An Independent Discriminant Network Towards Identification of Counterfeit Images and Videos
This work addresses the spread of false information online by detecting GAN-generated forgeries, which is an incremental improvement in the domain of digital forensics.
The paper tackles the problem of detecting counterfeit images and videos generated by GANs, proposing an independent discriminant network based on InceptionResNetV2 that can identify such forgeries, with potential applications in forensics for criminal activity identification.
Rapid spread of false images and videos on online platforms is an emerging problem. Anyone may add, delete, clone or modify people and entities from an image using various editing software which are readily available. This generates false and misleading proof to hide the crime. Now-a-days, these false and counterfeit images and videos are flooding on the internet. These spread false information. Many methods are available in literature for detecting those counterfeit contents but new methods of counterfeiting are also evolving. Generative Adversarial Networks (GAN) are observed to be one effective method as it modifies the context and definition of images producing plausible results via image-to-image translation. This work uses an independent discriminant network that can identify GAN generated image or video. A discriminant network has been created using a convolutional neural network based on InceptionResNetV2. The article also proposes a platform where users can detect forged images and videos. This proposed work has the potential to help the forensics domain to detect counterfeit videos and hidden criminal evidence towards the identification of criminal activities.