Error Level Analysis and Deep Learning For Detecting Image Forgeries
Dipak Agrawal, Hitesh Makwana, Shrinal S Dave, Sheshang Degadwala, Vidhi Desai · 2023
Social media is the major image sharing resource around the internet. On the other hand, the increasing availability of photo-editing software has led to the development of fake images. Since the compression ratio of the original and false images differ, this study has a developed a novel error level analysis model to determine the true compression ratio and detect the fake images. Image data analysis is another way available to determine its authenticity, though the metadata could have been altered. Here, the Deep Learning (DL) models are used to differentiate the real and fake images present in the dataset, as well as other parameters required for error rate analysis. For 10 epochs, the proposed model has produced the best training accuracy of 90.59%.