A coprime blur scheme for data security in video surveillance

Area of Science:

• Computer Science
• Image Processing
• Cryptography

Background:

• Traditional video encryption is computationally intensive.
• Existing methods often lack a balance between privacy and accessibility.
• Surveillance systems require robust data security measures.

Purpose of the Study:

• To propose a novel coprime blurred pair (CBP) model for enhanced data security in camera surveillance.
• To introduce a spatial encryption scheme using strategically blurred image/video content.
• To enable selective access to surveillance data while maintaining monitoring capabilities.

Main Methods:

• Developing a coprime blurred pair (CBP) model based on bivariate polynomial mapping in the (z) domain.
• Utilizing two distinct blurring kernels to create public and private video streams.
• Employing Bézout matrix rank analysis to ensure kernel coprimality and enable latent image recovery.
• Implementing a graphics processing unit (GPU) accelerated decryption scheme for real-time performance.

Main Results:

• Demonstrated effective protection of sensitive identity information in surveillance videos.
• Achieved faithful reconstruction of unblurred video streams using both CBP sequences.
• Validated real-time performance through GPU implementation.
• Showcased the practical applicability of the CBP model in surveillance scenarios.

Conclusions:

• The proposed CBP model offers an effective spatial encryption solution for camera surveillance.
• This approach balances data privacy with the need for continuous monitoring and forensic analysis.
• The real-time GPU implementation makes the solution practical for widespread adoption in security systems.