Skripsi

PENGEMBANGAN MODEL STEGANOGRAFI CITRA BERBASIS U-NET DENGAN CONVOLUTIONAL SPATIAL ATTENTION DAN RESIDUAL MAP PADA Y-CHANNEL

---

This study proposes an image steganography architecture based on a modified U-Net integrated with Convolutional Spatial Attention (CSA) on the cover stream and a Laplacian-based Residual Map on the Y-channel, designed to produce an embedding probability map that more selectively targets textured regions. The model generates a residual embedding which is inserted using a ±1 modulation scheme with top-K location selection guided by the learned probability values.Experiments conducted on the BOSSBase v1.01 and BOWS2 datasets with payloads ranging from 1–5% demonstrate PSNR values above 40 dB, SSIM values exceeding 0.98, and BER values approaching zero, indicating excellent imperceptibility and highly accurate bitstream extraction. Security evaluation using SRNet shows reduced detection sensitivity, confirming that the resulting embedding patterns are more adaptive and more resistant to identification by CNN-based steganalysis. These findings demonstrate that the combination of U-Net, CSA, and residual learning in the luminance domain significantly enhances representational efficiency and embedding security in modern generative steganography systems.

---

Availability

Detail Information

Series Title

-

Call Number

T1897202025

Publisher

Indralaya : Prodi Teknik Informatika, Fakultas Ilmu Komputer Universitas Sriwijaya., 2025

Collation

xviii, 231 hlm.; ilus.; tab.; 29 cm

Language

Indonesia

ISBN/ISSN

-

Classification

005.820 7

Content Type

Text

Media Type

-

Carrier Type

-

Edition

-

Subject(s)

Prodi Teknik Informatika
Steganografi--Data Hiding

Specific Detail Info

-

Statement of Responsibility

MI

Other version/related

File Attachment

Comments

---

You must be logged in to post a comment