitive data in the video feed or the audio feed. The sensitive data may be found in text indicia or images displayed in the video call field of view, individuals heard or seen in the background of the video call or the like.