Temporal forensic analysis of digital Camera sensor imperfections for picture dating

Lead Research Organisation: Northumbria University
Department Name: Fac of Engineering and Environment

Abstract

The research proposal aims to investigate the temporal variations of camera sensor imperfections in order to establish, for any given digital camera, a theoretical model that allows the analyst to estimate the acquisition date of digital pictures. This will advance the field of forensic science, with the potential for applications in high profile cases that require the extraction of evidential information for courtroom purposes. In this project, the analyst is assumed to have the camera device or a set of trusted pictures taken by the same camera and whose acquisition time is known. The goal is then to date one or more images supposedly obtained from the same source and whose acquisition time is unknown. This application will obviously help forensic investigators analyse incidents and link different events.
The novelty of the proposed approach lies in (i) New techniques for efficiently estimating the sensor pattern noise for the purpose of temporal analysis (ii) Classification and temporal analysis of three types of pixels, namely, normal pixels, abnormal but not yet defective pixels, and defective pixels. (iii) Use of machine learning techniques to model the temporal behaviour of camera sensor imperfections.

Planned Impact

This proposed research would be of great help to forensic scientists and investigators in police forces throughout the UK and beyond since the extraction of informaion from digital images can be used for courtroom purposes. It should also have a positive economic impact in future by enabling the leading UK industrial companies to develop internationally competitive products for picture dating as this has not yet been considered in the UK and elsewhere. In a local context, we will transfer the skills and techniques gained during our investigation to other group members in the Conputer and Electronic security Systems Lab at Northumbria University and to our collaborators in this project.

The results of this research project will be disseminated to the research community through publications at key international conferences and in high impact journals. Furthermore, under the main CESS website, a new webpage will be designed and updated regularly during the project for providing up-to-date details, experimental results, and free access to the standard dataset.
 
Description -Digital cameras age with time and because they leave a fingerprint on pictures, it is possible to use this finding to estimate the acquisition date of digital pictures.
-A New camera fingerprint estimation system has been developed. This has beaten the current state-of-the-art techniques and has shown reliable performance for use by forensic analysts and investigators.
-New database for use in temporal image forensics has been developed. This will be presented at a conference and be made available for use by the research community.
Exploitation Route -Automated systems will be developed to estimate the acquisition date of digital pictures.
-The proposed camera fingerprint estimation system has shown reliable performance for use by forensic analysts and investigators. It can also be used as a reliable source of evidence in the courtroom to solve cases and conflicts and enforce the law.
-The new temporal image forensics database will be made available for use by the research community to widen research in this area and promote the field for the purpose of law enforcement, promoting justice and civil security.
Sectors Digital/Communication/Information Technologies (including Software),Government, Democracy and Justice,Security and Diplomacy

 
Description 1- The system developed in this project has been tested and validated in source camera identification. The system can identify or verify the source camera that has been used to take a picture by just processing it. The system has been shown to be more reliable than existing ones in the literature. This could be adopted by digital forensics labs and companies in the courtroom to provide evidence which can solve criminal cases. 2- The last part of the project provides interesting findings about automatic picture dating. Forensics companies have shown interest in taking this further. This is still under discussion.
First Year Of Impact 2018
Sector Digital/Communication/Information Technologies (including Software),Government, Democracy and Justice
Impact Types Policy & public services

 
Title New Image database for Temporal forensics 
Description So far there are more than 20000 images that have been acquired on a regular basis (weekly or bi-monthly) by ten different digital cameras. This includes: 2 x Panasonic TZ20, 2 x Fujifilm S2950, 2 x Nikon L330, 2 x Canon IXUS 115 HS, 2 x Samsung PL120. Two cameras of the same make and model are used to create the intra-model challenge in the numerical experiments. The images are taken in a variety of indoor and outdoor environments, for near and far objects of various textures under uncontrolled conditions to simulate the real world. Accordingly, the images can vary in contents, exposure, optical depth and other settings. Extra attention is paid whilst taking the pictures to fulfill an ethical conduct and avoid breaches of privacy of any kind. The cameras have been tuned to their automatic settings at the highest pixel density. The images are saved in JPEG format. The database will be presented at a conference or journal paper before it is made public on the university website. 
Type Of Material Database/Collection of data 
Year Produced 2019 
Provided To Others? No  
Impact There will be a significant impact in image forensics when the database is made available because it is the first of its kind and to the best of our knowledge there is no such a database available for applications in temporal forensics.