An Improved Novel Framework for Image Forgery Detection Using CNN-LSTM Deep Learning Model

Abstract

In recent years, the popularity of capturing images has increased. The rapid advancement of technology has made effective image processing tools more accessible and making image forgery easier. Identifying real images from forged images has become an alarming obstacle. To detect image forgeries, numerous traditional techniques have been developed over time. Obtained efficiency and type of forgery and/or the image features. Image forgery detection uses different techniques depending on the detection requirements. This paper presents An Improved Novel Framework for Image Forgery Detection using a deep learning model based on CNN (Convolutional Neural Networks) and LSTM (Long Short-Term Memory). The described model uses CASIA.2.0 (Chinese Academy of Sciences, Institute of Automation), the database of image forgery, which contains a total of 12614 images, 5123 are Tempered Images, and the remaining 7491 are Genuine Images. In the image segmentation process, the K-means clustering model is utilised. Grey Level Co-occurrence Matrix (GLCM) features are then calculated to differentiate between the original and forged images. From the comparative results, the image forgery detection has performed better, achieving 98.5% Accuracy, 98% Precision, 97% Recall, and 97% Specificity, using the described CNN-LSTM Deep Learning Model, compared to other algorithms.