Article Per Year (5 Year)
p-Index From 2021 - 2026
0.23
P-Index
This Author published in this journals
All Journal Transaction on Informatics and Data Science
Riza Arifudin
Computer Science Department, Fakultas Matematika dan Ilmu Pengetahuan Alam, Universitas Negeri Semarang, Semarang, Indonesia
Computer Science & IT Decision Sciences, Operations Research & Management Engineering

Published : 1 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 1 Documents
Search
Journal : transaction on informatics and data science

 1 
Enhancing Image Tampering Detection Using ResNet50 with Error Level Analysis and Gamma Correction Hanief Widya Wardhana; Riza Arifudin
Transactions on Informatics and Data Science Vol. 3 No. 2 (2026)
Publisher : Department of Informatics, Faculty of Science and Technology, UIN Prof. K.H. Saifuddin Zuhri, Purbalingga, Indonesia

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.24090/tids.v3i2.16322

Abstract

Image manipulation has become increasingly sophisticated, making the verification of visual authenticity a critical challenge for preventing misinformation, fraud, and other malicious activities. Conventional detection methods often fail to capture subtle traces of tampering, motivating the use of deep learning combined with targeted preprocessing. This study proposes a detection approach that couples the ResNet50 architecture with two complementary preprocessing techniques—Error Level Analysis (ELA) and gamma correction—to sharpen the features associated with tampered regions. ELA exposes discrepancies in JPEG compression levels, while gamma correction enhances the contrast between authentic and manipulated areas; combining them yields a more discriminative feature space than either technique alone. On the CASIA 2.0 dataset (12,614 images), the model attains 82.60% accuracy without preprocessing, 95.16% with ELA, and 78.74% with gamma correction alone. Applying both techniques together produces the highest accuracy of 97.58%, a 14.98 percentage-point improvement over the baseline, because ELA isolates compression-level traces of tampering while gamma correction sharpens the contrast that makes those traces separable. These results demonstrate that carefully selected preprocessing significantly strengthens ResNet50-based image tampering detection.
Co-Authors Hanief Widya Wardhana