Robust Image Forgery Classification using SqueezeNet Network

The rapid advancement of digital manipulation tools has made it increasingly challenging to detect image forgery, especially with the emergence of complex and realistic deep imaging forgeries. Traditional techniques often struggle to identify these sophisticated manipulations. This research focuses on leveraging the SqueezeNet deep learning network to detect deep image forgeries. By training the SqueezeNet network on a dataset of real and fake images, distinctive features can be learned to enhance the detection process. The effectiveness of the SqueezeNet model is demonstrated through evaluation on a separate test dataset, although its performance may vary depending on the complexity of the forgeries. This study sheds light on the potential of using the SqueezeNet deep learning network to combat deep picture fraud and improve image analytics systems. The evaluation results with distinguishing forgery and original images with an accuracy of 98% which is 3-10% increase when compared with the other deep learning methods.