Page 47 - 2025S

40 UEC Int’l Mini-Conference No.54

Data Hiding Components for Solving Information Security Issues in
DICOM Medical Images

Juan Eduardo Mosco Garcia * 1 and Hayaru Shouno 2

1 UEC Exchange Study Program (JUSST Program)
2 Department of Informatics
The University of Electro-Communications, Tokyo, Japan

Abstract

In the context of secure medical image processing, patient authentication and data confidentiality are critical
challenges. This study presents an enhanced reversible data hiding approach in DICOM images by leveraging
halftone-based watermarking, deep learning denoising, and robust segmentation. We introduce a preprocessing
stage where the grayscale medical image is first converted into a halftone image using Jarvis error-diffusion
technique. A segmentation algorithm then isolates bright regions—ideal for embedding. Binary data is inserted
into the white areas using a modified version of the ”Data Hiding in Binary Images with High Payload” method.
To preserve visual quality, a denoising neural network restores the image with minimal distortion. The em-
bedding process follows the fragile-watermarking scheme, combining chaotic encryption with LSB insertion.
This enables embedding large watermarks and detecting tampering, as any alteration corrupts the embedded
data. Additionally, it allows extraction of the modified region. Experimental results show the proposed method
achieves a PSNR of up to 67.00 dB when comparing the restored grayscale image with the original, indicating
excellent visual fidelity and robustness against distortions such as salt-and-pepper noise. This work enhances
the efficiency and capacity of data embedding in medical images, while ensuring secure patient identification
and tamper detection.

Keywords: medical images, reversible watermarking, denoising network, data hiding, patient authentication
1 Introduction Among existing strategies, reversible data hiding
(RDH) provides the added advantage of restoring
In modern healthcare systems, the protection of the original image after the embedded information
patient data integrity and confidentiality has be- is extracted. This is crucial for clinical contexts,
come increasingly critical. The digitization of med- where even minimal alterations can lead to misdi-
ical records and the widespread adoption of Picture agnoses. Various RDH approaches, such as differ-
Archiving and Communication Systems (PACS) and ence expansion, histogram shifting, and interpola-
Hospital Information Systems (HIS) have brought tion, have been proposed [4–6], yet challenges re-
numerous benefits, but have also introduced signif- main in balancing high payload, robustness, and im-
icant vulnerabilities. Traditional security measures perceptibility.
such as firewalls and encryption are insufficient to
address the full scope of risks related to unautho- This study presents a reversible data hiding
rized access and data manipulation in DICOM im- method for DICOM medical images using halftone
ages [1–3]. To mitigate these challenges, data hid- watermarking with enhanced embedding capacity
ing techniques have emerged as powerful tools for and visual fidelity. Inspired by the ”Data Hiding
secure information embedding, ownership authenti- in Binary Images with High Payload” method [7],
cation, tamper detection, and patient identification. our approach introduces several key innovations.
First, the original RGB patient image is binarized
* The author is supported by JASSO Scholarship. using Jarvis error-diffusion method, chosen for its