This work was published in SPIE, showcasing its contribution to the field of digital watermarking and ownership protection.

As part of my Master’s thesis in Computer Science & Engineering, I designed and published a deep-learning-based watermarking algorithm leveraging convolutional neural networks (CNNs). The research focused on embedding and extracting watermarks in digital images, ensuring both invisibility and robustness. We proposed an autoencoder CNN architecture comprising two networks—embedding and extraction—to evaluate performance comprehensively. By employing advanced techniques like feature map concatenation and block-level transposed convolutions, the algorithm demonstrated superior resistance to various attacks while maintaining low computational cost. Comparative experiments confirmed its effectiveness, outperforming state-of-the-art approaches in robustness and invisibility.