ABSTRACT
A lot of fraudulent and manipulated photographs have been created in recent years and disseminated through the media and the Internet due to the accessibility and simplicity of image editing softwares. To judge the veracity of an image and, in some situations, to identify the parts that have been altered (forged), a variety of methods have been put forth. The most prevalent copy-move and splicing attacks are the focus of this paper’s examination of some of the most recent methods for detecting image forgeries that specifically rely on DL techniques. Deep learning-powered techniques seem to be the most relevant at the moment, as evidenced by the best overall performances they exhibit on the benchmark datasets that are now available, so this survey is especially pertinent at this time. These approaches’ key characteristics were discussed in addition to the datasets that were used to train and test them.
