Optimasi Steganografi dengan Prediction Error Expansion dan Regresi pada Sinyal Elektrokardiogram

Gautama, Pramudya Tiandana Wisnu (2024) Optimasi Steganografi dengan Prediction Error Expansion dan Regresi pada Sinyal Elektrokardiogram. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6025231061-Master_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 October 2026. Download (4MB) | Request a copy

Abstract

Teknologi telemedicine menjadi salah satu layanan kesehatan yang membutuhkan pertukaran informasi yang tinggi guna membantu diagnosis dan pengamatan yang akurat. Di samping itu, proses transfer memberikan permasalahan mengenai kerentanan pengiriman dan penyimpanan data pasien yang dapat berdampak pada kebocoran informasi pribadi. Salah satu aspek informasi kesehatan pasien yang dapat disalahgunakan adalah sinyal elektrokardiogram (ECG) di mana menampilkan diagnosis penyakit kardiovaskuler. Mekanisme menyembunyikan data personal ke dalam cover sinyal ECG atau steganografi dapat menjadi salah satu solusi. Namun, sinyal ECG memiliki keterbatasan pada proses penyematan di mana perlunya hasil watermark dikembalikan menjadi semula (reversible) serta menjaga struktur sinyal ECG hasil agar tetap memiliki fitur informasi krusial. Di samping itu, kebutuhan steganografi memerlukan kapasitas penyematan bit yang tinggi, konsumsi daya yang rendah, serta perbedaan hasil watermark yang minimal juga menjadi perhatian tersendiri.
Penelitian ini akan dirancang sebuah metode modifikasi prediction error expansion (PEE) dengan penggunaan machine learning. Beberapa model regresi akan dibuat untuk mempercepat proses prediksi error yang ada. Di samping itu, susunan PEE akan dioptimalkan dengan penggunaan konsep mirror-embedding agar mengurangi perbedaan hasil sinyal yang disebabkan oleh penggunaan threshold. Pengujian akan dilakukan dengan membandingkan beragam model regresi yang telah dibuat sebagai prediksi serta melakukan analisis perhitungan berdasarkan perbedaan sinyal original dengan hasil watermark berdasarkan ragam parameter yang ada, seperti model prediktor, besaran payload, dan maksimum sampel bit. Analisis penghitungan akan menggunakan percentage residual difference (PRD), normalized cross-correlation (NCC), dan signal-to-noise ratio (SNR). Setiap model regresi akan dibandingkan waktu rata-rata prediksinya agar didapatkan model terbaik.
============================================================
Telemedicine technology is one of the health services that requires a high level of information exchange to assist in accurate diagnosis and observation. In addition, the transfer process presents problems regarding the vulnerability of sending and storing patient data which can result in personal information being leaked. One aspect of patient health information that can be misused is the electrocardiogram (ECG) signal, which displays a diagnosis of cardiovascular disease. The mechanism of hiding personal data under the cover of ECG signals or steganography could be one solution. However, the ECG signal has limitations in the embedding process where it is necessary to return the watermark results to their original state (reversible) and maintain the structure of the resulting ECG signal so that it still has crucial information features. Apart from that, the needs of steganography requiring high bit embedding capacity, low power consumption, and minimal differences in watermark results are also particular concern.
This research will design a modified prediction error expansion (PEE) method using machine learning. Several regression models will be created to speed up the error prediction process. In addition, the PEE arrangement will be optimized by using mirror-embedding to reduce differences in signal results caused by the utilization of thresholds. Testing will be carried out by comparing various regression models that have been made as predictions and carrying out calculation analysis based on the difference between the original signal and the watermark results on various existing parameters, such as the predictor model, payload size, and maximum sample bits. The calculation analysis will use percentage residual difference (PRD), normalized cross-correlation (NCC), and signal-to-noise ratio (SNR). Each regression model will be compared with the average prediction time to obtain the best model.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	elektrokardiogram, PEE, regresi, reversible, steganografi, electrocardiogram, PEE, regression, reversible, steganography
Subjects:	Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines. Q Science > QA Mathematics > QA278.2 Regression Analysis. Logistic regression Q Science > QA Mathematics > QA76.9.A25 Computer security. Digital forensic. Data encryption (Computer science) R Medicine > RC Internal medicine > RC683.5.E5 Electrocardiography T Technology > T Technology (General) > T57.5 Data Processing
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55101-(S2) Master Thesis
Depositing User:	Pramudya Tiandana Wisnu Gautama
Date Deposited:	08 Aug 2024 08:21
Last Modified:	08 Aug 2024 08:21
URI:	http://repository.its.ac.id/id/eprint/111959

Actions (login required)

View Item