Evaluation of Superior Accuracy of Faster RCNN over Conventional Forensic Techniques in Image Tampering Detection

The demand for digital image fraud detection is rapidly increasing in society due to the importance of having certified images. The analysis encompasses the identification of picture copy move, image splicing, image retouching, and image resampling forgeries. A tampering detection system is crucial for detecting image tapering and verifying the validity of photos. In this work, an effective algorithm called as Faster Recurrent Convolutional Neural Network (F-RCNN) is proposed to enhance the accuracy of the detection of an image tampering. Despite the fact that several detection technologies are now available, their accuracy and reliability fall well short of expectations. Current methods rely on a single network topology with many neurons and deep layers, which is insufficient for detecting forgeries. The proposed algorithm is working well with the detection system of image tampering with the integration of Naive Byes classifier along with the proposed algorithm. The faster RCNN is consists of deep layers of CNN that is significance to extract features and to detect the tampered image. The proposed algorithm ensures its efficient detection of an image tampering by providing the accuracy level of $\mathbf{92 \%}$ with the various levels of epochs.