AuCFSR: Authentication and Color Face Self-Recovery Using Novel 2D Hyperchaotic System and Deep Learning Models

Area of Science:

• Digital Image Forensics
• Cybersecurity
• Computer Vision

Background:

• Color face images transmitted over public networks are susceptible to malicious tampering.
• Existing methods may lack robust authentication and effective recovery capabilities for compromised facial imagery.

Purpose of the Study:

• To develop a novel scheme, Authentication and Color Face Self-Recovery (AuCFSR), for authenticating color face images and recovering tampered regions.
• To enhance the security and integrity of digital facial data against unauthorized modifications.

Main Methods:

• AuCFSR embeds authentication and recovery data using a novel two-dimensional modular sine-cosine map (2D MSCM) hyperchaotic system.
• Data is concealed within the least significant bits of color image pixels.
• Deep learning models, CodeFormer and DeOldify, are employed for visual enhancement and colorization of recovered images.

Main Results:

• The proposed AuCFSR scheme effectively detects tampered color face images.
• It achieves a high level of security and produces high-quality output images.
• Recovered images exhibit improved visual quality and colorization.

Conclusions:

• AuCFSR offers superior performance in tamper detection accuracy, security, and visual recovery compared to existing schemes.
• The integration of hyperchaotic systems and deep learning provides a robust solution for color face image authentication and self-recovery.