Mulyadi, Hartawan Bahari (2026) Perlindungan Privasi Citra Digital dari Model Klasifikasi menggunakan Integrasi Image Poisoning dan Steganografi DCT-QIM dengan Secret Embedding Key. Masters thesis, Institut Teknologi Sepuluh Nopember.
![[thumbnail of 6025241067-Master_Thesis.pdf]](http://repository.its.ac.id/style/images/fileicons/text.png) |
Text
6025241067-Master_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (4MB) | Request a copy |
Abstract
Perkembangan sistem klasifikasi citra berbasis kecerdasan buatan (AI) meningkatkan risiko pelanggaran privasi karena citra digital dapat dianalisis dan diidentifikasi tanpa izin. Guna mengatasi hal tersebut, penelitian ini mengusulkan metode perlindungan citra yang menggabungkan image poisoning menggunakan GAN berbasis MIM-PGD dengan steganografi DCT-QIM. Perturbasi halus diterapkan pada citra untuk mengacaukan prediksi model AI tanpa menurunkan kualitas visual. Area perturbasi tersebut kemudian dimanfaatkan sebagai lokasi penyisipan pesan rahasia, sehingga proteksi dan komunikasi tersembunyi dapat dilakukan secara bersamaan.
Proses penyisipan dikendalikan oleh secret embedding key yang memetakan area embedding secara deterministik agar pesan dapat diekstraksi secara akurat dan citra tetap dapat dipulihkan (reversible). Evaluasi dilakukan menggunakan metrik Peak Signal-to-Noise Ratio (PSNR), Structural Similarity Index Measure (SSIM), Defence Success Rate (DSR), dan Reversibility Accuracy (RA) pada dataset ImageNet, CIFAR-100, dan USC-SIPI. Pendekatan ini dirancang untuk meningkatkan keamanan citra, menjaga kualitas visual, serta memungkinkan penyisipan pesan yang stabil dan sulit dideteksi, sehingga digunakan pada lingkungan digital yang membutuhkan perlindungan privasi tinggi.
=======================================================================================================================================
The advancement of AI-based image classification systems increases the risk of privacy violations, as digital images can be analyzed and identified without authorization. To address this issue, this study proposes an image protection method that integrates image poisoning using GAN-based MIM-PGD with DCT-QIM steganography. Subtle perturbations are applied to the image to disrupt AI model predictions without degrading visual quality. These perturbation areas are then utilized as embedding regions for hiding secret messages, enabling both protection and covert communication simultaneously.
The embedding process is controlled by a secret embedding key that deterministically maps the embedding areas, ensuring that the hidden message can be accurately extracted and the image can be restored (reversible). The method is evaluated using Peak Signal-to-Noise Ratio (PSNR), Structural Similarity Index Measure (SSIM), Defence Success Rate (DSR), and Reversibility Accuracy (RA) on the ImageNet, CIFAR-100, and USC-SIPI datasets. This approach is designed to enhance image security, preserve visual quality, and support stable and hard-to-detect message embedding, making it suitable for digital environments that require strong privacy protection.
| Item Type: | Thesis (Masters) |
|---|---|
| Uncontrolled Keywords: | Domain Transformasi, Image Poisoning, Steganografi, Perturbasi, Perlindungan Citra Digital |
| Subjects: | Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science) Q Science > QA Mathematics > QA76.9 Computer algorithms. Virtual Reality. Computer simulation. Q Science > QA Mathematics > QA76.9.A25 Computer security. Digital forensic. Data encryption (Computer science) Q Science > QA Mathematics > QA9.58 Algorithms T Technology > T Technology (General) > T57.5 Data Processing T Technology > TA Engineering (General). Civil engineering (General) > TA1637 Image processing--Digital techniques. Image analysis--Data processing. |
| Divisions: | Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55101-(S2) Master Thesis |
| Depositing User: | Hartawan Bahari Mulyadi |
| Date Deposited: | 29 Jul 2026 02:21 |
| Last Modified: | 29 Jul 2026 02:21 |
| URI: | http://repository.its.ac.id/id/eprint/139478 |
Actions (login required)
 |
View Item |