ASCA-squeeze net: Aquila sine cosine algorithm enabled hybrid deep learning networks for digital image forgery detection

The basic element in resolving numerous challenges, particularly social concerns like those in court cases and Facebook, is image forgery detection. The primary objective of this study is to build and develop an effective system for detecting digital image forgery utilising the recently proposed tec...

Full description

Saved in:
Bibliographic Details
Published in:Computers & security 2023-05, Vol.128, p.103155, Article 103155
Main Authors: Nirmalapriya, G., Maram, Balajee, Lakshmanan, Ramanathan, Navaneethakrishnan, M.
Format: Article
Language:English
Subjects:
Aquila optimizer
Congruence coefficient
Deep convolutional neural network
Sine cosine algorithm
Squeeze net
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:The basic element in resolving numerous challenges, particularly social concerns like those in court cases and Facebook, is image forgery detection. The primary objective of this study is to build and develop an effective system for detecting digital image forgery utilising the recently proposed technique called the Aquila Sine Cosine Algorithm (ASCA). The forgery from the digital image is detected in this study using a hybrid deep learning technique that incorporates Deep Convolutional Neural Network (DCNN) and Squeeze Net. Additionally, the training time and computational complexity of the detection process are decreased by updating the weight of both the DCNN and the Squeeze Net using the developed ASCA technique. Additionally, the developed ASCA is produced by combining the update functions of the Aquila Optimizer (AO) with the Sine Cosine Algorithm (SCA). As a result, the hybrid deep learning classifier provides the classified output as either the authentic image or the forged image using a copy-move forgery detection dataset. The experimentation of the developed model has provided higher performance, as shown by testing accuracy, True Positive Rate (TPR), and True Negative Rate (TNR) of 0.980, 0.976, and 0.956, respectively. Furthermore, by varying the iteration, testing accuracy, TNR, and TPR obtained by the devised technique are 0.944, 0.947, and 0.936, and by varying the population size obtained testing accuracy values of 1, TNR values of 1.003, and TPR values of 0.991, respectively, by algorithmic analysis.
ISSN:0167-4048
1872-6208
DOI:10.1016/j.cose.2023.103155