
Resilient File System (ReFS) is a new file system, released by Microsoft with Windows Server 2012 (supported in Windows 8), with the purpose of supporting extremely large data sets, and in the long term superseding what is currently one of the most widely used file systems; NTFS.
As there was very little documentation and research into ReFS, especially from a computer forensics point of view, this project was undertaken with the intention to investigate the file system and subsequently provide documentation which will be able to help investigators and computing professionals in their understanding of ReFS.
The approach taken was to reverse engineer the behaviour of ReFS, performing experiments on a ReFS formatted drive and then mimicking these on NTFS and FAT drives, enabling me to see if ReFS behaved in a similar way to the other file systems. A hexadecimal dump of a whole drive before the experiment would be taken and then compared to the hexadecimal of the drive after the experiment. This allowed me to examine and analyse the data and how it changed during the experiment, thus building understanding of how it was working.
Many important parts of the on disk structure were analysed as well as how the file system worked when operations were performed on it through the chosen operating system; Windows Server 2012. In addition to this a similar category method of analysis was adopted like Brian Carrier in the book File System Forensic Analysis (2005).
Individual sections of analysis included comparing the MBR and VBRs between the file systems, an investigation into the Recycle Bin and how that worked in ReFS compared to NTFS and FAT and research into the potential meaning of an inclusion of a registry reference, which was new to ReFS.
Other new features were discovered such as the inclusion of FSRS (File System Recognition Structure) in ReFS and the use of 16KB metadata blocks rather than clusters for addressing and storing data. Both of these were very different from the behaviour in the other file systems compared, so good knowledge and information has been uncovered and documented.
Experiments were grouped, some involved folders, some files and some a mix of both. The experiments were performed on all three file systems so the behaviour could be compared. Creating, renaming, adding content, deleting and zipping a folder were some of the experiments performed with folders. Understanding of the file system was developed by understanding what had changed and how, after these actions had taken place. Information such as byte flags and bytes representing permissions were discovered.
Files with different extensions were added to the system to see how these compared. This involved adding .exe, .doc and .txt files, comparing how the file system dealt with each one and what differed. Other actions were performed on each of the files, this included adding content, changing the content and then deleting the file.
Points of interest included how a file's content is stored very much differently in ReFS compared to the method used by NTFS and how everything, from the drive label, to folders and files are stored in the same way.
This report discloses information about ReFS, such as how it stores drive labels and folders, how the MBR is constructed and the general on-disk structure and how it differs from that of the NTFS and FAT file systems.