Abstract
With the recent development of Cognitive Internet of Things (CIoT) and the potential of Cyber Physical System (CPS), people's daily activities become smarter, and intelligent. The combination of CIoT and CPS can greatly enhance the quality of people's life. To this end, this article proposes CIoT-CPS that comprises of two main models: user activity cognitive model and image authentication model.The user activity cognitive model (UACM) is a machine-learning model to have the meaningful data. The image authentication model is to verify the authenticity of images captured by various devices, such as smart phones, digital cameras, and other camera-embedded portable devices. The authenticity of an image is breached when parts of images are assembled to produce a new image (known as a splicing forgery), or a part of an image is copied or pasted into another part of the same image (known as a copy-move forgery). In the proposed verification method, an opposite color local binary pattern (OC-LBP) texture descriptor is applied to a questioned image. The image is first decomposed into an RGB (red, green, blue) and a luminance and chroma color spaces. The OC-LBP measures the interrelation between pixels of different color components. The intensive computation involving six color components and a gray version is performed in the cloud, where a server can be dedicated to doing this job. The histograms of the OC-LBP are concatenated with weights to produce a final feature vector of the image. A support vector machine is applied as a classifier, which classifies the image as authentic or forged. Several experiments were performed to verify the suitability of those models or approaches. The proposed approaches show a good accuracy compared to other competing approaches.