Image Forgery Detection using Cryptography and Deep Learning

Oke, Ayodeji; Babaagba, Kehinde O.

Abstract

The advancement of technology has undoubtedly exposed everyone to a remarkable array of visual imagery. Nowadays, digital technology is eating away the trust and historical confidence people have in the integrity of imagery. Deep learning is often used for the detection of forged digital images through the classification of images as original or forged. Despite many advantages of deep learning algorithms to predict fake images such as automatic feature engineering, parameter sharing and dimensionality reduction, one of the drawbacks of deep learning emanates from parsing bad examples to deep learning models. In this work, cryptography was applied to improve the integrity of images used for deep learning (Convolutional Neural Network-CNN) based prediction using SHA-256. Our results after a hashing algorithm was used at a threshold of 0.0003 gives 73.20% image prediction accuracy. The use of CNN algorithm on the hashing image dataset gives a prediction accuracy of 72.70% at 0.09s. Furthermore, the result of CNN on the raw image dataset gives a prediction accuracy of 89.08% at 2s. The result shows that although a higher prediction accuracy is obtained when the CNN algorithm is used on the raw image without hashing, the prediction using the CNN algorithm with hashing is faster.

• Presentation Conference Type: Conference Paper (Published)
• Conference Name: EAI BDTA 2023 - 13th EAI International Conference on Big Data Technologies and Applications
• Start Date: Aug 23, 2023
• End Date: Aug 24, 2023
• Acceptance Date: Jul 27, 2023
• Online Publication Date: Jan 31, 2024
• Publication Date: 2024
• Deposit Date: Aug 16, 2023
• Publicly Available Date: Feb 1, 2025
• Publisher: Springer
• Pages: 62-78
• Series Title: Lecture Notes of the Institute for Computer Sciences, Social Informatics and Telecommunications Engineering
• Series Number: 555
• Series ISSN: 1867-8211
• Book Title: Big Data Technologies and Applications. BDTA 2023
• ISBN: 978-3-031-52264-2
• DOI: https://doi.org/10.1007/978-3-031-52265-9_5
• Keywords: Image Forgery Detection, Machine Learning, Deep Learning, Cryptography, Hashing
• Public URL: http://researchrepository.napier.ac.uk/Output/3169411
• Related Public URLs: https://infoscale.eai-conferences.org/2023/