Evaluating the Effectiveness of PRNU-Based Forensic Techniques in Mobile Devices using Peak Correlation Coefficient

In the digital age, this pioneering investigation is the initial endeavor to confront a crucial challenge in image forensics, marking a notable advancement. The implementation of photoresponse non-uniformity (PRNU) and convolutional neural networks (CNNs) has demonstrated remarkable effectiveness in their respective applications, leading to a noteworthy enhancement in forensic techniques for image analysis. The increasing difficulty of manipulating images underscores the need for forensic methods that are both accurate and efficient. This research specifically delves into the combined use of photo-response non-uniformity (PRNU) and convolutional neural networks (CNNs) to boost the efficiency of picture forensics. The methodology involves PRNU analyzing visual noise patterns and comparing them with CNN's expertise in eliminating distinguishing elements to determine the source of an image. Results from the experiments highlight that the integration of PRNU and CNNs achieves an exceptional level of efficiency, evident in a peak correlation coefficient (PCC) value of 0.989 and an accuracy rate of 89.47 percent on the CIFAR-10 dataset. This integrated approach proves beneficial across various forensic applications, including forgery detection, identification of image modifications, and tracking manipulated images. However, limitations arise due to the inherent characteristics of the gadget used in the study. Ongoing research efforts are directed towards establishing more advanced models, incorporating additional forensic techniques, and developing machine learning algorithms. This ongoing work aims to address these limitations and significantly enhance the precision of photo-matching in image forensics.