Advanced methods for source authentication and attribution in digital forensics

We live in the information age. Equipped with capable and feature-rich information processing devices that were impossible to imagine just a few years ago, people can now have affordable, fast, and almost constant access to information. Smartphones, online platforms, computers, and broadband internet access have made the internet the de-facto source for all kinds of information, tandem with the people's never-ending ambition for content creation and sharing. These developments brought adverse outcomes as well. Protecting the privacy of data has become an almost impossible feat. "Trusted" online platforms were revealed to be monetizing or found lacking in taking good care of user information, often-times siphoned by malicious parties. On a similar note, opposing parties have had more opportunities as new methods to avoid scrutiny become available for incriminating information. This PhD thesis presents advanced digital forensics methods for source identification and attribution of digital cameras based on "Photo-Response Non-Uniformity" (PRNU) noise. This type of noise is caused by a non-uniform response of the sensor to the incident light. The proposed methods relate to privacy protection as well to the re-attribution of cloaked incriminating information. The first method relates to the protection of privacy. Anonymously published photographs can be clustered and traced back to camera owners with the use of PRNU based source camera identification techniques. Since PRNU is multiplicative, its removal was deemed not possible with known techniques. The proposed method makes the removal possible and provides a way to prevent such harm. The second method relates to the re-identification of images processed with an algorithm called Patch-Match, which was shown to impede PRNU based techniques. The third refers to image transformations. We have presented a method that can compute the reverse transform parameters of such applications using the PRNU noise. Therefore, we can regain the ability to identify the source camera of transformed images. We believe the proposed methods taught in this study will help relieve the negative aspects of this age and serve as a guide for the researchers of the field to develop better digital forensics tools.