Screen-Cam Imitation Module for Improving Data Hiding Robustness.

Area of Science:

• Computer Science
• Cryptography
• Machine Learning

Background:

• Improving the robustness of neural network-based data-hiding schemes is crucial.
• Existing attack simulators often fail to capture complex, real-world distortions like those in screen-cam scenarios.
• Manual replication of screen-cam attacks is time-consuming and limits scheme development.

Purpose of the Study:

• To propose a novel generator neural network for simulating screen-cam distortions.
• To demonstrate the utility of this simulator in enhancing data-hiding scheme robustness.
• To provide a more efficient and accurate method for evaluating data-hiding security.

Main Methods:

• Developed a generator neural network to model complex screen-cam distortions.
• Integrated the simulator into an attack-simulation module for data-hiding schemes.
• Evaluated the impact on an existing data-hiding scheme's robustness.

Main Results:

• The generator network successfully simulated screen-cam distortions, replacing manual experiments.
• The robustness of the tested data-hiding scheme was improved by 15% in terms of bit error rate.
• The proposed method offers a more efficient approach to developing and evaluating data-hiding techniques.

Conclusions:

• Generator neural networks can effectively simulate complex real-world distortions for data-hiding research.
• The developed simulator enhances the robustness of data-hiding schemes against screen-cam attacks.
• This approach facilitates the creation of more secure and practical data-hiding solutions.