Digital Forensic Methods Using Systematic Artifacts of Imaging

摘要：该项目的目标是开发新的数字法医技术，并通过实验验证为法医调查员提供支持。数字取证的目标是以数字形式（包括数字图像、视频或音频）建立证据的起源、完整性、处理历史和意义。这是一个新兴且快速发展的领域，通常使用信号处理、估计和检测与机器学习相结合的方法。由于传感器设计、相机内信号处理和传感器制造过程本身的缺陷，该项目的主要目标是开发依赖于成像传感器系统伪影的法医方法。这些伪影相当于数字传感器指纹。通过检测数字图像中的指纹，可以将给定图像与拍摄该图像的特定相机/传感器联系起来，从而确定图像的来源和完整性，或揭示图像处理的历史。一些研究方向是高度相关和复杂的问题，以前没有解决过，如检测伪造指纹的可能性。其他任务包括以新颖的方式使用现有概念或对现有技术进行实质性改进。其中一种方法是在没有摄像头的情况下，根据源摄像头对图像进行分组。所开发的方法需要在1000多台相机获得的10万张图像上进行大规模实验验证。

Abstract:The objective of this project is to develop novel digital forensic techniques and support it with experimental validation for later technology transfer to forensic investigators. The goal of digital forensics is to establish the origin, integrity, processing history, and meaning of evidence in digital form that includes digital images, video, or audio. This is an emerging and rapidly developing field typically approached using methods from signal processing, estimation, and detection combined with machine learning. The main thrust of this project is development of forensic methods that rely on systematic artifacts of imaging sensors due to sensor design, in-camera signal processing, and imperfections of the sensor manufacturing process itself. These artifacts form an equivalent of a digital sensor fingerprint. By detecting the fingerprint in a digital image, one can link a given image to the specific camera/sensor that took it and establish thus the image origin and integrity or uncover the image processing history. Some of the research directions are highly relevant and complex problems that have not been addressed before, such as the possibility to detect attempts to forge the fingerprint. Other tasks involve using existing concepts in a novel manner or substantially improve upon existing technology. One of them is to group images according their source camera without having the cameras. The developed methods are subject to large scale experimental verification on a hundred thousand images obtained from more than 1000 cameras.