Source Camera Identification Based on Coupling Coding and Adaptive Filter

Abstract

Source Camera Identification (SCI) has been playing an important role in the security field for decades. With the development of Deep Learning, the performance of SCI has been noteworthily improved. However, most of the proposed methods are forensic only for a single camera identification category, e.g., the camera model identification. For exploiting the coupling between different camera categories, we present a new coding method. That is, we apply the multi-task training method to regress the categories, namely, to classify brands, models and devices synchronously in a single network. Different from the common multi-task method, we obtain the multi-class classification result by just one single label classification. To be specific, we classify the categories in a progressive way that the parent category classification result will be used in the child category classification (a detailed explanation will be given later in the main context). Also, by appropriately increasing the redundancy of the coding method for classifying new camera categories, the training time can be greatly reduced. To better extract camera attributes, we propose an adaptive filter. Additionally, we propose an auxiliary classifier that only focuses on the camera model re-classification, due to the low performance of the main classifier on certain models. Lastly, the extensive experiments show that our methods have a better performance than other existing methods. INDEX TERMS Source camera identification, deep learning, multi-task training, camera categories coupling coding, adaptive filter, auxiliary classifier.

A hybrid data fusion approach with twin CNN architecture for enhancing image source identification in IoT environment

“…The CSI‐CNN methodology entails the identification of the best patch within an image, subsequent calculation of SPN (Sensor Pattern Noise) across all patches, and the inclusion of residual blocks within the network architecture to heighten accuracy. A novel camera attribute classifier, as proposed in Reference 29, employs a recursive method to extract camera features across distinct CNN layers. Multiple deep learning models tailored for original source identification across diverse data sets are juxtaposed and scrutinized in the tabulated presentation delineated in Table 1.…”

Smartphone Camera Identification from Low-Mid Frequency DCT Coefficients of Dark Images

“…A significant number of works employ convolutional neural networks (CNN) for device identification, e.g., [ 18 , 52 , 53 , 54 ]. Source camera identification by brand, model, and device using coupling multitask training based on CNN is discussed in [ 55 ]. Due to low performance for camera model classification, the authors propose an auxiliary classifier used on the local neighborhood differences for the camera lens.…”

Comparative study of DL-based methods performance for camera model identification with multiple databases

“…ii) The triple classification is close to the basic scenario, as the goal is to identify the camera according to a label. But, as the name implies, it classifies cameras according to the brand, the model and the device, as in [8,9]. The triple classification aims to identify cameras based on all labels and is more global.…”

“…This overuse represents an advantage for performance evaluation because the methods can be compared on comparable terms. However, except for a few articles using an equivalent number of camera models (i.e., 27), such as [8], [9] for the triple classification, most methods classify disparate numbers of camera models. This difference is highlighted in the previous subsection as well as in the Tab.…”