Willis H. Ware Damaged 5.25” Floppy Disks

Author: David Tenenholtz

The RAND Corporate Archives serves as the organization’s corporate memory. The Archives collects, preserves, organizes and provides access to materials that document RAND’s story, people and impact from 1948 to the present. Within the archives is the Willis H. Ware Digital Collection (donated by the Ware family in 2014), which contains over three-hundred 5.25” floppy disks from the late 1980s that haven't been touched in thirty years. Dr. Willis Ware (1920-2013), the former head of the Computer Sciences Department at RAND, is considered a luminary in the disciplines of information security, data privacy, artificial intelligence, computer architecture, and much more. Having had a celebrated fifty-five year long career at RAND, this collection was deemed of high value and it was important to have sound methods to capturing and processing the data from the floppies, while also not subjecting the disks to any unintended harm during the process.

The disks themselves are not in good shape. The majority have mold growth and show signs of degradation after sitting in boxes in Ware's garage. After some research into best practices for recovering the data from this floppy disk format, it was established that: 1) floppy disks can shed some of their magnetic iron oxide layer during the disk capture process, resulting in any number of scratches to the surface of the disk, 2) one can slice open a floppy disk enclosure and pull out the magnetic media inside the sleeve to clean it, and (3) cleaning the magnetic media along with cleaning the drive heads repeatedly will yield the least error-ridden disk images.

Using a Device Side Data FC5025 floppy disk controller with a TEAC FD-55GFR disk drive, and a Digital Intelligence Forensic Recovery of Evidence Device (FRED), the overall process can be summarized in these steps:
1. Clean disk drive heads and clean floppy disk twice; wait 10-15 min for drive to dry.
2. Attempt a disk capture using the GUI application Disk Image and Browse. If there is audible squealing and excessive noise from the drive, cancel and clean drive again; wait to dry
3. Attempt more captures via the GUI, and hopefully get a good capture after anywhere from 3 to 6 attempts per disk. If a log file is needed (which allows for a count of read errors), and cancelling the capture job is not needed, do the capture job via the command line rather than using the GUI by running the ‘fcimage’ command.
4. Evaluate results of captures using sdhash in BitCurator. Run sdhash for all captures of a single disk and run a script to generate a “score” that is an average of the sdhash confidence values across all captures of that single disk. A high number close to 100 indicates the capture process yielded similar results; a low score (under 25) indicates that the captures were largely dissimilar from attempt to attempt and likely require more cleaning, or the disk may not be able to be captured reliably.
5. If needed, use Meld visual diff tool to explore file and directory differences. Meld can also be used to assemble access copies of files and directories that serve as a patched version of the original source data.

Laboring over these disks was slow-going, but the process improved over time after switching from cleaning the drive to cleaning the disk media as well as the drive. The process improved again after cleaning the disk media two times prior to an imaging attempt. This is an indication that the disk surface has stubborn residue on it even after the first cleaning, which will impact the quality of the capture.

As a result of carefully examining the procedures for dealing with degraded 5.25” floppy disk media, and applying these cleaning and evaluation methods, most of the disks in the collection of Ware’s floppies now have been successfully captured, often with perfect reads. Once the processing is completed, we will likely have a good representation of many of Ware's digital files during his last decade as Director of the Computer Sciences Department at RAND.

OSF Slide deck