Forensic Image Analysis: Mobile Device and Camera Classification Using EfficientNet and ResNet50 Models

In this article, we propose a novel approach to camera source identification that leverages deep learning and transfer learning techniques to address the growing need for accurate and reliable methods in digital forensics. With the increase of mobile devices and the increasing frequency of image sharing on social media platforms, the risk of image manipulation, forgery, and related cybercrimes has escalated. Traditional methods for camera identification, such as metadata analysis and sensor noise examination, have been foundational but are often vulnerable to tampering. Our research proposes an advanced model combining Convolutional Neural Networks (CNNs) and transfer learning to enhance the accuracy of camera and sensor detection. Utilizing the MICHE-I dataset, our approach demonstrates superior performance, with the EfficientNet model achieving 98.29% accuracy in device classification and the ResNet50 model excelling in sensor-level detection with 96.97% accuracy. This work significantly contributes to the field of multimedia forensics, providing a robust and efficient method for tracing the origin of digital images, which is crucial for verifying the authenticity of content in legal and security contexts.